Crojfe_40-2_2018

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Croatian Journal of Forest Engineering • Volume 39 • Issue 2

324

2018

Croat. j. for. eng. 39(2018)2

Contents Journal for Theory and Application of Forestry Engineering

Croatian Journal of Forest Engineering is a refeered journal distributed internationally, publishing scientific articles concerning forest engineer­ ing, both theoretical and empirical. The journal covers all aspects of forest engineering research, ranging from basic to applied subjects.

From volume 1 to 25 the journal was published under the title »Meha­ nizacija šumarstva«.

Publishers Forestry Faculty of Zagreb University, »Croatian forests« Ltd. Zagreb Croatian Chamber of Forestry and Wood Technology Engineers Copublishers FORMEC Publishing Council Mario Božić, Krunoslav Jakupčić, Vladimir Jambreković, Tibor Pentek, Tomislav Poršinsky (all from Croatia) Editorial Board Igor Anić, Damir Barčić, Ivan Balenović, Saša Bogdan, Zdenko Bogović, Jura Čavlović, Andreja Ðuka, Boris Hrašovec, Josip Ištvanić, Anamarija Jazbec, Ante P. B. Krpan, Josip Margaletić, Slavko Matić, Milan Oršanić, Zdravko Pandur, Ivica Papa, Renata Pernar, Stjepan Risović, Marijan Šušnjar, Damir Ugarković, Dinko Vusić, Željko Zečić, Marko Zorić International Editorial Board Dalia Abbas (USA), Mauricio Acuna (Australia), Stelian Alexandru Borz (Romania), Raffaele Cavalli (Italy), Woodam Chung (USA), Mehmet Eker (Turkey), Jörn Erler (Germany), Fulvio di Fulvio (Norway), Stefano Grigolato (Italy), Mohammad Reza Ghaffariyan (Australia), Hans Rudolf Heinimann (Switzerland), Dirk Jaeger (Germany), Martin Kühmaier (Austria), Matevž Mihelič (Slovenia), Tadeusz Moskalik (Poland), Ljupčo Nestorovski (Macedonia), Igor Potočnik (Slovenia), Hideo Sakai (Japan), Raffaele Spinelli (Italy), Karl Stampfer (Austria), Jori Uusitalo (Finland), Rien Visser (New Zeland) Editor’s Office P.O. Box 422, HR–10 002 Zagreb, CROATIA Tel. + 385 (0)1 235­24­17 Fax. + 385 (0)1 235­25­17 e­mail: crojfe@sumfak.hr Internet: http://www.crojfe.com Editor-in-Chief Tibor Pentek

Croatian Journal of Forest Engineering

•

Volume 39

•

Issue 2, 153–324

•

Zagreb, July 2018

Original scientific papers Okey Francis Obi, Rien Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency using a Multi­Step Data Envelopment Analysis Procedure

153

Eduardo Tolosana, Raffaele Spinelli, Giovanni Aminti, Rubén Laina, Ignacio López-Vicens Productivity, Efficiency and Environmental Effects of Whole­Tree Harvesting in Spanish Coppice Stands Using a Drive­to­Tree Disc Saw Feller­Buncher

163

Piotr S. Mederski, Mariusz Bembenek, Zbigniew Karaszewski, Zenon Pilarek, Agnieszka Łacka Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees 173 Zbigniew Karaszewski, Agnieszka Łacka, Piotr S. Mederski, Mariusz Bembenek Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes

183

Omar Mologni, Peter Dyson, Dzhamal Amishev, Andrea Rosario Proto, Giuseppe Zimbalatti, Raffaele Cavalli, Stefano Grigolato Tensile Force Monitoring on Large Winch­Assist Forwarders Operating in British Columbia

193

Václav Štícha, Jaroslav Holuša, Roman Sloup, Jan Macků, Jiří Trombik A Mobile Hydraulic Winch for Use in Small­Scale Forestry

205

Nopparat Kaakkurivaara, Tomi Kaakkurivaara Productivity and Cost Analysis of Three Timber Extraction Methods on Steep Terrain in Thailand

213

Matevž Mihelič, Raffaele Spinelli, Anton Poje Production of Wood Chips from Logging Residue under Space­Constrained Conditions

223

Milan Marusiak, Jindřich Neruda Dynamic Soil Pressures Caused by Travelling Forest Machines

233

Ahmad Solgi, Ramin Naghdi, Eric R. Labelle, Petros A. Tsioras, Ali Salehi Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density

247

Ehsan Abdi Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran

255

Editor Željko Tomašić

Mohammad Javad Heidari, Akbar Najafi, Seyedjalil Alavi Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression and Artificial Neural Networks

Technical Editor Mario Šporčić

Matija Landekić, Ivan Martinić, Matija Bakarić, Tibor Pentek, Tomislav Poršinsky, Mario Šporčić Current State and Improvement Potential of Forestry Workers Training in Croatia 289

Junior Editor Ivica Papa Editorial Advisor Dubravko Horvat Technical Editorial Board Andreja Ðuka, Zdravko Pandur, Dinko Vusić

Karlo Beljan, Stjepan Posavec, Jura Čavlović, Krunoslav Teslak, Thomas Knoke Economic Consequences of Different Management Approaches to Even­Aged Silver Fir Forests

All published scientific papers have been internationally reviewed Two issues of journal are published annually Circulation: 900 Prepress and Print »Laser plus« Ltd., Brijunska 1a, Zagreb Preparation ended 2018–07–27

299

Alan Antonović, Damir Barčić, Jaroslav Kljak, Josip Ištvanić, Tomislav Podvorec, Juraj Stanešić The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass for Biorefinery Products 313

Linguistic Advisers Maja Zajšek­Vrhovac (for English) Articles are abstracted by or indexed in CAB Abstracts, Compendex, GeoBase, Global Health, Paperchem, Science Citation Index Expanded, SCOPUS, VINITI

271

Cover photo Harvester in a Coniferous Stand (Photo: A. Ðuka) Publishing of this journal is co­financed by funds from Croatian Ministry of Science and Education Subscription: 80 € per year Subscription payment on behalf of: Forestry Faculty of Zagreb University, P.O. Box 422, HR–10 002 Zagreb, CROATIA Swift Code: ZABA HR 2X, Account Number: 2500–03281485, Details of Payment: 2–02–05 Contact: crojfe@sumfak.hr Subscription: 500 HRK/y (Local Payment) Recipient: Faculty of Forestry, University of Zagreb, p.p. 422, HR–10 002 Zagreb Giro account: 2360000–1101340148, Details of Payment: 2–02–05 Contact: crojfe@sumfak.hr

Original scientific paper

Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency using a Multi-Step Data Envelopment Analysis Procedure Okey Francis Obi, Rien Visser Abstract The performance of a harvesting crew in terms of its ability to transform inputs into outputs is influenced by discretionary factors within the unit’s control, such as the selection of machines and operators. However, factors associated with the operating environment, such as terrain slope and tree size that are outside the direct control of management, can also influence harvesting system efficiency. Using data on forest harvesting operations in New Zealand, this paper applies an established four-stage Data Envelopment Analysis (DEA) procedure to estimate the managerial efficiency of independent forest harvesting contractors, while taking into account the influence of the operating environment. The performance of 67 harvesting contractors is evaluated using seven inputs, one output (system productivity) and three operating environment factors in an input-oriented, variable return to scale DEA. The results show that the operating environment including terrain slope, log sorts and piece size influence the efficient use of inputs by harvesting contractors. A significant difference is observed between the mean managerial efficiency of the crews before and after controlling for the influence of the operating environment, the latter being higher by 11%. This study provides evidence that without accounting for the influence of the operating environment, the resulting DEA efficiency estimates will be biased; the performance of crews in favourable operating environment would be overestimated and those in unfavourable environment underestimated.

Keywords: data envelopment analysis, operating environment, forest harvesting, performance evaluation

1. Introduction Data envelopment analysis (DEA) has over the years evolved into a widely accepted research technique that the operations research community is increasingly applying to analyse and improve relative performance of private and public production entities (Liu et al. 2013).

DEA is a non-parametric mathematical programmingbased approach for performance estimation of production or decision making units (DMUs) addressed in Charnes et al. (1978) (Charnes, Cooper and Rhodes model – »CCR«) and extended by Banker et al. (1984) (Banker, Charnes and Copper model – »BCC«). It provides a framework for the estimation of best-practice frontier for production entities involving multiple in-

puts and outputs to allow for benchmarking and performance evaluation (Estelle et al. 2010). The overall efficiency of a production unit can be estimated using the CCR model under the assumption of constant returns to scale, while the technical or managerial efficiency of a unit can be estimated using the BCC model under the assumption of variable returns to scale. DEA classifies production units into efficient and inefficient units based on their selected inputs and outputs by maximizing the ratio between the weighted output and the weighted input (Sharma and Yu 2015). An efficient unit is assigned the maximum efficiency score of 1, while a unit with a score less than one is considered less efficient relative to its efficient peers. DEA is able to estimate the performance of production units in terms

Croat. j. for. eng. 39(2018)2 153

O.F. Obi and R. Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162)

of their ability to either minimize input usage under the production of given output (input orientation) or to maximize output production with given inputs (output orientation) (Li et al. 2017). It is important to note that DEA does not suggest that a unit with a score of 1 is absolutely efficient (operating at optimum output-input ratio), however, by comparing several units’ outputinput ratios (i.e. benchmarking), it can estimate that one or more units are more or less efficient than others (Sherman and Zhu 2006).

Researchers in the field of logging operations have only recently began to apply DEA in estimating performance of forest harvesting operations and it is gaining attention (Obi and Visser 2017a, Hailu and Veeman 2003, LeBel and Stuart 1998). The application of DEA in forest harvesting offers opportunities in examining harvesting efficiency owing to its flexibility, without requiring assumptions about the functional relationships among inputs and outputs, and its invariant nature to units of production factors (Macpherson et al. 2013). The effective application of DEA is based on the assumption that the production units whose performance is being estimated operate within a homogenous production environment (Carrillo and Jorge 2016). However, this assumption in practice does not hold for most harvesting operations as the ability of a harvesting crew/contractor to transform inputs into outputs is not only affected by discretionary inputs (i.e. controllable by the management) or managerial skills. It is also influenced by exogenous factors such as terrain slope, roughness or tree size (otherwise referred to as the operating environment) that are beyond direct managerial control (Obi and Visser 2017b, Aalmo and Baardsen 2015). These factors provide either a favourable or an unfavourable operating environment to the crews. An unfavourable operating environment would demand additional inputs from the production unit to produce the same level of output as a unit in a favourable environment in order to overcome the external disadvantage making the unit’s efficiency to be underestimated (Hu et al. 2011). This has been identified as a major problem in DEA studies as most performance studies do not account for differences in the operating environment of production units (Carvalho and Marques 2011, Fried et al. 2008).

In forest harvesting where operations are carried out in complex and unstructured operating environments (Di Fulvio et al. 2017), factors exogenous to harvesting crews’ control are likely to either positively or negatively influence the performance of harvesting operations. For example, steep terrain or terrain hindrance is expected to be more difficult for ground-based harvesting systems in terms of machine trafficability as opposed to flat or rolling terrain. As such, a relatively

154

e­ fficient crew in a harvest operation with high degree of terrain hindrance may be labelled as inefficient when benchmarked against another in an operation with low level of terrain hindrance. Without adequately controlling for exogenous factors, efficiency estimates in DEA will most often be biased as inefficiencies are assumed to be attributable to managerial skills (Macpherson et al. 2013). The managerial efficiency of units in adverse or unfavourable operating environments could be underestimated, conversely those in favourable environments could be overestimated (Yang and Pollitt 2009); thus potentially leading to inefficient allocation of resources. Accounting for differences in the operating environment of independent forest harvesting contractors is critical for objective and unbiased assessment of performance among harvesting crews.

There is an established four-stage DEA procedure developed by Fried et al. (1999) which is able to account for the factors that are not in direct control of the harvesting crews. However, existing studies on the application of DEA in the forest harvesting sector have so far focused on assessing performance without considering non-discretionary inputs, i.e. inputs beyond the managers’ control. Obi and Visser (2017a) examined the operational efficiency of 423 independent forest harvesting contractors in New Zealand over a period of 7 years using DEA. The authors considered five inputs, namely, number of harvest days, number of machines, total harvest area, number of log sorts and total volume of timber, and one output – system productivity in the production model. They reported that majority of the harvesting contractors operate at or near scale efficiency level, but the source of inefficiency in the industry is both technical and managerial.

They added that investment in technology and human capital could improve the overall efficiency of the industry. LeBel and Stuart (1998) applied DEA models to measure the technical efficiency of 23 fully mechanized loggers in Canada during the period 1988–1994.

The aggregate, technical, and scale efficiencies of the loggers were evaluated based on DEA models with capital, consumables, and labour as the inputs and tons of wood as output. They reported majority of the contractors to be efficient. Factors identified as influencing the technical efficiency of the loggers include low capacity utilization and scale of harvesting operations. Hailu and Veeman (2003) using panel data covering a period of 19 years (1977–1995) analysed the technical efficiency, technical change and total factor productivity in the logging industries for six boreal provinces in Canada using DEA. The study reported substantial technical efficiency differentials among the provinces. The authors identified some region-specific variables that influenced efficiency of logging Croat. j. for. eng. 39(2018)2

Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162) O.F. Obi and R. Visser

­ perations in the regions. The variables include forest o density, proportion of hardwood production, scale of operation measured as production per establishment, and engineering construction per area harvested.

Literatures on performance evaluation that account for the effects of exogenous factors on the efficiency of production entities in different industries exist (Zhu et al. 2016, Ferrera et al. 2014, Macpherson et al. 2013). There is however no literature controlling for the effects of the operating environment on efficiency estimates of forest harvesting operations. The objective of this study therefore, is to measure the managerial efficiency of independent forest harvesting contractors in New Zealand taking into account the effect of differences in their operating environment.

This removes the environment bias, and the resulting performance estimates are attributable purely to managerial efficiency. This is accomplished by applying the four-stage DEA procedure. This study extends the previous work of Obi and Visser (2017a) by introducing the operating environment factors in the performance evaluation procedure.

technical efficiency (Farrell 1957) defined as a measure of efficiency under the restriction that a linear combination of efficient units produces the same or more of all outputs and that the reduction in inputs is equiproportional. The efficiency scores are estimated without regard to the exogenous factors. This establishes a best-practice frontier for the harvesting crews based on the inputs and outputs included in the DEA. However, the efficiency estimates of crews operating in »good« operating environment are overestimated and that of the harvesting crews in »harsh« or »difficult« operating environments are underestimated. An input-oriented DEA framework with variable returns to scale (Banker et al. 1984) is adopted in the first DEA stage and can be represented by the following expression (Cordero-Ferrera et al. 2011):

Fried et al. (1999) developed an empirical technique termed the four-stage DEA procedure to separate managerial inefficiency from other inefficiency components beyond managerial control. The fourstage DEA procedure rests on the premise that production units operating in relatively unfavourable environments may be wrongly labelled as inefficient (Hu et al. 2011, Yang and Pollitt 2009). This procedure is able to control for the exogenous operating environment factors by compensating for their effects, and has been applied in previous literatures (Zhu et al. 2016, Ferrera et al. 2014, Yang and Pollitt 2009). Data on the original production factors are modified according to the effects of the exogenous factors, and the modified data are used for the final performance evaluation thus, providing a pure measure of managerial efficiency. The procedure is briefly described here so that the reader can follow the process through to the results. For detailed description of the four-stage DEA procedure, readers are referred to Fried et al. (1999).

2.1.1 Stage one DEA In the first stage, following a standard production theory set under variable returns to scale, a DEA production frontier is estimated using selected inputs and outputs for the production units which in the case of this study are independent forest harvesting contractors. The DEA estimator is used to estimate the Farrell

∑ ∑

Subject to

n

∑λ x

j ij

+ si− = θ 0 xij0

(1)

j =1

2. Methodology 2.1 The four-stage DEA procedure

m  s  Minθ 0 − ε  sr+ + si−    =  r 1=i 1 

n

∑λ y

j rj

+ − sr0 = yrj0

j =1

n

∑λ = 1 j

j =1

λj , si− , sr+ ≥ 0, i = 1, 2,… , m; r = 1, 2,… , s; j = 1, 2,… , n

Where: vector of inputs for unit j xij yrj vector of outputs for unit j q0 efficiency score e infinitesimal non-Archimedean constant lj weightings sr– inputs slacks sr+ outputs slacks.

2.1.2 Stage two The second stage is to estimate N input equations using an appropriate econometric method such as Tobit regression. The dependent variables are total input slacks (radial plus non-radial slack) estimated from the first stage DEA, while the independent variables are measures of the external operating environment. This quantifies the effect of the exogenous factors as it affects the excessive use of inputs so they can be adjusted accordingly. The radial input slack represents the reduction in the inputs of a relatively inefficient DMU

Croat. j. for. eng. 39(2018)2 155

O.F. Obi and R. Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162)

were it to operate efficiently beyond which no further reduction in inputs is possible without reducing output; whereas the non-radial slack represents the potential additional reduction in the inputs of a relatively inefficient DMU after proportionally reducing its current inputs to become efficient (Fried et al. 1999). The slack arise from two distinguishable effects: the technical inefficiency of the units and the influence of the exogenous factors which this approach aims to decompose and make adjustments on the original input values (Cordero et al. 2009). The sign of the coefficients estimated in the regressions provide information about the direction of the effects of the exogenous factors on each total input slack which may vary from one slack to another including in significance. Tobit regression is applied in this study, and has been applied in previous studies (Hung and Shiu 2014, Macpherson et al. 2013, Hu et al. 2011, Avkiran 2009, Fried et al. 1999). The N input equations are specified as follows:

(

)

ITSjk = f j Qjk , β j , ukj , j = 1,….., N ; k = 1,….., K (2) Where: ITSjk unit k’s total slack for input j based on the DEA efficiency estimates from the first stage Qjk vector of variables characterizing the external environment for unit k that may affect the utilization of input j bj vector of coefficients ujk disturbance term.

Eq. 4 creates a new dataset for each production unit wherein the inputs are adjusted for the influence of the operating environment. The maximum predicted slack is used to establish a base equal to the least favourable set of external conditions; thus a unit with external factors generating lower level of predicted slack would have its input adjusted upwards to put it on the same level with the unit operating in the least favourable environment. By increasing the unit’s input and leaving the output unchanged, its performance is purged of any advantage offered by its favourable operating environment.

2.1.4 Stage four DEA The fourth and final stage re-runs the DEA (Eq. 1) under the initial input–output production specification and generates new measure of efficiency by using the adjusted input dataset free from the influence of the operating environment. The new efficiency scores provide a measure of the efficiency that is attributable purely to managerial skills.

2.2 Dataset

This study uses a dataset on individual contracted harvesting operations (involving mechanized felling, extraction, processing of stems and loading out onto trucks) obtained from a large commercial forest company in New Zealand. The dataset contains detailed information on harvesting crews, stand, terrain, cost, harvesting system and productivity factors on harvesting operations from January 2016 to March 2017. The 2.1.3 Stage three data was collected at individual-contract level in order to capture the operating environment specific to each The third stage uses the estimated parameters from harvesting operation. Thus, it is able to capture the the second stage regression (Tobit regression) to pretrue reflection of the effect of the exogenous factors on dict new total input slack for each input and for each input requirement for the operations. The data were production unit based on the operating environment collated from the different regions of New Zealand factors applicable to that unit: amounting to a total of 67 entries on harvesting opk  ˆ k= erations executed by independent forest harvesting

(3) ITS f Q , β , j = 1, … .., N ; k = 1, … .., K j j j j crews. Due to the confidentiality agreement binding on the data, information on the identity of the harvestThe predicted total input slacks are used to adjust ing contractors are not provided; each harvesting conthe primary input data for each unit according to the tractor is assigned a unique identifier for ease of referdifference between maximum predicted slack and the ence. All the harvest operations were clear-fell in New predicted slack for each input: Zealand Radiata pine plantations. k adj ˆ k − ITS ˆ k,j = xj x kj +  Max k ITS 1, .., N ; k 1,….., K = … = j j    2.3 Production and exogenous factors

(

)

{

}

{ }

k  k ˆ k ˆ k 1,….., N ; k = 1,….., K = j +  Max ITSj − ITSj  , j =  

Where: value of unit k’s adjusted jth input xjk adj xjk value of unit k’s primary jth input Maxk{ITSjk} maximum predicted slack for unit k

156

(4)

Previous studies on performance evaluation in the forest harvesting sector have employed a variety of input–output factors. Based on available data and relevant literatures (Li et al. 2017, Obi and Visser 2017a, Visser and Spinelli 2012,Visser et al. 2011, Amishev et al. 2009), this study selects seven inputs, one output and three exogenous factors for the performance Croat. j. for. eng. 39(2018)2

Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162) O.F. Obi and R. Visser

e­ valuation of the harvesting crews. The factors are considered to practically reflect the harvesting process, considering the available data.

Input factors: These are factors over which the harvesting contractors have some level of control and they include (i) Number of workers (NMWOK) – this the average number of workers in a crew engaged in the harvesting operation of a defined forest area over the entire harvesting period; (ii) Number of machines (NMMCH) – defines the total number of machines deployed for a harvesting operation; (iii) Harvest days (HDAYS) – this is the total number of days of harvesting by a crew in a defined forest area; (iv) Net stocked area (NETAREA) – being the total actual harvest area size measured in hectares; (v) Total recoverable volume (TREVOL) – is the actual volume of stem harvested from a defined forest area measured in tonnes per hectare; (vi) Landings size (LNDSIZE) – this is the total landing size for a harvesting operation estimated from the product of average landing size and number of landings, and is measured in hectares; and (vii) Average haul distance (AVHUD) – this is the mean extraction haul distance measured in meters, and is obtained from the operational harvest plan.

Output factor: System productivity (SYSPROD) measured in tonnes per scheduled machine hour Table 1 Descriptive statistics of the factors for performance evaluation (N=67) Factors

Mean

SD

Min.

Max.

Inputs NMWOK

6.2

2.5

2

18

NMMCH

5.2

1.9

2

13

HDAYS

65.7

41

12

206

NETAREA, ha

32.2

27.3

5.6

153.8

TREVOL, tons/ha

555

125

298

902

LNDSIZE, ha

0.84

0.52

0.06

2.4

AVHUD, m

256.6

227.3

0

1937

11.2

9.6

59.5

Output SYSPROD, tons/SMH

31.7 Exogenous

AVSLOP

18.6

7.7

11

39.3

LGSORT

11.6

2.1

7

17

PESIZE, ton/stem

1.4

0.5

0.5

3.1

SD – Standard deviation

(tons/SMH) is considered the output of the harvest operations and is calculated as the total volume of harvested timber from a defined forest area divided by the total harvest time.

Exogenous factors: These are exogenously fixed factors within the operating environment of the harvest crews over which they do not have direct control. Three factors are identified as exogenous factors for the purpose of this study and they include (i) terrain slope (AVSLOP) – this the average slope of the harvested forest area measured in degrees, (ii) log sorts (LGSORT) – this is the number of log sorts from a defined forest area contracted to a harvesting contractor; and (iii) piece size (PESIZE) – is defined as the average piece size from a harvest area measured in ton/stem. Table 1 presents the descriptive statistics of all the factors.

2.4 Analysis Efficiency scores for the harvesting crews described in terms of the technical efficiency are estimated using DEAP software version 2.1 which also estimates radial and non-radial slacks for each production factor using a multi-stage process (Coelli 1996). Technical efficiency refers to the ability of a unit to utilize its limited inputs to produce the desired outputs and it is influenced by the use of technology (Coelli et al. 2005).

The number of production units in a DEA should at least be twice the number of inputs and outputs combined (Golany and Roll 1989) as a large number of inputs and outputs combined compared to the number of units diminishes the discriminatory power of DEA (Cook et al. 2014). This study has 67 production units (harvesting crews) and 8 inputs/output.

3. Results and discussion 3.1 First stage DEA without exogenous factors The first stage DEA results presented in Table 2 shows a large variation in efficiency estimates of the harvesting contractors. The mean efficiency score for the contractors is 0.79, theoretically suggests that the crews are currently operating at about 79% efficiency of their current input levels. Conversely, on average a harvest crew could reduce its current input usage by approximately 21%, were it to perform on the efficient frontier. A total of 18 crews (27%) are estimated as efficient, i.e. efficiency score = 1, while 14 crews (20%) have efficiency scores of 0.8 to 0.99. Most, 43% (N=29) are estimated to have efficiency scores of 0.60 to 0.79 (i.e. 60 to 79%). However, operations of independent harvesting contractors are often influenced by operating environment factors outside the control of the

Croat. j. for. eng. 39(2018)2 157

O.F. Obi and R. Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162)

crews (Obi and Visser 2017b, Hoffmann et al. 2016, Aalmo and Baardsen 2015). Crews operating in difficult environments may find it challenging to equal the performance of their counterparts in more favourable operating environment.

Table 2 Stage one efficiency scores statistics (N=67) Statistics

Efficiency rankings

N

% of crews

Mean

0.794

100%

18

27

SD

0.158

80–99%

14

21

Median

0.784

60–79%

29

43

Min.

0.519

40–59%

6

9

Max.

1

–

–

–

SD – Standard deviation

3.2 Second stage analysis In the second stage, total input slacks representing potential input saving for each of the inputs is regressed against the set of exogenous factors (independent variables) namely, average slope, log sorts and piece size using Tobit regression. There are seven regression models, one for each input slack. The parameters estimated are presented in Table 3. A positive exogenous factor coefficient on a total input slack suggests that the factor constitutes an unfavourable environment resulting in excess use of the input by the harvest crews; the reverse being the case for a negative

coefficient. In other words, an operating environment with a positive (negative) coefficient on a total input slack is associated with the inefficient (efficient) use of the input, and the sign and statistical significance can differ across the inputs (Fried et al. 1999). Consequently, an operating environment with a positive coefficient on an input slack tends to reduce harvesting efficiency as its measure increases, and vice versa for an operating environment with a negative coefficient.

As shown in Table 3, average slope (AVSLOP) has a positive coefficient on all the input slacks but its effect is significant only on the number of workers (NMWOK), number of machines (NMMCH) and average haul distance (AVHUD) slacks. Its positive coefficient on all slacks can be attributed to the enormous challenge it presents to forest harvesting operations irrespective of the system of harvesting adopted.

Number of log sorts (LGSORT) has a negative coefficient on all the total input slacks except on AVHUD slack, and it is significant on NMWOK and total recoverable volume (TREVOL) slacks. This suggests an increase in log sorts is favourable to the efficient use of all the inputs in the production model except AVHUD. Log sorts, thus can be said to improve harvesting efficiency as it increases. This makes practical sense in that harvest operations in New Zealand with high log sorts are usually associated with large forest areas, and is often characterized by high system productivity.

Piece size on the other hand has an insignificant positive coefficient on NMWOK slack and a significant positive coefficient on TREVOL slack. The coefficient is negative and insignificant on all other input slacks.

The significant positive coefficient of piece size on

Table 3 Estimation results of total input slacks using Tobit regression. Standard errors are shown in brackets Dependent variables, slacks Regressor

Constant AVSLOP, o LGSORT PESIZE, ton/stem Log-Likelihood

NMWOK

NMMCH

HDAYS

NETAREA, ha

TREVOL, tons/ ha

LNDSIZE, ha

AVHUD, m

5.36

3.21

78.8

28.5

201

0.58

–207

(2.35)

(1.73)

(43.9)

(24.2)

(112)

(0.47)

(219)

0.12**

0.07*

0.99

0.31

1.44

0.012

9.90*

(0.04)

(0.03)

(0.86)

(0.47)

(2.19)

(0.01)

(4.23)

–0.52**

–0.22

–2.85

–0.36

–18.3*

–0.03

15.0

(0.18)

(0.13)

(3.28)

(1.80)

(8.43)

(0.04)

(16.3)

0.23

–0.42

–21.3

–11.8

69.0*

–0.05

–70.1

(0.68)

(0.50)

(12.9)

(7.07)

(32.2)

(0.14)

(63.5)

–132

–117

–287

–256

–337

–55.3

–365

*significant at 95%, **Significant at 99%

158

Croat. j. for. eng. 39(2018)2

Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162) O.F. Obi and R. Visser

TREVOL is understandable in that for a given tree stand, increased piece size is expected to result in increased total recoverable volume. The varying effects of the exogenous factors on the input slacks justifies the need to correct the initial DEA scores for the influence of the factors. Otherwise, the impact of the operating environment on harvesting operations may consistently result in estimating crews in »good« operating environments as more efficient than those in »harsh« environments. In practical terms, the results present some insights as to the direction of the effects of these exogenous factors on the usage of harvesting inputs thus providing some guide as to the inputs that should be carefully managed under certain operating environments in order to improve overall harvesting efficiency.

3.3 Third stage analysis The estimated regression parameters presented in Table 3 are used in the third stage analysis to predict a new set of total input slacks for each of the crews according to the factors characterizing their operating environment (Eq. 3), and also to adjust the initial input data for each crew according to Eq. 4. The maximum predicted slack is used to set a baseline for the least favourable operating environment (Fried et al. 1999).

A crew with a predicted total input slack less than this value for an input will have its corresponding input factor adjusted upward. Table 4 presents a summary statistics of the adjusted inputs for the harvesting contractors. It can be seen that the mean value for each of the adjusted inputs (Table 4) is higher than its corresponding original mean value presented in Table 1.

This is because the adjusted input data controls for the influence of the three exogenous factors considered in this study, thus giving no advantage or disadvantage to any crew owing to a favourable or unfavourable operating environment in terms of input usage.

Table 4 Summary statistics of the adjusted input factors of the harvesting contractors

Mean

SD

Min.

Max.

NMWOK

9.6

2.2

4.8

19.2

NMMCH

7

1.7

4.5

14

HDAYS

98.3

41.6

36.5

231.3

NETAREA, ha

44.2

26.8

14.9

162.4

TREVOL, ton/ha

704

109

500.5

1003

LNDSIZE, ha

1.1

0.5

0.2

2.8

438.7

216.8

198.9

1950

AVHUD, m SD – Standard deviation

the adjustment (stage 1), 18 of the 67 contractors (27%) were efficient – 100% efficiency score (Table 2) and after the adjustment (stage 4) 23 crews (34%) were estimated to be efficient (Table 5). The mean and minimum efficiency estimates in stage four DEA also show that efficiency estimates are higher after adjusting for exogenous factors. The results indicate that it is important to include the effect of exogenous factors in the performance evaluation of harvesting operations.

A smaller variation in performance among the crews is observed as evident in the lower standard deviation of the performance estimates in the stage four DEA results (Table 5) compared to the stage 1 results (Table 2). The decrease in standard deviation reflects an overestimation of the performance of units in favourable conditions and an underestimation of Table 5 Stage four estimated efficiency score statistics (N=67)

3.4 Final stage DEA with exogenous factors The fourth and final stage of the approach is to rerun the DEA based on the initial input-output specification using the adjusted input data. This produces new efficiency estimates for the contractors attributable purely to managerial skills void of the influence of the operating environment factors considered in the analyses. Descriptive statistics of the results of the final stage DEA adjusted for the influence of the operating environment is presented in Table 5. Adjusting the inputs for the effect of exogenous factors on the performance of the harvesting crews results in an increase in the number of crews estimated as efficient, and in the number crews in the 80–99% efficiency ranking. Before

Statistics

Variables

Statistics

Efficiency rankings

N

% of crews

Mean

0.90

100%

23

34

SD

0.095

80–99%

32

48

Median

0.915

60–79%

12

18

Min.

0.68

–

–

–

Max.

1

–

–

–

Constant

19%

–

–

Increasing

78%

–

–

Decreasing

3%

–

–

Returns to Scale

SD – Standard deviation

Croat. j. for. eng. 39(2018)2 159

O.F. Obi and R. Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162)

those in more challenging environments in the first stage DEA. The average efficiency score increased by approximately 11% (79.4% to 90%) after controlling for environment effects on the efficiency score.

Approximately 19% of the crews operate under constant returns to scale while 78% operate under increasing returns to scale. This suggests that majority of the harvesting crews possess the capacity to improve their system productivity. It is important to note that a harvesting contractor estimated to be efficient (i.e. efficiency score = 100%) based on the four-stage DEA technique applied in this study does not interpret to mean that it has reach its maximum production efficiency or capacity. The DEA efficiency estimate of 1 assigned to the contractor means that among its peers based on their current input utilization and production outputs, the contractor outperformed its peers and can act as a benchmark for others in improving their managerial efficiency. The high percentage of the contractors operating under increasing returns to scale suggests the existence of opportunities to improve input utilization efficiency and consequently improve overall harvesting efficiency.

To statistically establish a difference between stages 1 and 4 DEA efficiency estimates, the Mann-Whitney U-test is applied. The Mann-Whitney U-statistics reject the null hypothesis of equality of the first and fourth stage efficiency scores (p-value = 0.0001). This implies that there exists a significant difference in the managerial efficiency of the harvesting contractors adjusted and unadjusted for differences in the operating environment. The slack adjusted new efficiency estimates represent potential minimum reduction in inputs if a crew operated in the worst environment and performed up to the efficient frontier (Fried et al. 1999).

The overall increase in the mean efficiency score in the fourth stage DEA suggests that crews in difficult ­operating environment exhibit better management skills but were adjudged poorly in the first stage DEA. In summary, including the operating environment ­factors in performance evaluation does make a signi­fi­ cant difference in the final technical efficiency estimates in forest harvesting operations.

4. Limitations of the study and future research Although this study achieved its objective of measuring impartially the technical efficiency (managerial efficiency) of forest harvesting contractors including quantitative environment factors, it presents some limitations worth acknowledging. The production model for the forest harvesting operations incorpo-

160

rated only seven inputs, one output and three environment factors. These factors are not exhaustive and was limited largely by availability of data. It would be interesting to incorporate additional factors, where data are available, in future studies including those endogenous to harvesting crews such as training, years in business, operator age, etc. The study did not consider statistical noise which is another phenomenon capable of influencing performance (described as the impact of good luck and bad luck), omitted variables and other related phenomena (Fried et al. 2002). Statistical noise is reflected in a random error term in stochastic frontier analysis-based performance evaluation of production units. This is left as a future line of study in performance evaluation within the forest harvesting industry.

5. Conclusions The four-stage DEA approach proposed by Fried et al. (1999) is applied in this study to account for the effect of non-discretionary factors, often exogenously fixed, on the performance of independent forest harvesting contractors. The very few DEA studies on performance within the harvesting sector have focused simply on estimating performance in terms of efficiency without taking into account the possible influence of the operating environment. The four-stage DEA approach simultaneously adjusts inputs factors to control for the operating environment factors and produces efficiency index attributable purely to managerial skills removing the bias introduced by the operating environment. Using data on forest harvesting operations contracted to 67 harvesting crews in New Zealand, this study demonstrates that benchmarking performance of harvesting crews without accounting for differences in the operating environment will lead to biased, inaccurate and misleading estimates. Significant difference (p<0.01) was observed between the mean managerial efficiency estimates unadjusted and adjusted for the effect of the operating environment with a mean increase of about 11% indicating the impact of the operating environment factors considered in this study. Previous studies also reported a difference between the mean efficiency score unadjusted and adjusted for the effect of the operating environment ranging from 10–23% (Macpherson et al. 2013, Kontodimopoulos et al. 2010, Wang et al. 2006, Fried et al. 1999). The study provides some useful decision support to forest companies, policymakers, and general industry stakeholders involved in the measurement and overall improvement of forest harvesting performance.

Croat. j. for. eng. 39(2018)2

Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162) O.F. Obi and R. Visser

Acknowledgments The authors are most grateful to the forest company in New Zealand and its managers who provided the data used in this study.

6. References Aalmo, G.O., Baardsen, S., 2015: Environmental factors affecting technical efficiency in Norwegian steep terrain logging crews: a stochastic frontier analysis. Journal of Forest Research 20(1): 18–23.

Amishev, D., Evanson, T., Raymond, K., 2009: Felling and bunching on steep terrain – a review of the literature. FFR Technical Note 1(7): 1–10.

Avkiran, N.K., 2009: Removing the impact of environment with units-invariant efficient frontier analysis: an illustrative case study with intertemporal panel data. Omega 37(3): 535–544.

Banker, R.D., Charnes, A., Cooper, W.W., 1984: Some models for estimating technical and scale inefficiencies in data envelopment analysis. Management Science 30(9): 1078–1092.

Carrillo, M., Jorge, J.M., 2016: A multi objective DEA approach to ranking alternatives. Expert Systems with Applications 50: 130–139.

Carvalho, P., Marques, R.C., 2011: The influence of the operational environment on the efficiency of water utilities.

Journal of Environmental Management 92(10): 2698–2707.

Charnes, A., Cooper, W.W., Rhodes, E., 1978: Measuring the efficiency of decision making units. European Journal of Operational Research 2(6): 429–444.

Coelli, T., 1996: A guide to DEAP version 2.1: a data envelopment analysis (computer) program. Centre for Efficiency and Productivity Analysis, University of New England, Australia. Available at: http://www.uq.edu.au/economics/cepa/ deap.htm (Accessed on August 11, 2016).

Coelli, T.J., Rao, D.S.P., O’Donnell, C.J., Battese, G.E., 2005: An introduction to efficiency and productivity analysis.

Springer Science & Business Media, 275 p.

Cook, W.D., Tone, K., Zhu, J., 2014: Data envelopment analysis: prior to choosing a model. Omega 44: 1–4.

Cordero-Ferrera, J.M., Crespo-Cebada, E., Murillo-Zamorano, L.R., 2011: Measuring technical efficiency in primary health care: the effect of exogenous variables on results. Journal of Medical Systems 35(4): 545–554.

Cordero, J.M., Pedraja, F., Santín, D., 2009: Alternative approaches to include exogenous variables in DEA measures: a comparison using Monte Carlo. Computers and Operations Research 36(10): 2699–2706.

Di Fulvio, F., Abbas, D., Spinelli, R., Acuna, M., Ackerman, P., Lindroos, O., 2017: Benchmarking technical and cost factors in forest felling and processing operations in different global regions during the period 2013–2014. International Journal of Forest Engineering 28(2): 94–105.

Estelle, S.M., Johnson, A.L., Ruggiero, J., 2010: Three-stage DEA models for incorporating exogenous inputs. Computers and Operations Research 37(6): 1087–1090.

Farrell, M.J., 1957: The measurement of productive efficiency. Journal of the Royal Statistical Society 120(3): 253–290.

Ferrera, J.M.C., Cebada, E.C., Zamorano, L.R.M., 2014: The effect of quality and socio-demographic variables on efficiency measures in primary health care. European Journal of Health Economics 15(3): 289–302.

Fried, H.O., Lovell, C.K., Schmidt, S.S., 2008: The measurement of productive efficiency and productivity growth. Oxford University Press Inc., New York, 638 p.

Fried, H.O., Lovell, C.K., Schmidt, S.S., Yaisawarng, S., 2002: Accounting for environmental effects and statistical noise in data envelopment analysis. Journal of Productivity Analysis 17(1–2): 157–174.

Fried, H.O., Schmidt, S.S., Yaisawarng, S., 1999: Incorporating the operating environment into a nonparametric measure of technical efficiency. Journal of Productivity Analysis 12(3): 249–267.

Golany, B., Roll, Y., 1989: An application procedure for DEA. Omega 17(3): 237–250.

Hailu, A., Veeman, T.S., 2003: Comparative analysis of efficiency and productivity growth in Canadian regional boreal logging industries. Canadian Journal of Forest Research 33(9): 1653–1660.

Hoffmann, S., Jaeger, D., Lingenfelder, M., Schoenherr, S., 2016: Analyzing the efficiency of a start-up cable yarding crew in Southern China under new forest management perspectives. Forests 7(9): 1–33.

Hu, J.L., Lio, M.C., Yeh, F.Y., Lin, C.H., 2011: Environmentadjusted regional energy efficiency in Taiwan. Applied Energy 88(8): 2893–2899.

Hung, C.L., Shiu, P.J., 2014: Evaluating project performance by removing external effects: implications to the efficiency of research and development resource allocation. Research Evaluation 23(4): 366–380.

Kontodimopoulos, N., Papathanasiou, N.D., Tountas, Y., Niakas, D., 2010: Separating managerial inefficiency from influences of the operating environment: an application in dialysis. Journal of Medical System 34(3): 397–405.

LeBel, L., Stuart, W., 1998: Technical efficiency evaluation of logging contractors using a nonparametric model. Journal of Forest Engineering 9(2): 15–24.

Li, L., Hao, T., Chi, T., 2017: Evaluation on China’s forestry resources efficiency based on big data. Journal of Cleaner Production 142: 513–523.

Liu, J.S., Lu, L.Y., Lu, W.M., Lin, B.J., 2013: Data envelopment analysis 1978–2010: a citation-based literature survey. Omega 41(1): 3–15.

Macpherson, A.J., Principe, P.P., Shao, Y., 2013: Controlling for exogenous environmental variables when using data

Croat. j. for. eng. 39(2018)2 161

O.F. Obi and R. Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency ... (153–162) envelopment analysis for regional environmental assessments. Journal of Environmental Management 119: 220–229.

Obi, O.F., Visser, R., 2017a: Operational efficiency analysis of New Zealand timber harvesting contractors using data envelopment analysis. International Journal of Forest Engineering 28(2): 85–93.

Obi, O.F., Visser, R., 2017b: Influence of the operating environment on the technical efficiency of forest harvesting operations. International Journal of Forest Engineering 28(3): 140–147.

Visser, R., Spinelli, R., Magagnotti, N., 2011: Landing characteristics for harvesting operations in New Zealand. International Journal of Forest Engineering 22(2): 23–27.

Visser, R., Spinelli, R., 2012: Determining the shape of the productivity function for mechanized felling and fellingprocessing. Journal of Forest Research 17(5): 397–402.

Wang, F., Hung, W., Shang, J., 2006: Measuring pure managerial efficiency of international tourist hotels in Taiwan. The Service Industries Journal 26(1): 59–71.

Sharma, M.J., Yu, S.J., 2015: Stepwise regression data envelopment analysis for variable reduction. Applied Mathematics and Computation 253: 126–134.

Yang, H., Pollitt, M., 2009: Incorporating both undesirable outputs and uncontrollable variables into DEA: the performance of Chinese coal-fired power plants. European Journal of Operational Research 197(3): 1095–1105.

Sherman, H.D., Zhu, J., 2006: Service productivity management: improving service performance using data envelopment analysis (DEA). Springer Science & Business Media, 328 p.

Zhu, W., Yang, X., Preston, J., 2016: Efficiency measurement of bus routes and exogenous operating environment effects on efficiency. Transportation Planning and Technology 39(5): 464–483.

Authors’ addresses:

Received: May 8, 2017 Accepted: January 1, 2018

162

Okey Francis Obi, M.Eng.* e-mail: profran6@gmail.com Assoc. prof. Rien Visser, PhD. e-mail: rien.visser@canterbury.ac.nz University of Canterbury New Zealand School of Forestry 8041 Christchurch NEW ZEALAND * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish Coppice Stands Using a Drive-to-Tree Disc Saw Feller-Buncher Eduardo Tolosana, Raffaele Spinelli, Giovanni Aminti, Rubén Laina, Ignacio López-Vicens Abstract Whole tree harvesting was conducted on two coppice stands with different tree composition (Q. ilex and Q. pyrenaica) in gentle terrain. Felling and bunching were performed by a drive-to-tree wheeled feller-buncher with disc saw head. Operations were analyzed on 17 plots 25x25 m2 in order to develop productivity models and to assess operational costs. The study also aimed at determining biomass collection efficiency and evaluating the impact of the new harvesting method on the soil, the remaining trees and stumps. The treatment consisted in a strong coppice thinning leaving standards. Productivity ranged from 2.8 to 4.6 odt/pmh in the Q. ilex coppice, and from 0.9 to 2.6 in the Q. pyrenaica stand. Tree species, dry weight per tree and percentage of removed basal area were the main independent variables affecting productivity. Approximately 50% of the standards showed damages. Most wounds were light, caused by the drive-to-tree work pattern, followed through GPS tracking. Soil damage was also light; in no plots, deep disturbances were found. However, most of the stumps were damaged. Forwarding and chipping productivity and cost were also evaluated. The slash left on the terrain averaged 3.0 and 1.5 odt/ha in Q. ilex and Q. pyrenaica, respectively, including scrub debris. As a conclusion, while this heavy feller-buncher can be useful in coppice heavy thinnings with larger trees, it would be a good option to try lighter disc saw felling heads mounted on the harvester boom tip, which probably would reach better productivity and reduce the frequency of stand damage.

Keywords: Whole tree harvesting system, Quercus ilex, Quercus pyrenaica, harvesting damages, operational cost, work study

1. Introduction Coppice management has been abandoned in many European countries after WWII because of the social and economic transformation of European society, which has made traditional coppicing practices less profitable (Carvalho et al. 2017). Abandonment resulted in the underutilization of the large coppice forest resources and in the loss of biodiversity ­(Müllerová et al.

2015). The densification and aging of forests have made them more vulnerable to disturbances such as storms or wildfires, which have significantly increased during the

past century (Schelhaas et al. 2003) and are estimated to increase by almost one million m3/y in Europe by 2030 (Seidl et al. 2014).

Coppice thinning would reduce wildfire suppression costs, especially if whole-tree harvesting (WTH) is adopted, because complete biomass removal reduces potential fire severity compared with other harvesting methods that leave large amounts of slash within the stand (Corona et al. 2015).

In Spain, coppice forests cover roughly 4 M ha and represent 20 % of the total forest area; Holm oak (Quercus

Croat. j. for. eng. 39(2018)2 163

E. Tolosana et al.

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

ilex) and Melojo oak (Q. pyrenaica) are the dominant forest species in Spanish coppice forests (Piqué and Vericat in Nicolescu et al. 2017).

The use of site-appropriate equipment and techniques is crucial to the financial sustainability of the whole supply chain (Enache et al. 2015). This is even more important for coppice forests, where operators must cope with small stem size and stump crowding, which increase operational costs (Spinelli et al. 2017a).

Coppice harvesting technology is evolving toward increased mechanization and larger and more efficient equipment (Spinelli et al. 2016). The growing mechanization level leads to higher productivity and lower unit costs for woody products from coppice forests (Laina et al. 2013). Moreover, increasing the mechanization level in forest operations contributes to reducing both the severity and the frequency of accidents and/or occupational diseases (Albizu et al. 2013).

Harvesting of coppice forests is technically and economically difficult, due to the difficulty encountered by a harvester head when approaching stems that are gathered in a clump on the same stump (Schweier et al. 2015). Also, the undesired potential effects of mechanization – damage to soil, residual stand, stumps and sprouting ability – have raised concern among forest managers and scientists (Pyttel et al. 2013, Spinelli et al. 2017b), and must be considered when implementing a mechanized harvesting technology.

One of the available technologies for mechanized felling consists of a feller-buncher head equipped with a disk saw. This technology has been tried with good results in SRC (short rotation coppice) by Iwarsson (2008). Its advantage lies in the high cutting speed and in the ability to manage multiple stems in a single pass.

This type of felling head has been tried recently in Mediterranean coppices, where it proved less effective than in SRC but highly capable to contain stump damage, when compared with shears (Schweier et al. 2015).

In 2017, the Spanish forest company SOMACYL began the field trial of a drive-to-tree disc saw fellerbuncher for use in coppice harvesting, which provided an ideal opportunity for conducting carefully designed time and motion studies for evaluating operational productivity, cost, product recovery and site damage.

The main goals of the present study were as follows: Þ developing productivity models based on significant explanatory factors and use the models to assess the operational costs Þ evaluating the impact of the new harvesting method on the soil, the remaining trees and stumps

164

Þ determining biomass collection efficiency (percent of total available biomass actually recovered) and biomass retention (i.e. amount of biomass left on the terrain).

2. Material and methods The selected base machine was a 130 kW John Deere 643J articulated carrier with a total mass of 12.7 tonnes, equipped with a felling head JD FD45, with 51 cm cut capacity, a 0.64 m2 accumulation capa­ city and a total weight of 2.2 tonnes. The felling head was mounted on the carrier front lift, so that the machine had to drive towards each of the tree clumps targeted for felling (i.e. drive-to-tree feller buncher).

The machine was tested on two separate coppice stands, one dominated by holm oak (Quercus ilex L.) and the other by melojo oak (Quercus pyrenaica Willd.).

Both sites were measured and characterized before and after harvesting.

2.1 Pre-harvest inventory Seventeen 25x25 m2 plots were randomly distributed across each forest site, 9 in the Q. ilex stand and 8 in the Q. pyrenaica coppice. On each plot, the diameter at breast height (DBH) of all the trees was measured, and the trees were marked with color paint. The limits of the plots – N–S and E–W lines – were marked with colored plastic tape and painted wooden poles at their corners. The silvicultural treatment was performed around the plots prior to the time study, so that the machine operator could work in close-to-real conditions within the plots. This work was planned together with the operation managers working for the enterprise concerned, SOMACYL.

2.2 Post-harvest inventory Just after the mechanized felling and piling, bunches were counted and the number of trees per bunch was estimated. The DBH and height of 10 felled trees per plot was determined, and 3–4 additional trees were weighed in order to fit the weight table. A sample was taken and weighed for moisture determination.

The DBH of all residual trees in the original 25x25 m plots was also determined.

2.3 Characterization, height-DBH equation and weight table fitting Treatment characteristics were obtained from the comparison of the pre- and post-harvest inventories.

Height-DBH equations and weight tables were fitted using standard statistic software Statgraphics Centurion Croat. j. for. eng. 39(2018)2

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

XVII. The selected dependent variable for the weight tables was dry weight, after estimating the moisture content of samples with the gravimetric method, according to ISO standard 18134-3:2015.

2.4 Time study and production evaluation A cycle-level time study was conducted on the feller-buncher, using a Husky Hunter hand-held field computer running the dedicated Siwork3 time study software (Kofman 1995). A cycle was defined as the time to process a single bunch. Productive time was separated from delay time.

To assess the production from each plot, the forwarder piled the whole trees from the plot in a separate roadside bunch that was marked and, afterwards, chipped and transported to the plant to be weighed and sampled for determination of moisture content.

An independent time-study of a complete work shift was performed outside the studied plots, in order to obtain a more reliable utilization factor than could be gained from the work on single 25x25 m2 plots.

During the forwarder extraction, a whole day shift was time-studied measuring the number of trips and the loaded bunches to get an approximate estimation of the extraction productivity and cost.

The machines were rented by the contracting enterprise SOMACYL. The actual hourly rental cost – or the unit cost in the case of chipping and chip transport - were also recorded.

2.5 Damage assessment To characterize soil and stand damage, an inventory for damages and stump status assessment was performed after the end of the extraction operation.

Damage to residual trees was determined by inspecting all remaining trees inside each of the 25x25 m2 plots, following the methodology proposed by Tavankar et al. (2013) that classifies the damages to the remaining trees according to their location, size and intensity, as reflected in Table 4.

Soil damage was determined with the method proposed by McMahon (1995), who proposed soil damage classes according to litter and/or topsoil removal and, in case of rutting, depending on rut depths, as reflected in Table 5. Observations were conducted inside circular sub-plots with a radius of 4 m, centered in the diagonal crossing point of each of the 25x25 m2 plots.

Researchers also counted the stumps within these same subplots, measuring their heights and evaluating their status.

E. Tolosana et al.

2.6 Collection efficiency measurement Inside these circular sub-plots, all the biomass left on the terrain was weighed, and a sample for moisture determination was collected in order to estimate the oven dry weight of the biomass. By doing so, it was possible to determine biomass retention and collection efficiency.

3. Results 3.1 Stand characterization The holm oak (Quercus ilex L.) coppice had an average initial density of 5257 trees/ha, a mean DBH of 5.7 cm, a mean height around 4 m, and an initial basal area of 13.3 m2/ha. The number of stools/ha was 956, with an average number of shoots per stool of 5.0. There were also 498 isolated Q. ilex oaks per ha. The treatment resulted in the removal of over 90 % of the trees and 70% of the basal area, and left 443 residual trees per ha.

The melojo oak (Quercus pyrenaica Willd.) coppice had an initial density of 4168 trees/ha, and a mean DBH of 7.1 cm, a mean height around 6 m, and a basal area of 16.8 m2/ha. The number of stools/ha was 1078, with an average number of shoots per stool of 2.9. There were also 1025 isolated oaks per ha. The treatment resulted in the removal of 85% of the trees and 55% of the basal area, and left 603 residual trees per ha.

Total harvest ranged between 29 and 77 fresh tonnes/ha, for the Q. Ilex stand and between 13 to 37 fresh tonnes/ha for the Q. pyrenaica coppice. Corresponding dry weights were 22–56 odt/ha (mean 36 odt/ha) for Q. Ilex and 9–29 odt/ha (mean 17 odt/ha) for Q. pyrenaica.

3.2 Height-DBH equation and Dry weight table To fit the height-DBH equation, 94 Q. ilex and 91 Q. Pyrenaica trees were measured. To build the dry weight table, 31 Q. ilex and 30 Q. pyrenaica oaks were measured and weighed, using different sample trees than those used for the height-DBH curve.

For the weight table, DBH and height were selected as independent variables, but the results showed the weak significance of height as an explanatory variable, so finally only DBH was selected as the main independent variable.

The best fit corresponds to the equations shown in Fig. 1.

3.3 Time study In the Q. ilex stand, the productivity of felling and bunching ranged between 2.7 and 4.8 odt/pmh (oven

Croat. j. for. eng. 39(2018)2 165

E. Tolosana et al.

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

Fig. 1 Height, fresh and dry weight curves for the studied oaks as a function of diameter at breast height (DBH) dry tonnes per productive machine hour). Delays were mostly absent from the felling time in the plots, because no incidents occurred during the study conTable 1 Ranges of productivity and explanative variables

Range

166

Species

DW/tree, kg

%ExtractedAB

Prod.

ODT/PMH

Q ilex (Holm oak = 1)

7.4–15.9

51–88

2.7–4.9

Q. pyrenaica (Holm oak = 0)

9.9–22.5

21–69

0.9–2.9

ducted there. The average productivity of mechanized felling was 4.0 odt/pmh inside the experimental plots.

In the Q. pyrenaica stand, the productivity of felling and bunching ranged between 0.9 and 2.9 odt/pmh.

Delays accounted for 5% of total worksite time. The average productivity of mechanized felling was 1.8 odt/pmh inside the experimental plots.

Besides the short-time study inside the plots, a longer time-motion study was conducted outside the experimental plots in order to get a better estimate of machine utilization. This longer study covered a full shift and showed that delays (including daily preparation Croat. j. for. eng. 39(2018)2

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

E. Tolosana et al.

Table 2 Productivity model for the feller-buncher in the studied coppices Multiple regression – Prod odt/PMH Dependent variable: Prod odt/PMH Explicative variables Dry W/tree %ExtractedBA Holm oak (1/0) Estimation

Standard error

t-statistic

p-value

Constant

Parameter

–1.66

0.83

–2.001

0.0667

%ExtractedBA

0.0464

0.009

5.110

0.0002

Dry W/tree

0.105

0.038

2.752

0.0165

Holm oak

1.105

0.283

3.902

0.0018

ANOVA Source

Squares sum

DF

Aver. square

F-ratio

p-value

Model

21.817

3

7.27233

39.31

0.0000

Residual

2.40526

13

0.18502

Total (Corr.)

24.2223

16

R-square = 90.1% R-square (adjusted by d. of. f.) = 87.8% Estim. standard error = 0.43 Average absolute error = 0.26 odt/hProd Durbin-Watson Coefficient = 2.47

and maintenance activity) accounted for 10% of total worksite time.

species is shown in Table 1, while the equation para­ meters are shown in Table 2.

The forwarder used to extract the whole trees was a 186 kW John Deere 1910E, with a load capacity of 19 tonnes. The forwarder was equipped with a compressing load deck designed to facilitate handling of bulky loads (Dutch Dragon PC-48). The forwarder was studied during a complete shift (8.7 pmh). Forwarder productivity was 7.0 odt/pmh. The incidence of delay time was 12%, including daily maintenance.

Prod(odt/PMH) = –1,66 + 0.0464 %ExstractedBA + 0.105 dryW/tree + 1.105 Holm oak (1/0)

3.4 Feller-buncher productivity model

3.5 Cost estimation

Multiple regression analysis was performed using the data from the 17 studied plots. The analysis tested the impact on productivity (dependent variable) deriving from the following independent variables: Tree species, Dry weight per tree – initial stand, Dry weight per extracted tree, Extracted dry weight per hectare, Initial number of trees per hectare, Extracted number of trees per hectare, Extracted basal area and Percentage of extracted basal area. Significant variables were Species (introduced as a dummy variable, with Q. ­pyrenaica as the baseline and Q. Ilex as the dummy), Dry weight per tree and Percentage of extracted basal area. The range of the variables for the two studied

The renting cost of the machines was established on an hourly basis (€/smh) for felling and bunching, and on a fresh tonne basis (€/fresh tonne) for chipping and chip transport to the power plant (transportation distance = 80 km, one way).

Unit cost estimates were based on the following assumptions: average moisture content of the chips produced in the study equal to 25 and 22%, respectively, for Q.ilex and Q.pyrenaica (wet basis) and machine utilization equal to 90%. The results of the calculations are reported in Table 3.

Total delivered cost was 68 €/odt for Q. ilex chips and 104 €/odt for Q. pyrenaica. If these figures were

(1)

Using the average productivity and the average removals, the required felling and bunching time in hours per hectare was estimated as 11.3 pmh (12.6 smh)/ha for Q. ilex and 11.9 pmh (13.2 smh)/ha for Q. pyrenaica.

Croat. j. for. eng. 39(2018)2 167

E. Tolosana et al.

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

Table 3 Average operational unit costs based on renting costs Average productivity, odt/pmh

Average unit cost – chipper renting & transport €/fresh tonne

Average unit cost, €/odt

Renting hourly cost, €/smh

Renting hourly cost, €/pmh

Q. ilex

Q. pyrenaica

90

100

3.18

1.45

–

71.5

79.4

6.99

6.99

–

Chipping

–

–

–

–

11.0

14.80

14.14

Chip transport (dist. = 80 km)

–

–

–

–

7.66

10.31

9.85

Total (direct cost)

–

–

–

–

–

67.92

104.32

+ 15% indirect and fixed costs

–

–

–

–

–

78.11

119.98

Operation

Felling and bunching Forwarding

inflated by 15% to reflect indirect and fixed costs, then the actual cost would increase to 78 €/odt for Q. ilex and 120 €/odt for Q. pyrenaica.

Regarding the influence of the significant independent variables on this cost, if the productivity eq. 1 (Table 2) were corrected using the ratio between the average, it would become: Prod(odt/PMH) = –1.31 + 0.0367 %ExstractedBA + 0.083 dryW/tree + 0.873 Holm oak (1/10)

(2)

And the unit cost of felling and bunching: Unit cost(€/odt) = 100/[–1.31 + 0.0367 %ExtractedBA + 0.083 dryW/tree + 0.873 Holm oak (1/0)] (3) Therefore, the total operational costs – without considering any revenues - in the average conditions for each of the studied stand types can be transformed in cost per hectare after adjusting for the different ­removals (36 odt/ha for Q. ilex and 17 odt/ha for Q. pyrenaica). The result would be 2820 €/ha for Q. ilex and 2076 €/ha for Q. pyrenaica.

3.6 Environmental effects The frequency and severity of residual tree wounding are shown in Table 4, separately for the two stand types. Different letters (a and b) for the two species show a statistically significant difference. The results indicate a greater level of damage in the Q. ilex stand, with special reference to crown damage (broken branches higher than 1 meter above the ground level).

Fortunately, injuries were mostly small and medium sized (surface smaller than 200 cm2). The main cause for such injuries was machine movement (bumping against the trees), while the accidental contact with the disc saw accounts for less than 10% of total wounding.

Soil and stump damages are summarized in Table 5. Soil damage was not severe (equal or less than 5%

168

Q. ilex

Q. pyrenaica

31.45

68.97 11.36

of the total surface showed disturbance deeper than 5 cm, even when the machine is moving all over the plot surface). Stump height was lower than the prescribed 10 cm in most of the cases (two thirds in Q. pyrenaica, three quarters in Q. ilex). Around 10% of the stumps were taller than that, but within the 20 cm mark, and only 10% exceeded this limit. Most of the stumps were severely damaged. In the Q. pyrenaica stand, over half of the stumps were split or fragmented, and the proportion increased to 70% in the Q. ilex stand.

3.7 Biomass collection efficiency Regarding collection efficiency, the actual harvest (chips dry weight) was between 70 and 90% of the estimated table weight for Q.ilex and Q.pyrenaica, respectively.

The slash left on the terrain averaged 3.0 odt/ha in Q. ilex and 1.5 odt/ha in Q. pyrenaica, including scrub debris.

4. Discussion The applied silvicultural operations were delayed – as both coppices were older than 35 years – traditional »mixed coppice« treatment, leaving a greater number of residual trees than in the »coppice with standards« traditional silviculture (Short and Campion 2014).

The studied harvesting system (multi-tree handling WTH) is one of the main trends for the mechanization of fuelwood harvesting (Erber et al. 2017).

The selected heavy drive-to-tree disc saw fellerbuncher has not been previously studied in the harvesting of Mediterranean coppice stands, despite the fact that powerful felling heads with a wide opening are considered necessary for work in coppice forests (Chakroun et al. 2016).

Croat. j. for. eng. 39(2018)2

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

E. Tolosana et al.

Table 4 Damages affecting the remaining trees, % Damage conditions, remaining trees, % Type

Qi

Qp

Bark

25.8 a 19.7 a

Wood

4.7 a

1.6 a

Broken branches

12.9 a

3.3 b

Destroyed

0.7 a

1.0 a

Total

44.1 a 25.6 b

Severe

5.4 a

2.6 a

Location

Qi

Qp

Height

18.0 19.0 a a

Bole

Low (0–0.30 m)

Crown

17.3 a

5.9 b

Medium (0.30–1.0 m)

Roots

0.4 a

0.0 a

High (>1.0 m)

Qi

Qp

Size

Qi

Qp

Small (<50 cm2)

16.1 a

7.3 b

8.9 Medium 14.1 a (50–200 cm2) a

7.5 b

Sawdisk injuries

8.6 a

3.6 a

10.1 a

Other

0.0 a

0.0 a

10.6 8.2 a a 7.8 a

16.8 7.8 a b

Large (>200 cm2)

5.5 a

Cause

Qi

Qp

Machine 91.4 movements a

96.4 a

Qi = Quercus ilex Qp = Quercus pyrenaica Different letters (a/b) show statistically significant differences for p=95%)

Table 5 Soil damages and stump condition after the treatment Soil damage, % of total surface No damage evidence

Litter still in place, evidence of minor disruption

Litter removed, topsoil exposed

Litter and topsoil mixed, <5 cm depth

Quercus ilex

6.1

51.7

41.7

0.5

Quercus pyrenaica

0.0

56.2

38.8

5.0

Specie

Stump height, % of stumps number

Stump status, % of stumps number

<10 cm

10–20 cm

>20 cm

No damages

Bark slightly removed

>50 % bark removed

Cracked stump

Destroyed stump

Quercus ilex

76.5

19.4

4.1

1.8

7.2

10.8

69.4

10.8

Quercus pyrenaica

66.7

23.3

10.0

3.9

16.1

22.8

48.3

8.9

Specie

The DBH-to-height and DBH-to-weight models show that Q. ilex is less slender than Q. pyrenaica, and is considerably shorter for equal DBH (Fig. 1). Nonetheless, Q. ilex has a much more developed crown than Q. pyrenaica (Ruiz-peinado et al. 2012), and this is the reason why, despite having less height, Q. ilex oaks do have more weight than Q. pyrenaica oaks for the same DBH (Fig. 1). This fact can explain why one of the most significant factors explaining productivity is the species, because Q. ilex trees are bigger and allow reaching a higher productivity than Q. pyrenaica oaks (1.1 odt/PMH more, over the whole range of other explanatory variables).

As a matter of fact, tree size (DBH, tree volume, stump mass or tree mass), is an established explanatory variable in most felling and bunching productivity

studies (Spinelli et al. 2007, Schweier et al. 2015, Erber et al. 2016, Chakroun et al. 2016, Spinelli et al. 2016).

Another very common explanatory variable for felling productivity in coppice operations is the removal (Spinelli et al. 2016), which is reflected in this study by the percent of the basal area.

The productivity of the disc saw felling head, tested in traditional Spanish coppice, is comparable with the productivity figures reported in other similar studies conducted on coppice stands across the Mediterranean region (Laina et al. 2013, Schweier et al. 2015).

In fact, the figures reported here for Q. ilex might be slightly higher than those found in the literature, if one takes into account the very small tree size (average DBH smaller than 6 cm) and the selective character of

Croat. j. for. eng. 39(2018)2 169

E. Tolosana et al.

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

the treatment, which resulted in the release of a dense residual stand.

The present prices per whole tree chips tonne for a moisture content of 25 and 22% in the Spanish market are 54 and 58 €, corresponding to equivalent prices per dry tone of 72 and 74 €, respectively (SOMACYL 2018).

In these conditions, the operation would not be economically sustainable, although one would get close to breaking even with Q. ilex. Using the present technology, the possibilities to achieve profitability would be with treating bigger sized stands – which are not usually available, except in the case of very old coppice forests – or by prescribing larger removals, leaving fewer residual trees, or none at all.

The difference between costs and revenues in the studied conditions would amount to a net loss of 221 €/ha for the Q. ilex stand and of 796 €/ha for the Q. pyrenaica stand.

Frequent residual tree wounding is partly due to the use of a large drive-to-tree machine in a dense residual stand. The working method resulted in the machine trafficking most of the plot surface (average length of the GPS track was 370 m inside the 625 m2 plots). Crown damage was more frequent with Q. ilex due to the specific tree architecture. In any case, only 5% of the damage in the Q. Ilex stand – and even less in the Q. pyrenaica coppice – reached deeper than the bark and/or were greater than 200 cm2.

Soil damage was not severe, despite the intense machine traffic. Both stands were located on flat terrain, on dry sandy soils - and the weather was dry for the whole duration of the study. However, it is reasonable to expect higher soil impact levels on steeper and/ or wetter ground conditions.

Stump damage was quite frequent, as is common when mechanized felling is introduced to coppice (Spinelli et al. 2017b), but the consequence of stump damage on stump mortality and sprouting vigor is still unclear: several studies indicate that stump damage may have no negative effects on sprouting (Pyttel et al. 2013, Spinelli et al. 2017b).

Finally, the different biomass recovery efficiency between the two oak species may be related to different tree architecture: Q. ilex trees have compact crowns with thicker branches, which may minimize handling losses. In contrast, Q. pyrenaica has weaker and wider crowns, which may drop branches if the handling is rough. However, the difference between estimated mass and actual removal in the Q. pyrenaica coppice is too high, and it is not consistent with a relatively small quantity of biomass left on the terrain. Therefore, collection efficiency figures must be impacted by the in-

170

evitable estimation error of the DBH-to-weight models, and therefore these values must be considered as wide approximations.

5. Conclusions Two DBH-to-height curves and DBH-to-weight table were developed for the two different coppice types, which are the most common in Spain. Although approximate, these curves provide an important tool for estimating biomass availability in these stands.

The comparison between the curves shows that Q. ilex is less slender than Q. pyrenaica and has a denser and wider crown, ultimately offering a higher biomass yield for the same DBH. For this reason, coppice thinning is significantly more productive in Q. ilex coppices than in Q. pyrenaica.

Feller-buncher productivity has been modeled as a function of tree species, tree weight and percentage of the extracted basal area. The model shows that productivity increases with stem size and removal intensity.

The analysis of the unit costs, based on the renting cost of the machines, the transport cost to a power plant located at 80 km from the coppice forest, and considering a 15% of indirect and fixed costs, shows that harvesting these stands with this system results in a small financial loss, particularly in Q. pyrenaica stands: therefore, selection thinning in small-tree coppice stands is only justified by specific management goals (e.g. fire proofing) and should be subsidized.

On the other hand, one could try to restore the financial performance of these operations by reducing machine cost through the use of lighter and cheaper machines - for instance by installing a smaller disc saw model on a mini-excavator such as skid-steer loader or a mini-excavator, the latter capable of swingto-tree operation (Spinelli and Nati 2009). Another possible opportunity for cost reduction could be the direct management of the operations – instead of renting the machines – and finally a reduction of the transportation distances.

Site damage was moderate but use of a heavy drive-to-tree feller buncher can cause a higher impact level than expected from a lighter swing-to-tree machine.

Most of the stumps were severely damaged, and even if the negative effect of stump damage on stump sprouting and shoot growth has not been ascertained, it would be safer to follow-up the regeneration on the study sites.

Croat. j. for. eng. 39(2018)2

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

Acknowledgements The authors are indebted to Rubén García and the operational managers and technicians Alberto Fernández-López and Carlos Martínez-Torres in SOMACYL public forest company in Castilla y León (Spain), whose help made possible the present study. Financial support for this research was provided by La Caixa through Fundación Patrimonio Natural de Castilla y León and also by the COST Action FP1301 (www.eurocoppice.uni-freiburg.de) within the scope of its STSM program.

6. References Albizu, P.M., Tolosana, E., Roman-Jordan, E., 2013: Safety and health in forest harvesting operations – Diagnosis and preventive actions - A review. Forest Systems 22(3): 392–400.

Carvalho, J., Magagnotti, N., Nicolescu, V.N., Ruch, P., Spinelli, R., Tolosana, E., 2017: Active management of traditional coppice forests: An interface between silviculture and operations. Eurocoppice COST Action report, Ed: Univ.

Freiburg, 4 p.

Chakroun, M., Bouvet, A., Ruch, P., Montagny, X., 2016: Performance of two shear heads for harvesting biomass in hardwood stands in France. Biomass and Bioenergy 91: 227–233.

Corona, P, Ascoli, D., Barbati, A., Bovio, G., Colangelo, G., Elia, M., Garfì, V., Iovino, F., Lafortezza, R., Leone, V., Lovreglio, R., Marchetti, M., Marchi, E., Menguzzato, G., Nocentini, S., Picchio, R., Portoghesi, L., Puletti, N., Sanesi, G., Chianucci, F., 2015: Integrated forest management to prevent wildfires under Mediterranean environments. Annals of Silvicultural Research 39(1): 1–22.

Enache, A, Kühmaier, M., Visser, R., Stampfer, K., 2015: Forestry operations in the european mountains: A study of current practices and efficiency gaps. Scandinavian journal of forest research 31(4): 412–427.

Erber, G., Holzleitner, F., Kastner, M., Stampfer, K., 2017: Effect of multi-tree handling and tree-size on harvester performance in small-diameter hardwood thinning. Silva Fennica 50(1): article id 1428, 17 p.

Iwarsson, M., 2008: Techniques and methods for small trees harvesting. In Suadicani, K., Talbot, B. (Eds), 2008: The Nordic-Baltic conference on forest operations: Copenhagen September 23–25. Forest & Landscape, University of Copenhagen, 39 p.

Kofman, P., 1995: User Guide. Danish Forest and Landscape Research Institute, Vejle, Denmark, 37 p.

Laina, R., Tolosana, E., Ambrosio, Y., 2013: Productivity and cost of biomass harvesting for energy production in coppice natural stands of Quercus pyrenaica Willd. in central Spain.

Biomass and Bioenergy 56: 221–229.

Mc Mahon, S., 1995: Accuracy of two ground survey methods for assessing site disturbance. International Journal of Forest Engineering 6(2): 27–33.

E. Tolosana et al.

Müllerová, J., Hédl, R., Szabó, P., 2015: Coppice abandonment and its implications for species diversity in forest vegetation. Forest Ecology and Management 343(1): 88–100.

Piqué, M., Vericat, P., 2017: Spain. In: Nicolescu, V. N., Bartlett, D., Buckley, P., Rossney, D., Pyttel, P., Unrau, A. (Eds.), 2017: National Perspectives on Coppice from 35 EuroCoppice Member Countries. Eurocoppice COST Action report, Ed: Univ. Freiburg, 70–71.

Pyttel, P.L., Fischer, U.F., Suchomel, C., Gärtner, S.M., Bauhus, J., 2013: The effect of harvesting on stump mortality and re-sprouting in aged oak coppice forests. Forest Ecology and Management 289: 18–27.

Ruiz-Peinado, R., Montero, G., del Rio, M., 2012: Biomass models to estimate carbon stocks for hardwood tree species.

Forest Systems 21(1): 42–52.

Schweier, J., Spinelli, R., Magagnotti, N., Becker G., 2015: Mechanized coppice harvesting with new smallscale fellerbunchers: Results from harvesting trials with newly manufactured felling heads in Italy. Biomass and Bioenergy 72: 85–94.

Schelhaas, M.J., Nabuurs, G.J., Schucks, A., 2003: Natural disturbances in the European forests in the 19th and 20th centuries. Global Change Biology 9(11): 1620–1633.

Seidl, R., Schelhaas, M.J., Rammer, W., Verkerk, P.J., 2014: Increasing forest disturbances in Europe and their impact on carbon storage. Nature Climate Change 4(9): 806–810.

Short, I., Campion, J., 2014: Coppice-with-standards: An old silvicultural system with new potential? Forestry & Energy Review 4(1): 42–44.

SOMACYL, 2018: Personal communication of the operational responsible of biomass supply from the Company.

Spinelli, R., Cuchet, E., Roux, P., 2007: A new feller-buncher for harvesting energy wood: Results from a European test programme. Biomass and Bioenergy 31(4): 205–210.

Spinelli, R., Nati, C., 2009: A low-investment fully mechanized operation for pure selection thinning of pine plantations. Croatian Journal of Forest Engineering 30(2): 89–97.

Spinelli, R., Cacot, E., Mihelic, M., Nestorovski, L., Mederski, P., Tolosana, E., 2016: Techniques and productivity of coppice harvesting operations in Europe: a meta-analysis of available data. Annals of Forest Science 73(4):1125–1139.

Spinelli, R., Magagnotti, N., Schweier, J., 2017a: Trends and perspectives in coppice harvesting. Croatian Journal of Forest Engineering 38(2): 219–230.

Spinelli, R., Pari, L., Aminti, G., Magagnotti, N., Giovannelli, A., 2017b: Mortality, re-sprouting vigor and physiology of coppice stumps after mechanized cutting. Annals of Forest Science 74(5): 5.

Tavankar, F., Majnounian, B., Bonya, A.E., 2013: Felling and skidding damage to residual trees following selection cutting in Caspian forests of Iran. Journal of Forest Science 59(5): 196–203.

Croat. j. for. eng. 39(2018)2 171

E. Tolosana et al.

Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish ... (163–172)

Authors’ addresses: Prof. Eduardo Tolosana, PhD. * e-mail: eduardo.tolosana@upm.es Rubén Laina, PhD. e-mail: ruben.laina@upm.es Ignacio López-Vicens, Eng. e-mail: ignacio.lopez@tbforest.com Technic University of Madrid C/José Antonio Novais 10 Campus Ciudad Universitaria 28040 Madrid SPAIN

Received: April 25, 2018 Accepted: June 29, 2018

172

Raffaele Spinelli, PhD. e-mail: spinelli@ivalsa.cnr.it Giovanni Aminti, PhD. e-mail: aminti@ivalsa.cnr.it CNR – IVALSA Via Madonna del Piano 10 I-50019 Sesto Fiorentino (FI) ITALY * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees Piotr S. Mederski, Mariusz Bembenek, Zbigniew Karaszewski, Zenon Pilarek, Agnieszka Łacka Abstract Harvester use in broadleaves has recently become more effective economically. However, difficulties with delimbing have shown that not all harvesting heads are suitable and efficient for broadleaved species. The typical obstacles are mainly large tree sizes, bends and forks in the trunks and large branches. For these reasons, it is difficult to obtain specific log lengths according to the settings in the harvester on-board computer. The objective of the research was to determine: 1) the accuracy of the log lengths from the bottom, middle and top parts of oak trees, and 2) harvester efficiency in the utilisation of the trunk for logs. The research was carried out on 61-year-old oaks from which logs with an expected length of 250 cm were processed.

To achieve this length, a margin of error was set in the harvester computer with minimum and maximum lengths of 252 and 257 cm. For thinning operations, a Ponsse Ergo harvester with a H7 harvesting head was used. After harvesting, manual log measurements were carried out on 280 logs: 69, 142 and 69, from bottom, middle and top parts of the trees, respectively.

The largest share of assortments satisfying the minimum requirement of 250–257 cm was obtained from the middle part of the trees (93%), followed by bottom logs (91%) and top logs (88%). The highest frequency of logs, which were too short, were found to be the top logs (9%), while bottom logs were most often too long (6%); therefore, different length settings should be applied to limit such inaccuracies. Analysis of the last log from the highest part of the tree indicated a strong goodness of fit between the top diameter and the DBH; the mean value of the top diameter was 13.3 cm over bark.

Keywords: oak (Quercus robur L.), calibration, cut-to-length, trunk use for logs

1. Introduction Since the 1990s, the increasing amount of harvested wood has been one of the factors influencing the expansion of mechanized harvesting in Poland. In 2017, the harvesting of merchantable timber in the State ­Forests National Forest Holding (SF NFH) aimed to achieve 40.5 million m3, a considerable increase ­compared to the 15.9 million m3 harvested in 1990 (Forestry 2006). The increase in forest growth leads to higher timber production and forest utilisation, changing unfavourable proportions of stands in all age classes (Borecki et al. 2016). Such conditions have been recognised as attractive by forest contractors, who popularised cut-to-length (CTL) technology, and in

2016, approximately 530 harvesters were in operation in Poland (Mederski et al. 2016b). The harvesters were able to cut and process about 30% of the available timber with highly effective thinning operations in pine stands (Mederski et al. 2016a). Among all forest species, Scots pine (Pinus sylvestris L.) is the most common in Poland, covering 60% of the land area, while among broadleaves, oak (Quercus robur L. and Quercus petraea (Matt.) Liebl.) claims the biggest share in growing stock, amounting to 8.1% and 6.8% of area and volume, ­respectively (The National Forest Inventory 2015).

When discussing mechanised harvesting, tree species cannot be ignored. The original harvesters were built to cut and process coniferous trees, mainly spruce

Croat. j. for. eng. 39(2018)2 173

P.S. Mederski et al.

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

in lowland areas. With time, the scope of operations of these machines was broadened to include steep terrains (Stampfer 1999, Frutig et al. 2007) as well as broadleaved stands (Martin et al. 1996 as cited by ­Sinneau and Cuchet 2001, Spinelli et al. 2002, Spinelli et al. 2009, Mederski 2013). The use of harvesters for broadleaved species makes the machine more universal in Central Europe or Baltic countries, where mixed forests are common, or where a larger share of broadleaves has been introduced (Mederski et al. 2009). In Central Europe, harvester relocation from one plot to another, exclusively among coniferous forest, very often lengthens trips, leading to increased costs. Mixed and broadleaved stands available to the harvester fill these gaps in machine relocation and contribute to better machine utilisation and, due to better logistics, lower fuel consumption and CO2 emissions. However, the mechanised harvesting of broadleaved trees has resulted in problems associated with delimbing ­(Labelle et al. 2016), taller stumps in coppice stands (Spinelli et al. 2017), but also larger shavings of the bark and the lateral surface of the processed assortments (Karaszewski et al. 2016). Furthermore, it seems that particular broadleaved species are more favourable for harvesters. In Estonia and Latvia, birch and alder cover approx. 40% of the stands; in Estonia, 95% of the timber is cut by harvesters in final fellings (80% in thinnings), while in Latvia, 70% of the wood is harvested with CTL technology (Moskalik et al. 2017). Not only has the wider use of harvesters been observed in recent years (Mederski et al. 2016b), but there has also been a higher annual use of CTL in European countries (Spinelli et al. 2011, Malinen et al. 2016). It is presumed that the current wider use and higher effectiveness of harvesters are, among other things, due to education, training, automation and organisational knowledge exchange (Häggström and Lindroos 2016).

When discussing the high effectiveness of harvesters, obtaining logs of accurate length has its economic importance. First of all, accurate processing leads to optimal trunk use for logs. However, economic losses can be observed when there are significant inaccuracies, not only due to inefficient use of the trunk: logs which are too long but are destined for the paper industry have to be shortened before debarking, which creates extra costs. In contrast, logs which are too short can be useless for the pallet industry, where an accurate, minimum length is necessary to produce a pallet, and meet safety and certification requirements. Basically, logs which are too short generate extra costs by being transported to the factory and then not being used for their intended purpose.

174

Bearing in mind the above considerations, the research presented in this paper was carried out to determine to what extent oak branching influences the effectiveness of using a harvester for thinning operations. In particular, the research was designed to mainly find out the influence of oak branching on the accuracy of the cross-cutting of top logs, but also to determine length accuracy in general, in different sections of the tree trunk. In other research, it was revealed that when harvesting aspen, the lengths of top logs (from tree crown where branches were removed) did not differ statistically from those of bottom logs (Mederski et al. 2008). This has been explained by the low density of aspen tree wood, which does not provide much resistance during delimbing, and consequently does not have a negative influence on the workings of the measuring device inside the harvesting head (allowing the trunk to remain in constant contact with the measuring wheel).

In light of these results, a hypothesis was formulated that the branching in oak (hardwood species) may lead to a decrease in the length accuracy of CTL timber assortments and could limit the processing of logs from the top parts of oak trees. Large branch diameter with a high wood density may present resistance to the delimbing knives, which could result in multiple attempts to cut the branches, thus decreasing the accuracy of the length measurement by the measuring wheel. At the same time, large diameter branches may limit log processing in the top parts of the trees.

Therefore, the objectives of this research were to find out: 1) the accuracy of log lengths from the bottom, middle and top parts of oak trees, and 2) harvester efficiency in the utilisation of the trunk for logs. Additionally, the study aimed to propose new variants of computer settings to obtain the largest possible share of logs with the expected length.

2. Materials and methods The research was carried out in compartment 590h (Lidzbark Forest District, northern Poland) during the thinning of a 61-year-old oak (Quercus robur L.) dominated stand. Prior to harvesting, the diameter at breast height (DBH) and height (h) of 441 trees selected for thinning were measured (Table 1). The DBH and h of the remaining trees were also recorded for stand characteristics. The trees selected for the study also had the height of the thickest branch (hb) measured (Table 1).

The DBH was measured once with a calliper (Haglof) with an accuracy of 1 mm, and the results were recorded with an accuracy of 1 cm (a result up to 0.5 cm was rounded down, while 0.6 cm or higher was rounded Croat. j. for. eng. 39(2018)2

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

Table 1 Characteristics of remaining stand, all harvested trees and those selected for study Trees

Remaining

Harvested

DBH

h

DBH

h

DBH

h

hb

cm

m

cm

m

cm

m

m

Mean

24

18.9

20

17.9

20

18.2

11.9

Minimum

13

11.2

6

8.9

12

11.2

2.2

Maximum

38

27.0

50

24.3

37

22.5

18.3

Standard deviation

5

2.4

6

2.5

6

2.7

2.8

137

137

441

441

69

69

69

Feature

N

Harvested & analysed

THI=0.84 DBH – diameter at breast height h – tree height hb – height of thickest branch on tree THI – thinning intensity, understood as mean diameter of harvested trees divided by mean diameter of remaining trees

up). The DBH of every tree was measured in a different direction, e.g. if one tree was measured along the N–S axis, the next was measured NE–SW, W–E, etc. Tree height and the thickest branch height were measured using a Vertex Laser (Haglof) with an accuracy of 0.1 m.

Subsequently, the study trees were selected to accurately represent all the DBH classes (to account for the smallest and largest diameters), but a higher frequency of trees was achieved in the more populated DBH classes. For this reason, the trees were selected in such a way that the selected study population (69 trees) presented a normal distribution of DBH. A ShapiroFrancia normality test set at a=0.01 was carried out, the result of which confirmed a normal distribution of DBH for the selected tree population. This was done not only to show the best representation of trees, but also to meet the best criteria for statistical evaluation.

The expected and acceptable assortment lengths for the customer were from 250 to 257 cm. This was applicable for all assortments, from bottom, middle and top tree sections. Therefore, the original harvester computer length settings were 252–257 cm, to meet customer expectations.

The aim was to process logs of a minimum top diameter of 7 cm over bark (minimum diameter of merchantable timber). From the originally selected 69 study trees, 280 processed assortments were measured after cutting: 69 bottom, 142 middle and 69 top logs.

More middle logs were cut as two or more assortments were obtained from the middle part of the trees depending on the length of the merchantable trunk.

P.S. Mederski et al.

The length measurements were made using a measuring tape and were accurate to 1 cm. Additionally, the smallest top diameter (one measurement) over the bark of the last assortment (from the highest part of the trunk) from each of the analysed trees was measured with a Digitech Professional electronic calliper with an accuracy of 1 mm. The timber was harvested in the middle of summer, in July, using a Ponsse Ergo single-grip harvester equipped with a H7 harvesting head (Table 2), which according to the manufacturer’s information, is suitable for coniferous and broadleaved species. Before thinning, calibration was carried out by an operator according to the current standards and with respect to species. The operator was 29 years old with 5 years’ experience working on harvesters, cutting pine and spruce to a large extent, occasionally the broadleaved species - birch and oak.

According to the manufacturer, the H7 harvesting head with three feed rollers (Fig. 1) is designed for harvesting both coniferous and broadleaved trees. Key elements of the machine include: additional feed Table 2 Technical specifications of H7 harvesting head Characteristics

Parameters

Measurements and basic parameters Weight incl. rotator and hanger

1200 kg

Length

1500 mm

Height with rotator

1680 mm

Width

1540 mm

Power consumption

130–140 kW

Operating pressure

28 MPa

Oil flow requirement

300 l min-1 Saw unit (hydraulic chainsaw)

Saw bar length

750 mm

Cutting diameter

640 mm

Chain speed

40 m s-1

Chain

0.404” Feeding unit

Feeding system

3 rollers

Largest opening

650 mm

Feed power

30 kN

Feed speed

0–5 m s-1 Delimbing unit

One stationary and four movable knives Separately controlled delimbing knives and feed rollers Proportional adjustment of rollers and knives

Croat. j. for. eng. 39(2018)2 175

P.S. Mederski et al.

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

assortments produced were the bottom logs at 254.1 cm; however, these were characterised by the smallest standard deviation (SD). The top logs had the biggest differentiation with the highest standard deviation (Table 3). The average assortment length of the middle logs (251.9 cm) did not fit the length intervals set in the harvester computer (252–257 cm), although it was within the range of expected length (250–257 cm).

Statistical analysis pointed towards a difference in the length of the top logs and middle logs in comparison to the bottom logs. At the same time, the difference in the top log and middle log lengths was not statistically significant (Table 3).

The largest share of assortments satisfying the minimum requirement of 250–257 cm was obtained from the middle part of the trees (93%), followed by the bottom logs (91%) and top logs (88%) (Fig. 2). At the same time, the bottom logs exhibited parameters Table 3 Measures of length variability in harvested logs Parameter

Fig. 1 Ponsse H7 harvester head (photo by Piotr S. Mederski) and measuring wheel (indicated with an arrow) in central part of head measure length of tree trunk by contact with its surface r­ ollers on the frame of the harvester head, upper and lower independently controlled hydraulic delimbing knives, as well as the controlled acceleration of the feed rollers. The length accuracy of the harvester ­measuring system was 1 cm.

The results obtained from the measured harvested assortments were analysed statistically. The data in the three analysed groups (bottom, middle and top parts) were of a different quantity and had no normal distribution. As a result of this, the non-parametric Kruskal-Wallis test was carried out, followed by a Dunn test for the multiple comparison of the means from each group. The analysis was carried out at a significance level of a=0.05 with STATISTICA 10.0. To graph the results, the program package R (3.0.2) was used (R Development Core Team, 2013).

Logs All

Bottom

Middle

b

a

Top

252.5

254.1

251.9

252.3a

Minimum, cm

243

249

243

245

Maximum, cm

266

259

266

266

Standard deviation

3.89

1.97

2.41

3.03

N

280

69

142

69

Mean, cm

Different letters next to mean values indicate statistically significant differences

3. Results 3.1 Log length The average assortment length was 252.5 cm, while the minimum and maximum lengths equalled 243 and 266 cm, respectively (Table 3). On average, the longest

176

Fig. 2 Proportions of logs from different parts of trees in respective assortment lengths Croat. j. for. eng. 39(2018)2

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

P.S. Mederski et al.

Fig. 3 Vioplot of log length data showing the frequency of data in groups; it presents a symmetrical reflection of kernel density estimation; the centre of each violin contains a boxplot

Fig. 4 Relation of top diameter over bark of last log (from highest tree part) to DBH (data from 69 trees: mean=13.3 cm; min.=8.2 cm; max.=29.0 cm; SD=4.1; N=69)

most closely complying with those set on the harvester computer: 83%, with only 57% and 54% for the middle and top logs, respectively (Fig. 2). The top logs also featured the largest number of logs that were too short (9%), while the overly long logs were generally found among the bottom logs (6%).

The most common log length was approximately 254 cm long. In other words, the harvester had a tendency to cut lengths within the middle of the 252–257 cm setting. For assortments from the middle part of the tree, a tendency for logs to be below the assigned value (namely 251 cm) was observed. The top logs tended to exhibit lengths from the middle of the assigned interval as well as constituting the most populous type of overly short assortment (Fig. 3).

4. Discussion

3.2 Effectiveness of trunk processing The oak trees were of a good quality with relatively short crowns representing an average of 35% of the total tree height (Table 1). An analysis of the last log from the highest part of the tree indicated a strong goodness of fit between the top diameter and the DBH (Fig. 4).

The mean value of the top diameter was 13.3 cm over bark; however, the largest reached as much as 29.0 cm over bark (Fig. 4). The relation between the DBH and the top diameter expressed by multiple R-squared was relatively high: 0.6062. A particularly dynamic increase in the top diameter of the last log was observed in the trees with a DBH above 25 cm; however, these were not very frequent in the present study (Fig. 4).

4.1 Log length The assortments obtained agreed to a large extent with the expected minimum (250 cm) and maximum (257 cm) length. Altogether, 91% of the assortments complied with these lengths. Although the bottom logs were most often (6%) excessively long (Fig. 2), the maximum excess in length was only 2 cm over the assigned values inserted into the harvester on-board computer (Table 3). At the same time, 93% of the assortments, whose length was between 250 and 257 cm, came from the middle part of the trunk.

The observed results point towards the fact that maintaining the maximum length of the thickest (and therefore the heaviest) assortments is the most problematic with respect to the influence of inertia. In the research analysing cut-to-length accuracy by Bembenek et al. (2015), birch and aspen bottom logs were also, on average, the longest. A similar tendency was observed when pine was harvested, where the influence of the assortment length on the accuracy of the cut log was analysed (Różański 1993). In the aforementioned research, the greatest number of assortments (96%) with deviations of ±2 cm was observed with the shortest assortments with an assigned length of 240 cm.

Among the longer assortments – 305, 405 and 505 cm – this condition was satisfied by 48%, 35% and 13% of the logs, respectively (Różański 1993). At the same time, the shortest assortments with the smallest diameter

Croat. j. for. eng. 39(2018)2 177

P.S. Mederski et al.

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

and length (240 cm) were obtained from the top parts of pine trees, which can be delimbed very readily (Mederski 2013).

Similar results were obtained by Nieuwenhuis and Dooley (2006), who described the influence of calibration on the precision of cut-to-length Sitka spruce (Picea sitchensis (Bong.)). These researchers compared the length measurements obtained using a measuring tape with the harvester measurements for two types of assortment: bottom logs and pulp wood (middle and top logs). In both instances, a smaller value was obtained from the tape measurements. Moreover, this pattern appeared more often in pulp wood, which means it is likely that the measuring wheel lost contact with the pulp log surface more often. At the same time, a reverse tendency was observed for assortment volume, where a smaller value was obtained when measured by the harvester.

It should also be noted that the research of ­Nieuwenhuis and Dooley (2006) showed that calibration did not always eliminate inaccuracies in measuring assortments, although it did have an influence on the correct determination of volume. Ultimately, however, the fluctuations in volume for pulp wood were unacceptable. Similar conclusions were drawn by Andersson and Dyson (2001), namely that the longer the assortment, the greater the inaccuracy in measuring its length. Moreover, the longest assortments were most frequently found to be too long.

It is important to point out that errors in assortment length measurements may not only be due to machine error. When manual harvesting is conducted using a chainsaw, factors such as branching and curviness have very limited impact (if any) on length accuracy.

However, logs that are too short can be more common when bucking with a chainsaw in comparison with a harvester (Zinkevičius et al. 2012).

Bearing in mind the fact that the bucking point is indicated by the measuring device located on the harvesting head, it is interesting to note that in the analysed research the bottom logs were the part of the assortment that most frequently complied with the assigned length. The bottom logs were also the most stable in maintaining lengths as the SD was the lowest, while the top logs were of a greater length variability with the highest SD (Table 3). During the experiment, it was also observed that the bottom ends of the trunks were without visible signs of bark damage. When large wounds are caused by thinning operations ­(Tavankar et al. 2017), they can also influence the accu­ racy of harvester measurement.

From the top parts of the trunk, 43% of the assortments were shorter than the assigned length in the

178

computer, however only 9% of the logs were shorter than the expected 250 cm. Overly long logs are obtained when the measuring wheel loses contact with the sides of the trunk measured. An explanation for the overly short logs seems a little more complicated, although two factors stand out from the rest and are worthy of note: surface unevenness (bends, stubs, and loss of bark), around which the measuring wheel rotates, and the lack of contact of the measuring wheel during the reverse movement of the feeding rollers with partially opened movable knives. The results obtained in this research can serve to explain the specific influence of these two factors. The largest percentage of logs with an inadequate length were the top logs.

This could be explained by stubs giving the measuring wheel a further distance to travel, hence decreasing the quality of delimbing. At the same time, this assortment most commonly required a retraction of the trunk by the feed rollers. This consequently reduced the impetus of the wood passing through the harvesting head needed to delimb the toughest branches. The repeated processing of the trunk and the multiple delimbing attempts (with the movable knives opening) can lead to the loss of contact of the measuring wheel with the sides of the trunk.

Another significant factor to consider in this research is the importance of the influence of the oak wood density and hardness in relation to increased resistance during delimbing. Mederski et al. (2008) did not observe statistical differences between the lengths of aspen bottom logs and top logs obtained in Lithuania. This was explained by the low density and hardness of that particular tree species and consequently its low resistance during delimbing. At the same time, low frequency rates of trunk retraction through the feed rollers (hence an increase in trunk impetus passing through the harvester) enabled the delimbing of the thickest branches. However, Bembenek et al. (2015) proved that there was a difference (statistically significant) between the length accuracy of thicker bottom logs and thinner top logs of aspen harvested in Poland. The differing results from research carried out on aspen in Lithuania (Mederski et al. 2008) and ­Poland (Bembenek et al. 2015) suggest that there could also be geographical and/or stand condition factors influencing the morphological features (branch thickness) of the same species, as indicated by Spinelli et al.

(2010). The branch size, angle between the branch and trunk, and trunk shape were indicated by Mederski (2013) as factors influencing the quality of delimbing and length accuracy.

These observations support Andersson’s (1999) conclusions suggesting that type, quality and assortment size (with no restriction to species) influence the Croat. j. for. eng. 39(2018)2

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

accuracy of bucking to a certain degree. At the same time, Andersson (1999) indicates that neither tree features, nor measurement method or climatic conditions influence bucking accuracy. According to Andersson (1999), the processing of broadleaved trees with a harvester, deep within the vegetation season, contributes to a substantial loss of bark, particularly in birch and alder. This leads to the trunk sliding through the harvester head, consequently resulting in length inaccuracies. Taking into consideration Andersson’s (1999) observations, it would be sensible to bear in mind these factors when trying to maintain steady assortment lengths in different environments.

4.2 Effectiveness of trunk processing A particularly important fact to emerge out of researching the harvesting of oak was the observation of a significant relationship between the top diameter of the assortments from the top part of the trunk and the DBH (Fig. 4). This relationship was present due to natural tree morphology – the thicker the tree, the bigger and longer the crown. Consequently, bigger and longer crowns had thicker branches growing lower on the stem (where there were bigger diameters), and these obstructed delimbing in the top section of the trees. In the case of all the remaining timber with a diameter of over 20 cm, of particular importance is the fact that there are significant levels of timber waste (tree tops). This also points towards certain limitations to harvester use in the processing of oak with larger diameter branches. As can be seen in research by ­Mederski (2013), the harvesting of pine in Central ­Europe allowed for the production of pine assortments with average minimum diameters of 9.4 and 10.6 cm over the bark (for 45 and 60-year-old pines, respectively), when the expected diameter was 7 cm over the bark. However, tree tops are in growing ­demand and can have a high value in the energy ­sector, especially when the quality of logging residues is satisfactory (Nilsson et al. 2016). As harvesters can have limitations in the processing of logs of smaller diameters from the tree tops, CTL technology in broadleaved stands can naturally contribute to a larger share of a renewable energy source. The relationship between the size of the top diameter of the last log and the DBH can be explained by the fact that trees with the biggest diameter have a stronger build, a longer crown and thicker branches. Thick branches, in ­particular, interfered and limited delimbing by the harvester head. It should also be assumed that the ­assortment length from the top logs of the thickest tree is even less accurate due to the lower quality of ­delimbing of the thickest branches (Mederski 2013).

P.S. Mederski et al.

4.3 Proposed corrections for error window Bearing in mind the tendencies in the bucking of the trunk into logs (with respect to the three aforementioned trunk sections), it is suggested that for top and middle logs different settings should be applied. In order to attain the optimal distribution, spread within the 250–257 cm limits with the minimum interval values for oak trees (of considered sizes and age), the following settings could be applied to the Ponsse Ergo (with the H7 head) harvester on-board computer: 254–257 cm for top and middle logs (the lower value moved up by 2 cm), and 252–255 cm for bottom logs (the higher value lowered by 2 cm). The proposed settings should result in fewer logs that are too short when obtained from the top and middle parts of the trunk, as well as a lower number of bottom logs that are too long.

5. Conclusions To summarise, the results obtained for the analysed harvester and species (oak), three dependencies can be clearly observed: Þ the larger (and heavier) the assortment, the greater the tendency to produce overly long timber, particularly in the case of bottom logs Þ the tested harvester had a tendency to produce top logs that were too short Þ the use of the trunk was limited, and the larger the DBH, the more of the top of the tree remained unprocessed Generally, the bottom logs were, on average, approximately 2 cm longer than the middle and top logs.

At the same time, the highest standard deviation of length appeared in the top logs, with the lowest in the bottom logs. The length of the bottom log assortments differed statistically from those of the other trunk ­assortments.

A Ponsse Ergo harvester equipped with a H7 harvesting head was an appropriate solution for harvesting a hardwood, broadleaved species, in a 61-year-old oak stand. The greatest accuracy of log length was achieved for the bottom logs, while the biggest share of logs that were too short were obtained from the top part of the trees. The mean length values indicated that the bottom logs were cut within the middle range of the limits of error, but the top and middle logs were close to the minimum length limit set in the harvester computer. To obtain better results, it is suggested that dif­ ferent length tolerances should be applied: 254–257 cm for thinner top and middle logs, and 252–255 cm for bottom logs. Some difficulties were observed when

Croat. j. for. eng. 39(2018)2 179

P.S. Mederski et al.

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181)

processing top logs from thicker trees. The mean top diameter of 13.3 cm over bark of the last log suggests that there is still potential for log processing from these parts of trees. The thicker the tree, the larger the tree top that remains. Wider use of a harvester, for coniferous and broadleaved stands in Central Europe, allows for the better utilisation of CTL technology within a given area by reducing machine relocation costs.

Acknowledgements This paper was written with the experience and knowledge gained during completion of the project: The potential of harvester application for hardwood species, EO-2717-22/13, financed by the General Directorate of the State Forests, 2013-2016. Additionally, the authors would like to thank Caiyin Wang for her help in data collection. Special thanks to Dariusz Szczawiński, PhD. current manager of Lidzbark Forest District for his help in arranging the field tests and to Kazimierz Durzyński for providing a harvester for this research. Finally, the authors would like to thank the two reviewers for their valuable comments that helped to improve this paper.

6. References Andersson, B., 1999: Log measuring accuracy of cut-tolength harvesters and processors. Forest Engineering Research Institute of Canada, Vancouver, Canada, Final Report of project CC97139-ORE, 18 p.

Bembenek, M., Mederski, P.S., Karaszewski, Z., Łacka, A., Grzywiński, W., Węgiel, A., Giefing, D.F., Erler, J., 2015: Length accuracy of logs from birch and aspen harvested in thinning operations. Turkish Journal of Agriculture and Forestry 39(6): 845–850.

Borecki, T., Stępień, E., Wójcik, R., Orzechowski, M., 2016: Verification of the principles of accounting for the size of the allowable fellings in forest management planning. Drewno 59(197): 61–75.

Cacot, E., Bigot, M., Cuchet, E., 2006: Developing full-mechanized harvesting systems for broadleaved trees: a challenge to face the reduction of the manual workface and to sustain the supply of hardwood industries. In Proc., 29th Council on Forest Engineering (COFE) Conference: Working Globally – Sharing Forest Engineering Challenges and Technologies Around the World July 22 to August 2, eds. W. Chung and H.S. Han, Coeur d’Alene, USA, 9 p.

Forestry, 2006: Central Statistical Office of Poland, Warsaw, 285 p.

Frutig, F., Fahrni, F., Stettler, A., Egger, A., 2007: Mechanisierte Holzernte in Hanglagen (Mechanised timber harvesting on slopes). Wald und Holz 4: 47–52.

Glöde, D., 1999: Single- and double-grip harvesters – productive measurements in final cutting of shelterwood. Journal of Forest Engineering 10(2): 63–74.

180

Häggström, C., Lindroos, O., 2016: Human, technology, organization and environment – a human factors perspective on performance in forest harvesting. International Journal of Forest Engineering 27(2): 67–78.

Karaszewski, Z., Łacka, A., Mederski, P.S., Noskowiak, A., Bembenek, M., 2016: Damage caused by harvester head feed rollers to alder, pine and spruce. Drewno 59(197): 77–88.

Labelle, E.R., Soucy, M., Cyr, A., Pelletier, G., 2016: Effect of Tree Form on the Productivity of a Cut-to-Length Harvester in a Hardwood Dominated Stand. Croatian Journal of Forest Engineering 37(1): 175–183.

Malinen, J., Laitila, J., Väätäinen, K., Viitamäki, K., 2016: Variation in age, annual usage and resale price of cut-tolength machinery in different regions of Europe. International Journal of Forest Engineering 27(2): 95–102.

Martin, P., Lapeyre, D., Douchet, O., Restoy, G., Guegand, G., 1996: Récolte mécanisée des taillis en bois ronds (Mechanised timber harvesting in coppice stands). AFOCEL Fiche Information-Forêt 4(540): 1–6.

Mederski, P.S., 2013: The potential of harvester use for thinning operations in mixed birch-pine stands. Poznań University of Life Sciences Press, Poznań, 109 p.

Mederski, P.S., Bembenek, M., Jakubowski, M., Zinkevičius, R., 2008: Length accuracy of aspen logs harvested with CTL 40HW harvester head designed for broadleaved species. Human and Nature Safety 14: 241–243.

Mederski, P.S., Bembenek, M., Karaszewski, Z., Łacka, A., Szczepańska-Álvarez, A., Rosińska, M., 2016a: Estimating and Modelling Harvester Productivity in Pine Stands of Different Ages, Densities and Thinning Intensities. Croatian Journal of Forest Engineering 37(1): 27–36.

Mederski, P.S., Jakubowski, M., Karaszewski, Z., 2009: The Polish landscape changing due to forest policy and forest management. iForest 2(4): 140–142.

Mederski, P.S., Karaszewski, Z., Rosińska, M., Bembenek, M., 2016b: Dynamics of harvester fleet change in Poland and factors determining machine occurrence. Sylwan 160(10): 795–804.

Moskalik, T., Borz, S.A., Dvořák, J., Ferencik, M., Glushkov, S., Muiste, P., Lazdiņš, A., Styranivsky, O., 2017: Timber Harvesting Methods in Eastern European Countries: a Review.

Croatian Journal of Forest Engineering 38(2): 231–241.

Nillson, D., Pettersson, R., Thörnqvist, T., Nylinder, M., 2016: The importance of accurate measurement of comminuted logging residues’ moisture contents for small-scale forest owners. Drewno 59(198): 99–110.

R Development Core Team, 2013: R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Różański, H., 1993: Wydajność i dokładność wyróbki drewna harwesterem wysięgnikowym Lokomo FMG 990/756 (Productivity and bucking accuracy of the Lokomo FMG 990/756 harvester). Przegląd Techniki Rolniczej i Leśnej 10: 18–19.

Sionneau, J., Cuchet, E. 2001: Mechanisation of thinnings in hardwood, the French experience. In Proc., International Croat. j. for. eng. 39(2018)2

Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees (173–181) Conference: Thinnings: A valuable forest management tool, IUFRO Unit 3.09.00 and FERIC, Natural Resources Canada and Canadian Forest Service, Québec, Canada 2001, 8 p.

Spinelli, R., Hartsough, B.R., Magagnotti, N., 2010: Productivity standards for harvesters and processors in Italy. Forest Products Journal 60(3): 226–235.

Spinelli, R., Magagnotti, N., Nati, C., 2009: Options for the mechanised processing of hardwood trees in Mediterranean forests. International Journal of Forest Engineering 20(1): 39–44.

Spinelli, R., Magagnotti, N., Picchi, G., 2011: Annual use, economic life and residual value of cut-to-length harvesting machines. Journal of Forest Economics 17(4): 378–387.

Spinelli, R., Magagnotti, N., Schweier, J., 2017: Trends and Perspectives in Coppice Harvesting. Croatian Journal of Forest Engineering 38(2): 219–230.

Spinelli, R., Owende, P.M.O., Ward, S.M., 2002: Productivity and cost of CTL harvesting of Eucaliptus globulus stands

P.S. Mederski et al.

using excavator-based harvesters. Forest Products Journal 52(1): 67–77.

Stampfer, K. 1999: Influence of terrain conditions and thinning regimes on productivity of a track-based steep slope harvester. In Proc., International Mountain Logging and 10th Pacific Northwest Skyline Symposium March 28 – April 1, eds. J. Sessions and W. Chung, Corvallis, Oregon, USA, 10 p.

Tavankar, F., Picchio, R., Nikooy, M., Lo Monaco, A., Venanzi, R., Iranparast Bodaghi, A., 2017: Healing rate of logging wounds on broadleaf trees in Hyrcanian forest with some technological implications. Drewno 60(199): 65–80.

The National Forest Inventory, 2015: Bureau for Forest Management and Geodesy, Sękocin Stary, Poland, 176 p.

Zinkevičius, R., Steponavičius, D., Vitunskas, D., Činga, G., 2012: Comparison of harvester and motor-manual logging in intermediate cuttings of deciduous stands. Turkish Journal of Agriculture and Forestry 36(5): 591–600.

Authors’ addresses: Assoc. prof. Piotr S. Mederski, PhD.* e-mail: piotr.mederski@up.poznan.pl Mariusz Bembenek, PhD. e-mail: mariusz.bembenek@up.poznan.pl Department of Forest Utilisation Poznań University of Life Sciences Wojska Polskiego 71A 60-625 Poznań POLAND Zbigniew Karaszewski, PhD. e-mail: z_karaszewski@itd.poznan.pl Wood Science and Application Department Wood Technology Institute Winiarska 1 60-654 Poznań POLAND Zenon Pilarek, PhD. e-mail: zpilarek@up.poznan.pl Department of Forest Technology Poznań University of Life Sciences Wojska Polskiego 71C 60-625 Poznań POLAND

Received: December 12, 2017 Accepted: March 3, 2018

Agnieszka Łacka, PhD. e-mail: agnieszka.lacka@up.poznan.pl Department of Mathematical and Statistical Methods Poznań University of Life Sciences Wojska Polskiego 28 60-637 Poznań POLAND *Corresponding author

Croat. j. for. eng. 39(2018)2 181

Original scientific paper

Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes Zbigniew Karaszewski, Agnieszka Łacka, Piotr S. Mederski, Mariusz Bembenek Abstract Harvesters have become a common solution for wood harvesting in coniferous and broadleaved stands. Unfortunately, not every customer will accept logs with damage on the lateral surface of the roundwood caused by feed roller spikes. The extent of the wood damage caused by the spikes of harvester heads depends mainly on the type of feed rollers and tree species. The objective of the study was to investigate the external damage to pine (Pinus sylvestris L.) roundwood from harvester head spikes depending on the season of the year and harvester engine RPM, as well as the significance and potential consequences of such damage. The scope of the study also included an analysis of wood damage depth in three stem sections. The experimental plots selected were all in an 85-year-old pure pine stand. Logging was performed using a Ponsse Beaver harvester with an H60e harvester head manufactured in 2006. The mean depth of wood damage at all the points of measurement was 4.1 mm, while the maximum depth of wood damage totalled 5.3 mm. The depth of wood damage depended on the season of the year in which the logging work was performed, the harvester engine RPM and the stem section from which the log was processed.

The damage was the deepest during summer operations and the shallowest during winter and springtime. The differences were statistically significant, however, the difference in the depth of damage was only 1 mm in average. Deeper wood damage was found at a lower engine RPM. Wood damage depth differed axially, and the least damage was found in the bottom logs.

Keywords: bark loss, harvesting head, mechanised logging, pilodyn, Pinus sylvestris L.

1. Introduction Nowadays, harvesters can be used for wood harvesting in coniferous and broadleaved stands (Mederski et al. 2016). Currently, due to cut-to-length technology (CTL), more roundwood of better quality is sold as short logs to sawmills, plywood and furniture ­manufactures. Unfortunately, not every customer will accept logs with damage on the lateral surface of roundwood (throughout the bark) caused by feed roller spikes. During mechanised harvesting, the bark of the processed logs may be removed, thus reducing the degree of protection it can offer to the wood. Regardless of species, bark adhesion is the strongest in winter and early spring, between December and April, and the weakest (nearly non-existing) in late spring and summer, between May and August (Simonov 1984, after Aniszewska and Więsik 2015), probably

because of the activity of the cambium (Uzunović et al. 1999). In mechanised logging processes, bark is usually peeled off longitudinally by the delimbing knives, which may remove strips of bark or both bark and wood. The season and the ambient temperature significantly affect resistance – as the temperature decreases, the shear strength of both the bark and wood increases (Aniszewska and Więsik 2015). Damage to the outer layers of wood associated with bark stripping is easily recognized and often criticised, even though it also occurs on logs prepared using a chainsaw (Spinelli et al. 2011) and is usually accepted. The extent of this damage may differ depending on the season of the year during which logging is performed and on the method of processing (short or long wood system). Perforation of the roundwood surface influences, for example, its technical properties, and causes the roundwood to dry out more rapidly, which is un-

Croat. j. for. eng. 39(2018)2 183

Z. Karaszewski et al. Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191)

desirable in paper production (Warkotsch 1994). Winter is considered a better time for logging (Lee and Gibbs 1996, Uzunović et al. 1999, Murphy and Pilkerton 2011, Murphy and Acuna 2017). During spring and summer, with a higher air temperature and moisture, secondary wood defects may occur, such as blue stain.

Wood damaged on the lateral plane by the harvester is susceptible to fast and deep penetration of fungi (Lee and Gibbs 1996). As buyers expect a consistent year-round supply of roundwood on the market, logging also takes place in those months when the risk of secondary depreciation of the product is higher.

Besides weather conditions, the extent of damage may be affected by technical considerations related to the use of logging machinery and its condition. Before work, operators set up harvester computers taking into consideration tree species, type of assortment, length and diameter of assortment, as well as length tolerances. Moreover, harvester engine revolutions per minute (RPM) may be adjusted by the operator in order to reduce fuel consumption. Lower RPM means that the hydraulic pumps operate with less power, which affects the speed and torque of the feed rollers. This may in turn reduce logging output, as well as decrease delimbing efficiency, and thus potentially cause increased damage to the roundwood due to repeated attempts to delimb. As the literature indicates, bark adhesion depends on the season and it is at its lowest during late spring and early summer. Therefore, it was hypothesised that roundwood damage from feed roller spikes would be most severe in summer. It was also hypothesised that a higher RPM might cause deeper spike penetration in the wood as more aggressive rotations could cause greater damage. Thus, the purpose of this study was to investigate and highlight any differences in terms of the external damage to pine (Pinus sylvestris L.) roundwood from the harvester head spikes depending on the season of the year and engine RPM, as well as the significance and potential consequences of such damage. In addition, the scope of the study included an analysis of wood damage depth in three stem sections (bottom, middle and top) and pilodyn penetration tests (Giefing 1985) in areas adjacent to the damage sites. Pine was selected as the species of analysis as it is a common species cut by harvester in Central Europe and Baltic countries (Moskalik et al. 2017).

2. Material and methods

the mean DBH was 27 cm, the height was 22 m, and the standing volume totalled 281 m3/ha (Forest Management Plan 2007). The stand was of moderate density and of an average technical quality. The thinning intensity was 30 m3/ha. Access to the stand was provided by a network of machine operating trails, spaced 20 m apart axis-to-axis. The processed assortments were stored near the trail.

The study was performed during the four seasons at various ambient temperatures: winter (February, air temperature from 0 to +1 °C), spring (April, from –2 to +4 °C), summer (September, from +14 to +23 °C), and autumn (November, from +8 to +10 °C). During the study, all the trees were processed into industrial wood, with a length of 2.50 m.

2.2 Machine characteristics Logging was performed using the Ponsse Beaver harvester equipped with an H60e head. The harvester had 27,000 hours of work at the beginning of the research (Table 1).

Table 1 Basic technical specifications of Ponsse Beaver harvester with H60e head Harvester Year manufactured

2006

Engine

MB OM 904 LA

Engine power, kW (hp)

170 (228)

Hydraulic fluid, l

208

Software

OPTI 4G

Operating pressure, MPa

20–24

Required oil flow, l/min

240–300 Harvester head H60e

Feed system

Bogie roller assemblies with spike rollers

Feed rollers, no

4

Gross feed force, kN

24

Feed speed, m/s

0–5

Spike length, mm

20

Spike width at base, mm

15

Hydraulically movable knives, no

3

Largest opening, mm

600

2.1 Area of study The experimental plots were selected in a 11.95 ha pure pine stand, on sandy soil located in Dąbrowa Forest District, central Poland. The stand age was 85 years,

184

The work was performed by one operator who worked day and night shifts interchangeably in each season. The operator had 500 hours of experience on Croat. j. for. eng. 39(2018)2

Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191) Z. Karaszewski et al.

the harvester. Firstly, the harvester head measuring system was calibrated (to obtain an accurate length and diameter) by a Ponsse technician giving instructions to the operator, who performed the subsequent calibrations.

tissue. Penetration depth without bark was measured only in the wood tissue, after manual removal of the bark. Pilodyn penetrations were taken twice for each type: with and without bark, and mean values for each group (with and without bark) were calculated.

2.3 Study design and measurements

2.4 Statistical analysis

The wood damage from the feed roller spikes, understood as wood damage depth, was tested with respect to the following criteria: Þ season of the year in which the logging was performed: winter, spring, summer, or autumn Þ harvester engine RPM (two of the most common setups): low (1600 RPM) or high (1750–1800 RPM) Þ section of the stem from which the log was cut: bottom, middle, or top. The bottom section was the first butt log, the middle section was any log located between the butt and top log, and the top section was the last log with a diameter of at least 7 cm overbark at the top end.

The wood damage depth was measured using the depth gauge of a Mitutoyo digital caliper. To obtain an accurate depth, the bark was removed and then a chisel was used to detach a chip of wood damaged by the feed roller spikes according to methods described in earlier research (Karaszewski et al. 2016b). The measurements were performed in the middle of a log from each tree section: bottom, middle and top. The six most severe damage points (the widest openings) were measured on each log. The caliper was also used to measure the thickness of the bark (with phloem) near the damage depth measurement site (one site being six neighbouring measurement points). The damage depth and bark thickness measurements were performed at an accuracy of 0.01 mm. For further analyses, the mean and maximum damage depths from each measurement site were used. The mean bark thickness was used to analyse the impact of the parameter on wood damage depth. Additionally, the overbark log diameter was measured at the damage depth measurement site, using a caliper with a precision of 1 mm.

Wood susceptibility to mechanical damage was tested using a Pilodyn 6J. This instrument has a steel cylinder (penetrator) of 2.5 mm in diameter and 600 mm in length, used for measuring the penetration depth in wood at a constant energy of 6 J. Pilodyn tests were performed on the cut logs, with and without bark, at the damage depth measurement sites. The penetration depth with bark was understood as the total length of penetration through the bark and wood

An analysis of variance was performed preceded by the Fligner-Killeen test for variance homogeneity and Pearson’s test for distribution normality. Post hoc tests were performed using Tukey’s test. If conditions for parametric analysis of variance were not met, the Kruskal-Wallis test and Dunn’s test would be used. A significance level of a=0.05 was used for all the analyses. A Pearson’s correlation matrix was calculated for the mean and maximum damage depths, log diameter, bark thickness, and pilodyn penetration depth with and without bark. Statistical analyses were provided using the R 3.3.1 software (R Core Team 2017).

3. Results of the study The depth of wood damage from the feed roller spikes was analysed on a total of 102 stems, cut into 306 logs (Table 2). In total, 1836 measurements of wood damage depth and 1836 measurements (306 logs x 6 measurement depths per log) of bark thickness were taken. A total of 612 Pilodyn penetration depth measurements were performed with bark and without bark. The mean depth of wood damage by the harvester head spikes at all the points of measurement was 4.1 mm, while the maximum depth of wood damage at all these points totalled 5.3 mm.

Table 2 Log parameters by stem section Logs

n

Length cm

Mid diameter cm±SD

Bark thickness mm±SD

Bottom

102

250

20.8±3.4

7.9±2.5

Middle

102

250

16.3±2.6

2.6±1.1

Top

102

250

12.3±2.5

2.0±0.6

The analysis of variance showed a significant interaction between the three parameters – season, harvester engine RPM, and the stem section, which impacted the mean damage depth (p=0.0468). The damage depth was the greatest in summer, at a lower RPM, and in logs cut from the top stem section (mean depth = 6.0 mm, maximum depth = 7.2 mm), while the least damage was caused in winter and autumn, at a

Croat. j. for. eng. 39(2018)2 185

Z. Karaszewski et al. Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191)

Fig. 1 Mean damage depth by season higher RPM, in logs from various sections (mean depth = 3.0 and 3.1 mm, and maximum depth = 4.7 and 3.9 mm, respectively; Fig. 1).

An analysis of all three factors showed a complex system of interrelations. In order to present clear differences, the factors were considered independently, in relation to season, engine RPM and section of the stem.

The mean damage depth varied according to the season and ranged between 3.7 mm in winter and 4.7 mm in summer (Table 3). The maximum damage

depth ranged from 5.0 mm in spring to 5.8 mm in summer (Table 3).

The damage depth was greater at a low RPM, both in terms of the mean (+0.4 mm) and maximum values (+0.6 mm, Table 3).

The mean and maximum damage depth on the processed pine stems differed axially and reached from the bottom to the top: 3.9, 4.1 and 4.3 mm on the butt, middle and top logs, respectively. The maximum depth was 1.1–1.4 mm greater than the mean values (Table 3).

Table 3 Mean and maximum wood damage depth according to harvesting season, engine RPM and stem sections (with basic statistics: minimum, maximum values and standard deviation SD)

Criteria

Sample size

Depth of damage

no of logs

Mean value

Min.

Max.

Maximum value

Min.

Max.

mm

mm

mm

mm

mm

mm

Winter

60

3.7

2.0

7.1

1.0

5.3

2.5

9.0

1.5

Spring

78

4.0

2.2

6.6

0.9

5.0

2.7

7.9

1.1

Summer

78

4.7

2.4

7.2

1.2

5.8

3.3

9.8

1.6

Autumn

90

4.1

2.0

6.8

1.0

5.2

2.7

8.3

1.2

Low RPM

147

4.4

2.0

7.2

1.1

5.6

2.7

9.8

1.4

High RPM

159

3.9

2.0

7.0

1.1

5.0

2.5

8.7

1.4

Bottom logs

102

3.9

2.0

7.1

1.1

5.0

2.7

9.0

1.4

Middle logs

102

4.1

2.0

6.6

1.0

5.3

2.5

8.9

1.2

Top logs

102

4.3

2.0

7.2

1.1

5.7

2.5

9.8

1.4

186

SD

SD

Croat. j. for. eng. 39(2018)2

Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191) Z. Karaszewski et al.

Fig. 2 Pilodyn penetration depth with and without bark by season for bottom, middle and top logs The mean pine bark thickness was 4.2 mm when combining all stem sections. The bark thickness decreased from the bottom to the top of the stem (Table 2).

The shallowest penetration depths with bark (20.1 mm) and without bark (18.0 mm) in all the sections were found in summer (Fig. 2). In winter, the penetration depth with and without bark was similar and considerable in all sections (20.3 mm and 20.4 mm, respectively). The penetration depth without bark tended to be the shallowest in the bottom logs, greater in the middle logs, and the greatest in the top logs (Fig. 2). Statistical analysis confirmed significant differences in the pilodyn penetration depth without bark between the bottom and middle logs (p<0.0001), and between the bottom and top logs (p<0.0001). The

Mean damage depth

Maximum damage depth

Mean bark thickness

Diameter

Table 4 Correlation matrix

Mean damage depth

1.0

0.9

–0.2

–0.3

Maximum damage depth

0.9

1.0

–0.2

–0.3

Diameter Pilodyn penetration with bark Pilodyn penetration without bark

–0.3

–0.3

0.7

1.0

0.1

0.1

0.1

0.1

0.0

0.1

–0.3

–0.2

Characteristic

penetration depths with bark were more similar, and significant differences were found between the bottom and middle logs (p=0.0020), as well as between the middle and top logs (p=0.0287).

No correlation was found between the mean or maximum damage depth and pilodyn penetration depth. As the bark thickness increased, the damage depth of the wood decreased, though the relation was weak, at r=–0.2 (tab. 4). As expected, the bark thickness increased along with log diameter, with r=0.7.

4. Discussion 4.1 Seasons The wood damage depth differed in the analysed seasons (Table 3), but in all the cases, its value varied widely, as reported in previous studies for pine and spruce (Nuutinen et al. 2010, Karaszewski et al. 2016a), birch (Nuutinen et al. 2010) and alder (Karaszewski et al. 2016b).

With regard to seasons, the wood damage depth was found to be shallowest at low air temperatures (between –2 and +4 °C), in terms of both the mean and maximum depths (Table 3). The decreased wood damage depth in winter might also have been due to better bark adherence during this season. Bark adherence to the wood affects the impact of the delimbing knives on their bark shaving, or may affect the presence of bark on the log. When the bark can be stripped with less force, as is the case during the late spring and summer in Europe, the delimbing knives may debark

Croat. j. for. eng. 39(2018)2 187

Z. Karaszewski et al. Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191)

the log and, as a consequence, the harvester head spikes scar the exposed wood. However, if the bark remains on the assortment after delimbing, the resistance of the bark during the short initial contact with the spikes prevents deeper damage to the wood itself.

According to Murphy and Pilkerton (2011), bark loss from processed wood may be five times higher in late spring and early summer than in winter. In order to limit bark loss, rubber rollers have been suggested as an alternative to the steel ones used in the present study (Lee and Gibbs 1996; Murphy and Acuna 2017).

However, currently this is not a popular choice under European conditions. Suggestions for: 1) a decrease in the length and increase in the number of feed roller spikes, aimed at improving the contact surface between the roller and wood (Warkotsch 1994), and for 2) the monitoring of machine performance (Gerasimov and Seliverstov 2010) are reflected in new technical (Sowa et al. 2013) and organizational solutions (Błuszakowska and Nurek 2016).

As expected, bark thickness changed axially, and was greatest in the bottom section. The depth of wood damage was negatively correlated (weakly) with bark thickness, although the least damage was found in the bottom logs, where the bark was the thickest, and the most damage was found in the top logs, with the thinnest bark. These results are comparable to the findings and observations made by Nill et al. (2011), who reported a decreasing susceptibility to damage in tree species, which tend to form thicker bark layers. According to Nill at al. (2011), among deciduous species, beech, which has a relatively thin periderm layer, was the most susceptible to damage during logging. Moreover, in coniferous tree stands, species with thick bark, such as Douglas fir, pine, or larch, were much less susceptible to damage than spruce, which has a relatively thin bark. However, damage seen in the thin-barked fir and spruce differed significantly, which indicates that there are other significant factors affecting a species’ susceptibility to damage. Despite the similar bark thickness, fir was more resistant to damage during logging than spruce (Nill et al. 2011). This is supported by the findings of Kohnle and Kändler (2007), who pointed to the potential significance of anatomical bark features, such as sclerenchyma cells, often found in fir bark (but not in spruce bark), which may offer better resistance against spike penetration.

4.2 RPM At a lower engine RPM, the feed rollers of the harvester head caused significantly more damage while processing the logs than at a higher RPM (Table 3). A lower RPM slowed the movement of the spikes on the

188

round wood surface. As a result, this probably gave the spikes more time to penetrate the wood tissue, and, therefore, the damage was more severe compared to the cases when a higher RPM was applied. The larger mean damage found in the top logs (compared to the logs from other stem sections, Table 3) was consistent with observations by other authors (Uzunović et al.

1999, Gerasimov and Seliverstov 2010, Karaszewski et al. 2016b), although this finding was likely not only due to the branching of this section, but also due to its thinner bark and the lower mechanical resistance of the juvenile wood (Tomczak et al. 2010). This was confirmed by the pilodyn tests (Fig. 2), with statistically significant differences between the bottom, middle and top logs. Van der Merwe et al. (2015) reported yet another reason for the larger damage found in top logs, pointing to the relationship between the frequency of the feed roller movement along the processed stem (which is usually the greatest for top logs) and increased wood damage, also associated with fibre loss.

4.3 Pilodyn In order to investigate possible associations between wood resistance to penetration and damage depth, pilodyn tests were performed. The pilodyn penetration depth in fresh pine wood is only slightly affected by moisture content (Giefing and Kokociński 1991), which motivated the choice of pilodyn as a testing instrument. Contrary to expectations, the pilodyn penetration in the wood without bark was deepest during winter (Fig. 2). This divergence between the initial assumptions and findings was also reflected in the lack of correlation between the parameters studied, which was also true in the case of pilodyn penetration with bark (r=0.1 in both cases; Table 4).

The maximum wood penetration depth for Pilodyn 6J is 40 mm (Giefing 1985). Pilodyn penetration depth is correlated with wood density in coniferous trees (Giefing and Jabłoński 1988), and in dry wood it is correlated with its hardness, r=–0.70 (Giefing and ­Romanowska 1992). However, pilodyn penetrates wood at a constant energy of 6 J, while the pressure of harvester head feed rollers can be higher or simply variable, which may also explain the unexpected ­results and lack of correlations. Summarising the ­pilodyn penetration tests, it should be concluded that factors other than resistance to pilodyn penetration were associated with wood damage depth, and that the instrument did not satisfactorily simulate the ­damage caused by harvester head spikes.

Aside from the scientific investigation of the differences in mean and maximum damage depth, practical implications can be derived from the assessment.

Croat. j. for. eng. 39(2018)2

Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191) Z. Karaszewski et al.

Damage on the roundwood surface is a natural consequence of harvester use and has a negative impact on wood quality (i.e. there is a risk of blue stain rendering the wood unattractive after further processing). Although the differences between damage depths in each season were statistically significant, the observations may be considered of little importance. In fact, although a spike damage depth between 1.0 and 1.5 mm may be insignificant for most timber-sector customers, the fact that damage occurs is important.

In practice, such defects caused by harvester-related damage (spike damage) up to 2 cm depth, defined by technical specifications for medium-sized logs, are considered allowed in some European countries.

These results provide an opportunity for further development of harvester heads (feed rollers in particular) for coniferous and broadleaved species in lowland forests (Moskalik 2002a, Moskalik 2002b, Więsik et al.

2005, Mederski et al. 2016, Moskalik et al. 2017) and highland forests (Visser and Stampfer 2015, Enache et al. 2016).

5. Conclusions The following conclusions were drawn from the results: Þ The wood damage caused by the harvester head spikes depended on the analysed factors: the season of the year in which the logging work was performed, harvester engine RPM and the stem section from which the log was processed.

These factors had a varying impact on wood damage and their interaction was statistically significant. Þ The wood damage depth differed according to the season. The damage was the deepest in summer and the shallowest during winter and springtime. Þ Contrary to the hypothesis, deeper wood damage was found at 1600 RPM: a mean of 4.35 and a maximum of 5.61 mm, respectively. When 1750–1800 RPM was used, spike penetration was shallower: 3.92 and 5.04 mm (mean and maximum), respectively. Þ The wood damage depth differed axially: the least damage was found in the bottom logs (mean 3.9 mm, max. 5.0 mm), more in the middle logs (mean 4.1 mm, max. 5.3 mm), and the most damage was found in the top logs (mean 4.3 mm, max. 5.7 mm). Þ The pilodyn penetration depth was not correlated with wood damage depth. However, the

penetration depth increased axially from bottom to top, which was consistent with the axial increase in damage from the harvester head spikes (without considering the bark). Þ In practical terms, seasonal differences in the depth of damage from feed roller spikes may have little impact on customers’ decisions. They may be more significant in the case of plywood processing, where the outer layer is the most valuable, or when the roundwood is used as the final product. The use of a lower harvester RPM results in lower surface roundwood quality – a higher RPM should be used to limit damage to the surface of the roundwood from feed roller spikes.

Acknowledgements The authors would like to thank Artur Karetko, Forest District Manager, for providing access to the experimental plot, Magdalena Kaczmarek and Jacek Godlewski for their assistance with the field work, and Witold Urbaniak for technical support with the harvester calibration.

The paper has been written on the basis of partial results from: Karaszewski, Z., Noskowiak, A., 2016: Impact of season on damage to mechanically harvested pine wood. ST-1-BDZ/2016/N. Research and technical documentation. Wood Technology Institute, Poznań, Poland.

6. References Aniszewska, M., Więsik, J., 2015: Korowarki (Debarkers). In: Urządzenia techniczne w produkcji leśnej. Tom 2. Maszyny i urządzenia do pozyskiwania i transportu drewna. Więsik, J., (ed) Wydawnictwo SGGW, Warszawa, 278–282.

Błuszkowska, U., Nurek, T., 2016: Effect of organizational factors on harvester exploitation efficiency. Sylwan 160(6): 443–451.

Enache, A., Kuhmaier, M., Visser, R., Stampfer, K., 2016: Forestry operations in the European mountains: a study of current practices and efficiency gaps. Scandinavian Journal of Forest Research 31(4): 412–427.

Forest Management Plan for Dąbrowa Forest District 2007: 2007–2016.

Gerasimov, Y., Seliverstov, A., 2010: Industrial round−wood losses associated with wood harvesting systems in Russia.

Croatian Journal of Forest Engineering 31(2): 111−126.

Giefing, D.F., 1985: Construction of pilodyn and its application. Sylwan (129)7: 63–68.

Giefing, D.F., Jabłoński, K., 1988: Susceptibility of wood of live pine trees to the penetration with the pilodyn needle

Croat. j. for. eng. 39(2018)2 189

Z. Karaszewski et al. Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191) according to the density and humidity of the wood. Roczniki AR w Poznaniu 193: 7–21.

Giefing, D.F., Kokociński, W., 1991: Zmiany podatności drewna sosny na penetrację iglicą pilodynu w następstwie jego wysychania (Changes in susceptibility of pine wood to the pilodyn penetration due to wood drying). Prace Komisji Technologii Drewna 12: 17–25.

Giefing, D.F., Romanowska, K., 1992: Susceptibility of wood of live spruce trees to the penetration with the pilodyn needle according to the density and humidity of the wood. Roczniki AR w Poznaniu 241: 17–26.

Karaszewski, Z., Łacka, A., Bembenek, M., Mederski, P.S., 2016a: Wood damage and value loss of alder plywood harvested by Valmet 911.4 harvester with a 360.2 head. Sylwan 160(12): 1002−1009.

Nill, M., Kohnle, U., Sauter, U., 2011: Rindenschäden mit mutmaßlichem Bezug zur Holzernte im Spiegel der Betriebsinventuren in Baden-Württemberg. Forstarchiv 9(6): 216– 224.

Nuutinen, Y., Väätäinen, K., Asikainen, A., Prinz, R., Heinonen, J., 2010: Operational efficiency and damage to sawlogs by feed rollers of the harvester head. Silva Fennica 44(1): 121–139.

R Core Team, 2017: R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Accessed September 12, 2017. Available at: https://www.R-project.org/.

Simonov, M.N., 1984: Mechanizacija okorki lesomaterialov (Mechanisation of forest products’ debarking). Lesnaja Promyszlennost, Moscow, Russia.

Karaszewski, Z., Łacka, A., Mederski, P.S., Noskowiak, A., Bembenek, M., 2016b: Damage caused by harvester head feed rollers to alder, pine and spruce. Drewno 59(197): 77–88.

Sowa, J.M., Gielarowiec, K., Gaj-Gielarowiec, D., 2013: Characteristics and development of the construction of logging harvester heads. Forestry Letters 105(3): 57–76.

Kohnle, U., Kändler, G., 2007: Is Silver fir (Abies alba) less vulnerable to extraction damage than Norway spruce (Picea abies)? European Journal of Forest Research 126(1): 121–129.

Spinelli, R., Magagnotti, N., Nati, C., 2011: Work quality and veneer value recovery of mechanized and manual log-making in Italian poplar plantations. European Journal of Forest Research 130(5): 737–744.

Lee, K., Gibbs, J.N., 1996: An investigation of the influence of harvesting practice on the development of blue-stain in Corsican pine logs. Forestry 69(2): 137–141.

Mederski, P.S., Karaszewski, Z., Rosińska, M., Bembenek, M., 2016: Dynamics of harvester fleet change in Poland and factors determining machine occurrence. Sylwan 160(10): 795−804.

Moskalik, T., 2002a: Development of technics and technologies of mechanised wood harvesting. Sylwan (146)10: 31–38.

Moskalik, T., 2002b: Wood harvesting with harvesters in Poland. Sylwan (146)11: 103–110.

Moskalik, T., Borz, S.A., Dvořák, J., Ferencik, M., Glushkov, S., Muiste, P., Lazdiņš, A., Styranivsky, A., 2017: Timber Harvesting Methods in Eastern European Countries: a Review.

Croatian Journal of Forest Engineering 38(2): 231–241.

Murphy, G., Acuna, M., 2017: Effect of harvesting season, system and equipment on in-forest Pinus radiata bark removal in Australia and New Zealand. International Journal of Forest Engineering 28(1): 10–17.

Murphy, G.E., Pilkerton, S.J., 2011: Seasonal impacts on bark loss by mechanized processors in Oregon. International Journal of Forest Engineering 22(1): 35–41.

190

Tomczak, A., Jelonek, T., Zon, L., 2010. Comparison of selected physical properties of the juvenile and mature wood of Scots pine (Pinus sylvestris L.) from mature stands. Sylwan 154(12): 809–817.

Uzunović, A., Webber, J.F., Pearce, A.J., Dickinson. D.J., 1999: The role of mechanized harvesting in the development of blue stain in pine. Canadian Journal of Forest Research 29(2): 242–251. van der Merwe, J-P., Pulkki, R., Ackerman, P., 2015: Fibre losses during debranching and debarking of Eucalyptus pulp logs using a single-grip harvester. Southern Forests 77(4): 309–313.

Visser, R., Stampfer, K., 2015: Expanding Ground-based Harvesting onto Steep Terrain: A Review. Croatian Journal of Forest Engineering 36(2): 321–331.

Warkotsch, W., 1994: The Impact of Harvesting Operations on Timber Quality: Causes and Remedies. South African Forestry Journal 169(1): 33–48.

Więsik, J., Nurek, T., Dybcio, M., 2005: Badania procesu pozyskiwania drewna harwesterem na zrębie zupełnym (Tests on logging operations by harvester on the clear cut). Technika Rolnicza i Leśna 11: 25–28.

Croat. j. for. eng. 39(2018)2

Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes (183–191) Z. Karaszewski et al.

Authors’ addresses: Zbigniew Karaszewski, PhD.* e-mail: z_karaszewski@itd.poznan.pl Wood Technology Institute Wood Investigation and Application Department ul. Winiarska 1 60-654 Poznań POLAND Agnieszka Łacka, PhD. e-mail: agnieszka.lacka@up.poznan.pl Poznań University of Life Sciences Faculty of Agronomy and Bioengineering Department of Mathematical and Statistical Methods ul. Wojska Polskiego 28 60-637 Poznań POLAND

Received: September 12, 2017 Accepted: March 13, 2018

Piotr S. Mederski, PhD. e-mail: piotr.mederski@up.poznan.pl Mariusz Bembenek, PhD. e-mail: mariusz.bembenek@up.poznan.pl Poznań University of Life Sciences Faculty of Forestry Department of Forest Utilisation ul. Wojska Polskiego 71A 60-625 Poznań POLAND * Corresponding author

Croat. j. for. eng. 39(2018)2 191

Original scientific paper

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia Omar Mologni, Peter Dyson, Dzhamal Amishev, Andrea Rosario Proto, Giuseppe Zimbalatti, Raffaele Cavalli, Stefano Grigolato Abstract The forest industry around the world is facing common challenges in accessing wood fiber on steep terrain. Fully mechanized harvesting systems based on specialized machines, such as winch-assist forwarders, have been specifically developed for improving the harvesting performances in steep grounds. While the mechanization process is recognized as a safety benefit, the use of cables for supporting the machine traction needs a proper investigation. Only a few studies have analyzed the cable tensile forces of winch-assist forwarders during real operations, and none of them focused on large machines normally used in North America. Consequently, a preliminary study focused on tensile force analysis of large winch-assist forwarders was conducted in three sites in the interior of British Columbia during the fall of 2017.

The results report that in 86% of the cycles, the maximum working load of the cable was less than one-third of the minimum breaking load. The tensile force analysis showed an expected pattern of minimum tensile forces while the forwarders were traveling or unloading on the road site and high tensile forces when operating on steep trails, loading or traveling. Further analysis found that the maximum cycle tensile forces occurred most frequently when the machines were moving uphill, independently of whether they were empty or loaded. While the forwarders were operating on the trails, slope, travel direction, and distance of the machines from the anchor resulted statistically significant and able to account for 49% of tensile force variability. However, in the same conditions, the operator settings accounted for 77% of the tensile force variability, suggesting the human factor as the main variable in cable tensile force behavior during winch-assist operations.

Keywords: Steep slope harvesting, ground-based extraction, cut-to-length system, cable tensile force, winch-assist

1. Introduction Robust and sound forest engineering practices specifically developed for different terrain and stand characteristics are a crucial element of sustainable forest management systems; such practices must be technically feasible, economically viable, environmentally sound and socially acceptable (Heinimann 2004, ­Marchi et al. 2018). The forest industry around the world is facing common challenges in accessing wood fiber on steep terrain (Visser and Stampfer 2015, Mologni et al. 2016). Steeper slopes require motor-manual felling and yarding with systems such as cable (Cavalli 2012,

Visser and Harril 2017) and helicopter (Lyons and ­McNeel 2004, Grigolato et al. 2016), but these options are more expensive and much more hazardous compared to fully mechanized ground-based harvesting operations (Hert 2016).

Specialized forestry machines can often exceed the upper slope limits established in safety codes in many countries throughout the world (Alam et al. 2013, Visser and Berkett 2015, Session et al. 2017). Current regulations in the province of British Columbia (Canada), for example, restrict the use of ground-based logging equipment to slopes not exceeding 40% (B.C. Reg.

Croat. j. for. eng. 39(2018)2 193

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193â&#x20AC;&#x201C;204)

296/97) following specific steep slope protocols for stability and safety concerns. The slope is not the only limiting factor, however, and fully mechanized ground-based systems are often limited by other terrain factors, such as soil strength and/or roughness (Amishev et al. 2009).

There is a considerable interest and recent worldwide effort to improve traction of forestry machines when operating on steep slopes, especially in western North America (Amishev 2016). Various steep-slope harvesting machines with specialized undercarriages and carriers have been shown to safely access and operate on terrain up to 70% slope without external support or anchoring (Cavalli 2015). However, there is a limit about the physical feasibility of operating machines on steep slopes (Berkett 2012) that needs to be better defined and understood. A way to increase traction and stability on steep terrain is through assisting forestry machines by winch and cable to various anchor types. Options for extending mechanized forestry operations to steep slopes were examined during the 1970s through a feasibility study of a self-contained

cable tether system (McKenzie and Richardson 1978).

Steep terrain winch-assist machinery for forestry have been commercially available in Europe since the 1990s and since the early 2000s numerous commercial options have been developed for harvesters (Visser and Stampfer 2015). In New Zealand, the first winch-assist system was pioneered in 2006, while in North America, the first winch-assist unit was designed and manufactured in 2012.

The subsequent developments in purpose-built winch-assist machines over the last decade have led to the application of this concept as a well defined harvesting system (Cavalli and Amishev 2017) with potentiality for improving the efficiency of harvesting operations (Dyson and Boswell 2016), as well as for improving machine mobility and reducing soil disturbance through the reduction of slip (Visser and Stampfer 2015). However, even if scientific and anecdotal ­evidence provided for increased knowledge and understanding, there is still a limited quantitative framework with which to evaluate the relationship between cable tensile force, stability, ground pressures, and

Fig. 1 Location of study sites

194

Croat. j. for. eng. 39(2018)2

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

slip, especially in terms of site operative conditions and machine specifications (Sessions et al. 2017).

Forestry machines operating in steep terrain should be able to stop in full control at all times without reliance on the cable. Lots of winch-assist systems utilize alert devices which sound an alarm in the operator’s cab if the anchor moves. These devices are generally based on systems able to detect the anchor movements or the sudden drop of the rope tensile force, usually occurring when anchors have moved, or wire rope/connectors have failed (Amishev 2016).

However, there are currently no high-resolution onboard information systems installed to provide detailed information about the actual tensile force on the cable/s. Holzleitner et al. (2018) developed a scientific approach with a robust workflow for in-depth monitoring and analysis of cable tensile force for harvesters and forwarders equipped with integrated winch-assist systems. The Authors tested the methodology and presented results from a short-term study of tensile forces of harvester and forwarder operations in Central Europe. The forwarder analyzed by Holzleitner et al.

(2018), following the classification system reported in Brunberg (2004), was a medium-size machine (load capacity 12 ton). There is no evidence of research studies reporting tensile force monitoring on large forwarders (load capacity >14 ton).

The main objective of this study was to analyze the performance of large winch-assist forwarders equipped with integrated winches operating on ordinary harvesting sites in British Columbia, focusing on the tensile force monitoring of the support cables. A better understanding of factors affecting cable tensile force during harvesting operations was another objective of the study.

2. Materials and methods 2.1 Study sites and machine description The study was conducted in three different harvesting blocks located in the interior of British Columbia (Fig. 1), between Clearwater, Kamloops, and Prince George. All three harvested study sites were comprized of old-grow forests dominated by hybrid spruce (Picea glauca var. albertiana (S.Br.) Sarg.), subalpine fir (Abies lasiocarpa (Hook.) Nutt.) and lodgepole pine (Pinus contorta Douglas ex Loudon), with varying stand characteristics (Table 1). The study took place in ordinary harvesting operations where stands were clearcut using a combination of fully mechanized winch-assist and conventional ground-based operations. The winch-assist operations were limited to the

O. Mologni et al.

Table 1 Timber cruise stand and stem volume Parameter

Site A

Site B

Site C

Total harvested area, ha

50.2

17.1

116.2

Winch-assist harvested area, ha

29.0

9.7

81.0

-1

Wood volume, m ha

364

409

374

Average stem net merchantable volume, m3

0.63

0.45

1.10

Average DBH, cm

32.9

27.2

41.3

577.0

909.3

339.7

–

–

–

–

7

–

63

10

68

–

50

–

37

33

31

–

–

1

3

-1

Stand density, stems ha

Species composition by volume, % Douglas-fir Pseudotsuga menziesii (Mirb.) Franco Hybrid spruce Picea glauca var. albertiana (S.Br.) Sarg.

Lodgepole pine Pinus contorta Douglas ex Loudon Subalpine fir Abies lasiocarpa (Hook.) Nutt.

Western hemlock Tsuga heterophylla (Raf.) Sarg.

steepest sections of the blocks, interesting between 57% and 70% of the areas.

Three John Deere 1910E eight wheels winch-assist forwarders were monitored. The 21.8 ton forwarders have a maximum payload of 19.0 ton and are powered by a 186 kW engine. The cabs are self-leveling and rotate automatically following the crane movements.

A Haas synchro-winch was mounted on the rear frame of each forwarder. The capstan style winch has a drive drum which provides tensile force to the cable, separately from the storage drum. The winch provides a consistent pulling power synchronized with the forwarder wheel rotation. The operator can adjust the tensile force to ten different settings ranging from 0 to 90 kN. Forwarder maximum speed is 6 km h-1 when winch-assist is active. The store drum holds 400 m of cable with a diameter of 14 mm and a minimum breaking load of 211 kN. The Haas winch system, including the cable, weighs 1.9 ton.

2.2 Data collection The data collection was based on the cable tensile force monitoring, synchronized with the time and motion study and with the collection of the machine positions on known corridor profiles. The approach pro-

Croat. j. for. eng. 39(2018)2 195

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

posed by Holzleitner et al. (2018) for monitoring tensile force in winch-assist operations was adapted and integrated with the profile field survey and the load volume estimation. The study was conducted in the fall of 2017 in conditions of snow and freezing temperatures.

The cable tensile force was measured by a CableBull® SR22/800 XR sensor (manufactured by the ­Honigmann Industrielle Elektronik GmbH) positioned close to the anchor. The rated load capacity was 200 kN and the resolution was 0.0127 kN. The recording frequency could be set to four different modes 100, 200, 1000, and 2000 Hz. The survey frequency was set to 100 Hz. A diameter compensator enabled pre-calibrating the sensor to measure cable ranging from 14 to 22 mm in diameter. An analogic-digital converter unit was used to connect the sensor to a laptop where the data was recorded through the HCC-Easy software (version 6.02.23). The connections of the sensor and the converter were reinforced and isolated to protect the system from the elements and hard contact with the ground.

Two video cameras were installed in the forwarder cabs. One camera recorded the work cycles and motion elements and the second camera recorded the winch settings chosen by the operators. Both cameras were set to acquire videos at 720 ppm with 30 frames s–1.

A GNSS sensor, integrated and synchronized with an Inertial Measurement Unit (IMU) – composed by an accelerometer and gyroscope – on a single board microcontroller Arduino®, was installed in the forwarder cabs. This sensor measured the forwarder’s speed, position, and inclination at a registration frequency set to 5 Hz; as the cabs were self-leveling no machine inclination data was collected. High-quality photos were taken of each load to estimate forwarder load volume.

Both back and lateral photos were taken orthogonally to the machine, reducing to the minimum the distortions, when the cable was on the ground with no tensile force. Forwarding corridor terrain profiles were surveyed by measuring slope and distance using a Truepulse®200 rangefinder at each significant change in ground slope. Anchor positions and survey points were marked using the AvenzaMap® app. Anchor trees species, diameters, and heights were noted.

2.3 Data analysis The tensile force analysis was completed by combining recorded tensile forces with video analysis, load volumes, and corridor profile data. The clock time on the laptop recording the tensile force data was synchronized before each survey session using a web connection. The R software (version 3.2.4) was then

196

used to synchronize the two cameras and analyze the whole set of data.

The Analysis of Variance (ANOVA) was used for analyzing the differences between different work elements in both mean tensile forces and in average peak tensile forces, utilising the forwarder cycles as the observational unit. If the normality of the data distribution and the homogeneity of the variance assumptions were violated, non-parametric tests were used to perform the analysis. The relationships between the tensile force and potential influencing factors were tested through simple and multiple linear regression analyses at a significance level of 0.05.

The in-cab videos were analyzed identifying the forwarder travel direction – uphill, downhill or stationary – and the different work cycles, work elements, and delays following procedures defined in the Basic Time Concepts (Björheden 1991). If two elements overlapped, the activity with the higher priority (1: highest; 3: lowest) was recorded following the method described by Fernandez-Lacruz et al. (2013) and Erber et al. (2016), adapted to the present time study. The work elements were: Þ Travel empty: time spent moving (empty) to the loading site; starts when the forwarder wheels begin to rotate and ends when the boom begins to swing (priority 2) Þ Loading: time spent loading logs in one trip; starts when the boom begins to swing and finishes when the boom stops swinging (priority 1) Þ Driving – loading: time spent moving between the different loading spots; starts when the forwarder wheels begin to rotate and ends when the boom begins to swing (priority 2) Þ Travel loaded: time the loaded machine spends moving to the landing; starts when wheels begin to rotate and ends when the boom begins its swing (priority 2) Þ Unloading: time spent unloading logs at the landing; starts when the boom begins to swing and finishes when the boom stops its swing (priority 1) Þ Driving – unloading: time spent moving between different sort piles at the unloading site; starts when the forwarder wheels begin to rotate and ends when the boom begins to swing (priority 2) Þ Winch control: time the operator spent changing the force settings on the winch control panel (priority 3) Þ Delay: includes all delays less than 15 minutes.

Croat. j. for. eng. 39(2018)2

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

O. Mologni et al.

Table 2 Corridors features Slope, %

Corridor

Site

Cycles

Operator

Forwarding direction

Max

Horizontal length, m

Inclined length m

Mean

SD

1

A

6

A–B

Down

25.4

8.1

39.1

210.4

218.6

2

B

3

C

Up

36.6

19.0

75.8

238.6

261.3

3

C

2

D

Down

44.2

12.3

68.2

201.4

224.9

4

C

2

D

Down

40.6

12.2

58.0

208.7

232.3

5

C

4

D

Up

32.6

8.8

42.3

89.4

95.8

6

C

4

D

Down

36.1

19.6

55.0

204.4

225.7

7

C

5

D

Down

45.3

10.9

58.3

203.4

226.6

8

C

2

D

Down

37.1

19.7

66.4

200.0

220.5

Forwarder load photos were uploaded into CAD software, following a perspective correction through Adobe Photoshop® to reduce eventual minimal distortions. Field measurements of forwarder bunk dimensions were used for scaling the photos. Load volume was calculated from log diameters and average log lengths, measured directly in the CAD software, assuming the logs as pure cylinders. Profile data were analyzed through the R software to define terrain slope at any distance from the landings. The tensile force data was combined with the video analysis, the GNSS-IMU data, and the corridor terrain profile data to obtain a data set showing: the cable tensile force, clock time, metric coordinates, distance from the midroad (from which calculate the distance from the anchor), ground slope, machine speed, cycle number, and work element. The videos recording the operator settings were analyzed only for one of the operators working in two different corridors.

3. Results and discussion A total of 14.7 hours of forwarding operations were monitored, recording more than 5.3 million rows of tensile force data. The study included 28 forwarder loads driven by four different operators on eight different corridors. All operators were less than 40 years of age and had at least two years’ experience in winchassist operations. The average slopes of the corridors ranged between 25 and 45%, but is six of them, the maximum slopes exceeded 55%. The corridor lengths varied between 90 and 260 m. Downhill forwarding direction was the most commonly observed method (21 on 28 cycles). According to the operators, this was

because of lower fuel consumption, better traction, and overall improved efficiency compared to uphill forwarding. For this reason, most of the time cut blocks were planned with landings at the bottom and a road/ trail to the top for access to an anchor. Details on corridors features are shown in Table 2.

The average work cycle was 31.6 minutes long, with the loading element accounting for more than one-third of the total time (Fig. 2). Loading and unloading accounted for 52.2% of the total recorded time, while traveling was 44.8%. Forwarders operated for 72.0% of the time on trails and 28.0% of the time on forest roads.

The forwarder travel distances were measured as the horizontal distances from the forest road centre line in proximity to the trail, assumed as the reference starting point. The distance traveled on the trails, measured as the farthest loading point of the cycle, ranged from 42.5 m to 203.1 m, with an average of 149.1 m (SD 46.4 m). Forwarder distances traveled on the roads, measured as the farthest unloading point of the cycle, ranged from 14.0 m to 203.9 m, with an average of 67.8 m (SD 52.9 m).

The total wood volume forwarded during the study was 523 m3. Volume forwarded by corridor showed a wide range, from 28 to 110 m3, mainly because of different number of monitored cycles per corridor. The average wood volume extracted for each cycle was 18.7 m3 cycle-1 (SD 4.2 m3 cycle-1) ranging between 10.9 to 24.1 m3 cycle-1. A total of 1054 logs were counted and measured by photo analysis. The average log size was 0.50 m3, ranging from 0.13 to 1.60 m3. Table 3 shows details of forwarder travel ­distances and loads aggregated per corridor.

Croat. j. for. eng. 39(2018)2 197

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193â&#x20AC;&#x201C;204)

Fig. 2 Average duration of each work element. The error bars report the 95% confidence interval Table 3 Average forwarding distances and load features aggregated per corridors. Values between the brackets report the standard deviation Corridor Cycles

Mean trail distance (SD) m

Mean road distance (SD) m

Total wood volume m

Mean load volume (SD) m3

Mean log size (SD) m3

1

6

141.8 (52.8)

60.8 (26.5)

95.9

16.0 (5.0)

0.34 (0.26)

2

3

136.5 (49.4)

75.1 (21.7)

50.6

16.9 (4.8)

0.22 (0.06)

3

2

158.7 (37.3)

24.2 (4.4)

32.7

16.4 (1.9)

0.55 (0.11)

4

2

149.3 (36.6)

37.0 (9.3)

28.7

14.4 (4.2)

0.50 (0.10)

5

4

80.2 (9.6)

47.2 (10.5)

84.4

21.1 (3.4)

0.74 (0.22)

6

4

168.8 (34.6)

138.6 (81.2)

75.9

19.0 (2.5)

0.66 (0.13)

7

5

194.7 (3.2)

64.3 (75.4)

109.8

22.0 (2.6)

1.02 (0.45)

8

2

164.7 (19.0)

60.9 (12.5)

45.0

22.5 (2.0)

0.95 (0.54)

3.1 Working load of the cable and tensile force variability The tensile force was less than 30% of the minimum breaking load of the cable in almost all the cycles. In only four cycles, the peak tensile forces exceed onethird of the minimum breaking load for a total of just 7.3 seconds. The maximum working load (expression of the tensile force as percentage of the minimum breaking load of the cable) recorded in the study was 40.1% and it happened while a forwarder was traveling empty. The distribution of working load of the cable shows a characteristic bimodal shape (Fig. 3), similar to the working load distribution presented by Holzleitner et al. (2018). Low working loads occurred mainly during the unloading element at the landing,

198

where tensile force ranged from 8 to 15 kN. Higher working loads were measured during the work elements operating on steep trails, and in particular during the loading (Fig. 4), which was also the longest element in the average cycle.

For most of the work elements, cable tensile force ranged from 15 to 55 kN, with peak tensile forces occasionally exceeding 70 kN (Fig. 5). Minimum tensile forces, less than 10 kN, were recorded during unloading and traveling on the road. The winch control element occurred most frequently while the forwarders were operating in the loading area, setting the winch considering the trail conditions. Thus, tensile forces recorded during this element assumed values similar to loading and driving on the trails.

Croat. j. for. eng. 39(2018)2

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193â&#x20AC;&#x201C;204)

Fig. 3 Working load of the cable for the whole dataset The cable tensile force of a forwarder completing one cycle forwarding logs downhill is shown in Fig. 6.

The first five minutes of data show tensile force peaking and then receding to about 5 kN. This was because the forwarder had to return to the road and lower the tensile force as the cable was wedged in a stump. The

O. Mologni et al.

Fig. 4 Working load of the cable divided for the different work elements. Winch control and delay are excluded for better clarity of the graph sequence of spikes between the loading and driving on the trail is due to the non-perfect synchronization between the machine movements and the winch. Tensile force decreased by approximately 10 kN when the

Fig. 5 Rope tensile force distribution for each work element Croat. j. for. eng. 39(2018)2 199

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

Fig. 6 Rope tensile force plotted over time for a whole cycle. Continues line shows the distance of the forwarder from the mid road. Two dashed lines represent the safe working load of the cable (33 and 40% of the minimum breaking load) forwarder began moving uphill after being stationary.

Again, when the machine started moving downhill after a stationary element, there were positive spikes due to the machine moving in the opposite direction of the winch force.

3.2 Mean and peak tensile forces The highest mean tensile forces were recorded for loading and driving between different loading spots (driving – loading), considering both the whole set of data and the separate subsamples of downhill oriented cycles (logs forwarded downhill to the landing) and uphill oriented cycles (logs forwarded uphill to the landing). Lower and similar tensile forces were recorded for travel empty, travel loaded, winch control and delay. The minimum values were recorded for the unloading elements (unloading and driving – unloading) at the landing.

Regarding the average peak tensile forces, which represent one of the main concerns in tensile force analysis, the highest and similar values were recorded again during the work elements operating on steep trail and in particular during loading and travel empty. Travel empty also represented the element during which the absolute highest peak tensile force of 84.6 kN was recorded. However, analyzing separately the subsample of uphill oriented cycles, the highest average peak tensile forces were recorded for the travel loaded element. These considerations suggested the absence of correlation between the total machine weight and the tensile force, while the travel direction

200

prevailed. Table 4 shows the results of the KruskalWallis and Mann-Whitney-Wilcoxon non-parametric tests, which were applied instead of the ANOVA because of the violation of the normality of data distribution and the homogeneity of variance.

The frequency, with which a work element recorded the highest tensile force value in the cycle, changed Table 4 Tensile force data aggregated per work element Work element Travel empty

Mean tensile force, kN

Average peak tensile force, kN

Max peak tensile force kN

31.6 (14.85) ab

48.0 (17.4) a

84.6

a

69.9

b

50.6 (14.6)

Loading

38.5 (16.6)

Driving – loading

37.9 (16.4) b

46.7 (17.1) ab

69.6

Travel loaded

27.3 (10.9) a

46.2 (15.7) a

70.8

Unloading

8.1 (7.0) c

12.6 (13.4) c

53.1

Driving – unloading

6.8 (4.8) c

10.2 (10.2) c

41.4

b

68.2

35.8 (16.3) b

63.0

ab

Winch control

27.6 (14.4)

Delay

25.0 (15.6) a

37.5 (21.2)

»Mean tensile force« and »Average peak tensile force« represent the data aggregate per work element calculated as the mean of different cycles The »Max peak tensile force« represents the maximum tensile force recorded in the whole dataset for any work element (just one value for any work element) The standard deviation is reported between the brackets. Different letters close to the brackets represent statistical differences between the work elements

Croat. j. for. eng. 39(2018)2

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

O. Mologni et al.

Table 5 Correlation coefficients Coefficient

0.0270

228.905

<0.001

Slope

0.936

0.0001

1864.653

<0.001

–1.200

0.0147

–81.805

<0.001

3.644

0.0181

201.255

<0.001

–0.050

0.0000

–411.133

<0.001

Travel direction Uphill Anchor distance

3.3 Tensile force influencing factor analysis When the forwarders were operating on the trails, the slope, travel direction, and distance from the anchors (derived by the distance from the mid-road) resulted as the main significant variables influencing the tensile force and able to explain 49% of the data variability (F=929371.9, p<0.001, Adjusted R2=0.493).

p-value

6.172

Stationary

considering the forwarding direction. For uphill oriented forwarding operations, the maximum tensile force was recorded during the travel loaded element in 71.4% of the cycles. Instead, for downhill oriented operations, which represent the majority of the analyzed cycles, the elements which recorded the main frequency in reaching the maximum tensile force was travel empty (47.6% of the cycles), followed by loading (28.6%). Thus, in both the configurations, the maximum tensile forces were recorded while the forwarders were traveling uphill, during travel empty for downhill oriented operations, and during travel loaded for the uphill oriented ones. This suggested again that the total machine weight (and thus also the wood volume transported during the travel loaded element) did not influence the peak tensile force, while the punctual travel direction did. Indeed, considering the cycle as the observational unit, no statistically significant relationships were found between the maximum tensile forces and the load volume, nor for the forwarding direction (uphill or downhill orientation), and the maximum trail slope. Only the average trail slope was significant (F=22.1, p<0.001, R2=0.460).

t-value

Intercept

Travel direction

Fig. 7 Tensile force and settings relationships. Different sizes represent the point density. Dashed line represents the linear regression

Estimate of Standard error coefficient of estimate

Estimates, coefficients, standard errors, and significant levels are reported in Table 5. While increasing slope is easily connected to the tensile force increase, the increase in the tensile force during moving uphill could be associated with the necessity to contrast the slipping of the wheels. The reduction in the tensile force at an increased distance from the anchor could be mainly connected to losses due to the friction of the cable on the ground and stems, which exceed the effect of tensile force increasing due to an increased difference in altitude (rope weight effect).

In the subsample of data, where setting information have been analyzed, settings chosen by the operator were able to explain alone 94% of the tensile force variability measured at the anchor (F=26889960.5, p<0.001, R2=0.944), representing the main factor influencing the cable tensile forces. The differences between the recorded tensile forces and the operator settings were normally less than plus or minus 15 kN. Fig. 8 shows the prevalence of an extra tensile force ranging from 2 to 6 kN when the forwarders were operating at the forest road. Considering that most of the cycles were downhill oriented, the forwarders at the forest road were located at about 50 m less in altitude and about 200 m of distance. The weight per unit of length of the cable, multiplied for the difference in altitude and reduced by the partial force losses for friction on the ground (limited in case of very low tensile forces), caused this almost constant extra tensile force during unloading at the forest road. Limiting the analysis to the subsample of data related to forwarding operations on steep trails (thus excluding operations occurred when the machine was located on the forest road), the relationship between the tensile force and the operator settings (Fig. 7) decreases in strength but still maintains high determination (F=4131298.5, p<0.001, R2=0.774). The relationship increased when only the data related to active machine movements

Croat. j. for. eng. 39(2018)2 201

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193â&#x20AC;&#x201C;204)

uphill or downhill (F=3727998.5, p<0.001, R2=0.841) were considered.

The analysis of the main factors influencing the residuals of the tensile force-setting regression did not show a satisfying result because of the limited data available (one-day recording on two corridors). Slope, travel direction, and distance from the anchor were able to explain just 5% of the residuals variability (F=15560.4, p<0.001, Adjusted R2=0.049). However, the same limited dataset was able to explain only 15% of the tensile force variability (F=54052.9, p<0.001, Adjusted R2=0.152), compared to the 49% of the whole dataset on the trail. Thus, the analysis of operator settings, considered as the main factor influencing the cable tensile force in winch-assist forwarding, should be expanded for identifying the main influencing variables.

between the winch and the wheel movement was not perfectly matched.

Cable tensile force analysis showed an expected pattern of minimum tensile force while the forwarders were traveling on the road or unloading and high tensile forces when operating on steep trails, loading or traveling. Further analysis found that maximum cycle tensile forces were recorded most frequently during traveling uphill, empty or loaded, independently of the load, forwarding direction orientation, or maximum corridor slope. Only the average corridor slope was significant and able to explain 46% of the peak tensile force variability. Analyzing the set of data related to the forwarder operations on steep trail, slope, travel direction, and distance of the machine from the anchor resulted statistically significant and able to account for 49% of tensile force variability. However, in the same conditions, the operator settings account for 77% of tensile force variability, representing the main factor in determining the actual tensile forces.

Cable tensile force is an important factor in assessing long-term cable performance and developing safety guidelines for winch-assist harvesting machines. This study found that the tensile force settings chosen by the forwarder operators account for most of the tensile force variability, suggesting the need for considering the human factor in the safety procedures.

Deeper investigation should also be focused on the analysis of the relevance of the operator experiences in winch-assist operations and its relationship with the settings chosen. The present study involved four different operators; however, most of the cycles were monitored on a single operator (operator D) and all the operators had similar age and experience. Again, considering the fact that cable tensile forces never reached critical values, further research should consider the development of protocols to analyze localized cable damages and natural anchor stability, which could represent the worst safety limit in winchassist operation based on integrated-winch machines.

4. Conclusion

5. References

The study found that in 24 of 28 (86%) forwarder cycles, cable tensile forces were less than one-third of the minimum breaking load of the cables. This limit was exceeded for only 7.3 seconds during the 14.7 total operating hours monitored in the study. A series of short duration spikes in cable tensile force was observed when the forwarder was loading and then moved. This was attributed to a time lag between the time when the wheels started moving and the response of the winch, indicating that synchronization

Alam, M., Acuna, M., Brown, M., 2013: Self-levelling fellerbuncher productivity based on LiDAR-derived slope. Croatian Journal of Forest Engineering 34(2): 273â&#x20AC;&#x201C;281.

Fig. 8 Differences between tensile force measured at the anchor and tensile force set by the operator in the winch control panel

202

Amishev, D., 2016: Winch-assist technologies available to Western Canada. Technical report No. 37, FPInnovations, Vancouver, British Columbia (Canada).

Amishev, D., Evanson, T., Raymond, K., 2009: Felling and bunching on steep terrain â&#x20AC;&#x201C; A review of the literature. Harvesting Technical Note HTN01-07, Future Forests Research Limited, Rotorua, 10 p.

Croat. j. for. eng. 39(2018)2

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204) B.C. Reg. 296/97: Workers compensation act. Occupational health and safety regulation. Part 26 – Forestry operations and similar activities. Equipment operations – 26.16 Slope limitations.

Berkett, H., 2012: An examination of the current slope gradients being experienced by ground-based forest machines in New Zealand plantation forests. M.Sc. Thesis, University of Canterbury, Christchurch, New Zealand, 75 p.

Björheden, R., 1991: Basic time concepts for international comparisons of time study reports. Journal of Forest Engineering 2(2): 33–39.

Brunberg, T., 2004: Underlag till produktionsnormer för skotare [Productivity-norm data for forwarders]. Skogforsk, Uppsala, Sweden. Redogörelse 2004(3): 12 p.

Cavalli, R., 2012: Prospects of research on cable logging in forest engineering community. Croatian Journal of Forest Engineering 33(2): 339–356.

Cavalli, R., 2015: Forest Operations in Steep Terrain. Proceedings of the Conference CROJFE 2015 Forest Engineering – Current situation and future challenges, March 18–20, Zagreb, Croatia.

Cavalli, R., Amishev D., 2017: A case study: ground yarding in mountainous terrain. Proceedings of the 27th Club of Bologna Member’s Meeting – Session 3 KNR 3.2, November 12–13, Hannover, Germany.

Dyson, P., Boswell, B., 2016: Winch-assisted feller-buncher equipped with a continuous-rotation disc saw: short-term productivity assessment. Technical report No. 46, FPInnovations, Vancouver, British Columbia (Canada).

Erber, G., Holzleitner, F., Kastner, M., Stampfer, K., 2016: Effect of multi-tree handling and tree-size on harvester performance in small-diameter hardwood thinnings. Silva Fennica 50(1): Article id 1428, 17 p.

O. Mologni et al.

video camera system. Forest Science and Technology 12(2): 88–97.

Heinimann, H.R., 2004: Harvesting: forest operations under mountainous conditions. In: Jeffery, B. (Ed.), Encyclopedia of Forest Sciences. Elsevier, Oxford, 279–285 p.

Hert, R., 2016: Risk management issues for mechanized harvesting. Steep slope Conference, Richmond, British Columbia (Canada).

Holzleitner, F., Kastner, M., Stampfer, K., Holler, N., Kanzian, C., 2018: Monitoring cable tensile forces of winch-assist harvester and forwarder operations in steep terrain. Forests 9(2): 1–13.

Lyons, K.C., McNeel, J., 2004: Partial retention and helicopter turn volume. Forest Products Journal 54(1): 58–61.

McKenzie, D., Richardson, B., 1978: Feasibility study of selfcontained tether cable system for operating on slopes of 20–75%. Journal of Terramechanics 15(3): 113–127.

Marchi, E., Chung, W., Visser, R., Abbas, D., Nordfjell, T., Mederski, P.S., McEvans A., Brink, M., Laschi, A., 2018: Sustainable forest operations (SFO): a new paradigm in a changing world and climate. Science of the Total Environment 634: 1385–1397.

Mologni, O., Grigolato, S., Cavalli, R., 2016: Harvesting systems for steep terrain in the Italian Alps: state of the art and future prospects. Contemporary Engineering Sciences 9(25): 1229–1242.

Sessions, J., Leshchinsky, B., Chung, W., Boston, K., Wimer, J., 2017: Theoretical stability and traction of steep slope tethered feller-bunchers. Forest Science 63(2): 192–200.

Visser, R., Berkett, H., 2015: Effect of terrain steepness on machine slope when harvesting. International Journal of Forest Engineering 26(1): 1–9.

Fernandez-Lacruz, R., Di Fulvio, F., Bergström, D., 2013: Productivity and profitability of harvesting power line corridors for bioenergy. Silva Fennica 47(1): Article id 904, 23 p.

Visser, R., Harrill, H., 2017: Cable yarding in North America and New Zealand: a review of developments and practices.

Croatian Journal of Forest Engineering 38(2): 209–217.

Grigolato, S., Panizza, S., Pellegrini, M., Ackerman, P., Cavalli, R., 2016: Light-lift helicopter logging operations in the Italian Alps: a preliminary study based on GNSS and a

Visser, R., Stampfer, K., 2015: Expanding ground-based harvesting onto steep terrain: a review. Croatian Journal of Forest Engineering 36(2): 321–331.

Croat. j. for. eng. 39(2018)2 203

O. Mologni et al.

Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia (193–204)

Authors’ addresses: Omar Mologni, MSc. e-mail: omar.mologni@phd.unipd.it Prof. Raffaele Cavalli, MSc. e-mail: raffaele.cavalli@unipd.it Assoc. prof. Stefano Grigolato, PhD * e-mail: stefano.grigolato@unipd.it Università degli Studi di Padova Department of Land Environment Agriculture and Forestry Viale dell’Università, 16 35020 Legnaro ITALY Dzhamal Amishev, PhD. e-mail: dzhamal.amishev@fpinnovations.ca Peter Dyson, BSc. e-mail: peter.dyson@fpinnovations.ca FPInnovations 2665 East Mall Vancouver British Columbia V6T 1Z4 CANADA

Received: May 02, 2018 Accepted: June 13, 2018

204

Assist. prof. Andrea Rosario Proto, PhD. e-mail: andrea.proto@unirc.it Prof. Giuseppe Zimbalatti, MSc. e-mail: zimbalatti@unirc.it Università degli Studi di Reggio Calabria Località Feo di Vito 89122 Reggio Calabria ITALY *Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

A Mobile Hydraulic Winch for Use in Small-Scale Forestry Václav Štícha, Jaroslav Holuša, Roman Sloup, Jan Macků, Jiří Trombik Abstract Winches have recently been used to extract timber from forests. Winches are often components of tractors, but tractors cannot be used on difficult terrain and are generally too expensive for small forest owners. The current study considers the use of an experimental winch for the extraction of timber from small plots with difficult terrain. The mobile hydraulic winch used in this study weighs 50 kg and has a pulling force of up to 53 kN, a 12 V motor, and a 64x224 mm drum. The associated power unit is a gasoline, single-cylinder, four-stroke, air-cooled KIPOR KG 390D (400D), 389 cm3 engine, with 7.7 kW of power, and a torque of 22.6 Nm at 2500 rpm.

The engine powers a high-pressure oil pump with an output pressure of 3 MPa and a flow rate of 60 litres per minute. The input torque of the pump shaft is 25 Nm at 3000 rpm. The hydraulic winch can be fixed to tree trunks, stumps, or wooden pegs by lashings. The winch was tested at three locations with different assortments of wood. The results showed that the experimental winch was practical for timber extraction and that <250 kN of force was needed for successful extraction. At the test sites, the expense of lumber removal was on average 140% greater with the winch than with a horse but the advantage of the hydraulic winch is high pulling force. Because of its small size and low weight, the unit can be easily handled by two workers, easily moved at short distances in small plots with rough terrain, and easily transported among plots. With a one-man crew, the percentage of direct costs represented by wages dropped to 56%, and the percentage represented by fuel increased to 40%.

Keywords: farm-forestry, extraction, productivity, cost analysis

1. Introduction The technology used to transport timber out of forests, i.e., to extract timber from forests, mainly depends on the type of timber and the terrain. The force required can be generated by gravity, machinery, or by humans or various animals. In contrast to aerial extraction by helicopter, ground extraction is done by pulling or towing (Zloch 1971).

In recent years, extraction from larger forests plots on relatively even terrain has often involved the use of winches (Akay 2005, Russell and Mortimer 2005, Gellerstedt 1997), which are usually located on universal farm tractors or forestry tractors (Laurier et al.

2002). Although tractors with winches can be very useful, their manoeuvrability is inadequate for use in small forest plots, and they cannot be used on steep slopes (Akay 2005). Moreover, the owners of small forest plots seldom own tractors.

In many countries, forest properties are often small and divided among multiple owners. Currently, about 150,000 individuals own forest land in the Czech Republic. The average area of forest land owned per person is about 3 ha. Only 0.3% of the owners owned more than 50 ha of forest in 1990s (Jánský 2000). For such small owners, the use of large machinery is not economical. Even when the owner belongs to a cooperative, the fragmentation of the land often precludes the economical use of large machinery (Ottaviani Aalmo et al. 2016). At least in other countries, banks do not want to lend to small owners, and if they do, the owner often has a big problem repaying the loan (Mitchell-Banks 2001). For these reasons, small owners have become interested in buying smaller, more versatile equipment.

An alternative to tractor-mounted winches, are small portable winches (Visser and Stampfer 2015).

For example, the LD-52UV was introduced in 1952 and the VSKII was introduced in 1955. These winches,

Croat. j. for. eng. 39(2018)2 205

V. Štícha et al.

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

which had drums that could hold 350–400 m of 8 mm thick rope, were gradually replaced by iron horses (KAPSEN, MULA) but with short ropes (Balcar 2004).

As load manoeuvring is unimportant for thinning, modern mobile winches, such as the LPV-20 VNAD-2 VNAD-D, were developed. However, these machines cannot be used in areas with difficult terrain (Horek 1993). A basic limitation of portable winches is their low pulling force, which ranges from 7–10 kN (Neruda and Zemánek 2013). As a consequence, portable winches cannot be used to extract large logs or large bundles of timber.

For studying the resistance of trees to mechanical stress (Peltola et al. 1999, Gardiner et al. 2000), Macků et al. (2016) designed an experimental mobile winch with commonly available components. As assessing the resistance of full-grown trees to mechanical stress requires substantial force, the experimental winch had high power. The aim of the current study was to assess the use and costs of this experimental winch for timber logging in small-scale forestry.

2. Material and methods 2.1 Descriptions of the winch and power unit The experimental hydraulic winch and its power unit are described in Fig. 1. The hydraulic winch can be fixed to tree trunks, stumps, or wooden pegs with lashings (Fig. 2).

The winch can generate a 53 kN pulling force. The drum is 64x224 mm. The winch, which is equipped with both wireless and wired controls, weighs about 50 kg.

The power unit is a KIPOR KG390D (400D) one-cylinder, 389 cm3, four-stroke gas engine. The engine has a torque moment of 22.6 Nm at 2500 rpm. The nominal engine power is 7.7 kW at 3600 rpm. The engine powers a high-pressure oil-pump with an output pressure of 3 MPa and a flow of 60 litres per minute. The input torque of the pump shaft is 25 Nm at 3000 rpm (Fig. 2).

The mobile winch and the power unit were constructed as prototypes (Macků et al. 2016). The total cost was about 2000 EUR. Other equipment included

Fig. 1 Hydraulic winch (A) and power unit as seen from above (B)

206

Croat. j. for. eng. 39(2018)2

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

V. Štícha et al.

Fig. 2 Photos of different ways of moving logs with the experimental winch a wood splitting hammer, axe, chain saw, textile tiedown straps, strain gauge with datalogger, woodchuck, timer, notebook, and camera.

2.2 Study area To assess the use of the experimental winch for log extraction, five tests were conducted at three locations (Únětice, Liboc, and Svatý Jan pod Skalou) in the central Czech Republic (Fig. 3). Each test consisted of 10 »work cycles« (10 independent extractions). All three locations are at 250–350 m a.s.l. The tests were con-

ducted in dry weather between 10:00 am and 3:00 pm and in parts of the forest where timber has been felled and left in place. Background information on the tests, terrain, and the properties of the extracted timber is provided in Table 1.

In test C, the timber was manually extracted. The pile of firewood was formed of bolts of black locust (Robinia pseudoacacia) with a length of 1 m, an average thickness of 16.5 cm (range = 10 to 28 cm), and a moisture content of about 25–30%. In the other four tests, the timber was extracted with the experimental winch.

Croat. j. for. eng. 39(2018)2 207

V. Štícha et al.

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

Fig. 3 Localities where the experimental winch was tested for timber extraction; location of the study area Tests A and C, which involved the same location, slope, and timber properties, enabled a direct comparison of extraction by hand vs. by winch.

2.3 Data collection As noted in the previous section, 10 work cycles were performed for each test. Each work cycle consisted of the following separate operations: Þ rope unreeling (from starting the engine to stopping the unwinding) Þ load assembly (from preparing the chokers to moving personnel from the unsafe space) Þ winching (from the start of winching by controls until the end of winching)

Þ load release (from the release of chockers to the start of unreeling the rope again).

A strain gauge and datalogger were used to record the maximum and average pulling forces. The times required to perform all operations of the work cycle, from winch and power unit preparation to deposition of the timber at the roadside landing, were recorded.

2.4 Cost calculation The cost of extracting timber with the experimental winch was calculated. A standard method was used to calculate fixed and operating costs (Miyata 1980) but it was also modified as needed for the winch. In

Table 1 Background information on the tests and localities used for assessing the experimental winch (extraction distances are listed in Table 2) Latitude; Longitude

Slope, °

Tree species

Tree age

Length of timber, m

Date of experiment

A Únětice

50.1500703N, 14.3643139E

35

Black locust

60

1

March 29, 2016

B Únětice

50.1500703N, 14.3643139E

<5

Black locust

60

Bundle (Fig. 2 bottom left and bottom center)

March 29, 2016

C Únětice D Svatý Jan pod Skalou E Liboc

50.1500703N, 14.3643139E 49.9709597N, 14.1193853E 50.0991686N, 14.3172789E

35 <5 <5

Black locust Larch Spruce

60 130 120

1 4 (Fig. 2 bottom right) 8

May 19, 2016 May 1, 2016 June 10, 2016

Test and locality

208

Croat. j. for. eng. 39(2018)2

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

these calculations, service time of the winch was set at 500 h per year, the depreciation period was set at 10 years, and the salvage value was set at EUR 100. As the winch and power unit are relatively inexpensive (acquisition cost was about EUR 2000), the purchaser was assumed to pay the entire purchase price. There is no need to pay insurance or taxes on this type of device.

Wage costs were based on the average wage of workers, including other employee payments, which, with all other required payments, was about EUR 5.77 per hour in the Czech Republic at the time of the study.

That wage is lower than wages paid to workers in Western European countries (Eurostat 2017).

Fuel prices were based on the average prices in 2016. The average CZK exchange rate to EUR in 2016 was used (27 CZK = 1 EUR). Repairs and maintenance costs were obtained from logging machine operators and were set at 10% of the daily depreciation of the machine. General overhead expenses were set at 30% of direct expenses, which is the average value in forestry in the Czech Republic. The calculation includes both total costs and performance costs (total costs + profit) for other entities (with a 15% expected profit).

A standard 8-hour shift was used.

Based on the average times of individual operations (Table 2), the total time required to extract 1 m3 of timber was determined. The volume of timber extracted per shift was calculated based on the average time of extraction in each test. A 7-hour shift was assumed because 1 hour is required to prepare the equipment. The direct costs to extract 1 m3 of timber was then calculated.

Costs were compared for extraction using the experimental winch vs. a horse. In the case of horse extraction, actual contract prices of timber extraction in forest enterprises were used (School Forestry Com-

V. Štícha et al.

pany in Kostelec nad Černými lesy, Vojenské lesy a statky ČR, s.p., Forests of the Czech Republic, s.p.).

Cost was also determined when the experimental winch was operated by two workers in accordance with Executive Order (Nařízení vlády 2002) and by one worker.

3. Results Extraction distances ranged from 10–25 m (Table 2). Unreeling the rope was the least time consuming operation, followed by loading and forming the load.

In most localities, the most time-consuming operation was winching the load. The time required for winching depended on the load and the terrain. Winching a 4 m long log required an average pulling force of about 2 kN when the terrain was flat (test D) but almost 7 kN on a slope with 1 m long log (test A); winching an 8 m long log required over 20 kN when the terrain was flat (test E) (Table 2).

The extraction of wood by hand in test C (10.5 m on a 35° slope) required an average of 55.5 min, which was almost 5 times longer than when using a winch at the same location in test A (Table 2).

The cost per hour calculations (averaged over the four tests with winch) show that over 70% of the direct costs were for wages and 26% were for fuel. Other direct costs were relatively insignificant. With a two-man crew, the direct costs were 16.16 EUR per hr, and cost price (full standard cost + profit) per hr of operation, was 24.16 EUR/hr (Table 3). With a one-man crew, the total direct costs dropped to 10.39 EUR per hr, and the performance-related price was 15.53 EUR per hr (Table 3). With a one-man crew, the percentage of direct costs represented by wages dropped to 56%, and the percentage represented by fuel increased to 40%. The costs

Table 2 Time required for winching operations and pulling forces generated by the experimental winch in five tests Extraction Volume of Test and locality

distance, m load, m3

Time of

Time of load

Time of

Time of load

Total extraction

Logged pulling

unreeling, min

assembly, min

winching, min

release, min

time, min

force, kN

Average

Average/Minimal/Maximal

A Únětice

10.3±2.5

1.075

0:50/0:45/0:55

4:30/4:05/5:10

4:20/3:45/5:10

1:50/1:30/2:15

11:30/10:05/13:30 6.56/4.44/10.96

B Únětice

16±2

1.191

1:05/1:02/1:09

4:52/4:10/5:45

7:05/6:50/7:32

2:10/2:50/3:01

15:12/14:52/17:27

6.72/4.53/7.82

C Únětice

10.5±0.5

1.079

ns

ns

ns

52:10/55:05/58:20

ns

ns

15.1±2

0.4

1:02/0:56/1:07

2:10/1:50/2:44

7:02/6:55/7:25

0:33/0:30/0:35

10:47/10:11/11:51

2.37/1.5/3.4

25.2±4.4

0.6

1:33/1:24/1:39

2:05/1:41/2:36

9:10/8:43/10:15

0:35/0:30/0:40

13:23/12:18/15:10 17.35/16.16/20.15

D Svatý Jan pod Skalou E Liboc ns – non studied

Croat. j. for. eng. 39(2018)2 209

V. Štícha et al.

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

Table 3 Direct costs and operational costs of the experimental winch with one or two workers Economic indicator

One worker

Two workers

Investment costs, EUR

2000

2000

Salvage value, EUR

100

100

Lifespan, years

0.37

0.37

Expected yearly usage, hours

18.50

18.50

Fuel consumption, litres/hr

0.14

0.14

Fuel costs, EUR/litre

1.09

1.09

Gross pay-out, EUR/hr

4.31

4.31

Yearly depreciation, EUR/year

190

190

Depreciation, EUR/hr

0.38

0.38

Fuel, EUR/hr

4.15

4.15

Lubricant, EUR/hr

0.04

0.04

Repairs, EUR/hr

0.05

0.05

5.77

11.54

10.39

16.16

3.12

4.85

Full standard cost, EUR/hr

13.51

21.01

Expected profit (15%), EUR/hr

2.03

3.15

15.53

24.16

124.28

193.29

Pay of operators including mandatory fees, EUR/hr Total direct costs, EUR/hr Overhead expenses (30% of direct cost total), EUR/hr

Cost price (full standard cost + profit), EUR/hr Cost price per work shift, EUR/shift

for extracting bundles of timber because extraction of bundles by horse is not a common practice mainly because of occupational safety regulations (Executive Order Nařízení vlády (2002)) (Table 4).

For extraction of logs, cost price would be 3.5 times higher with the experimental winch (with a two-man crew) than with a horse (Table 4).

4. Discussion

of logging one timber pile by one worker are comparable to logging with a horse (on average only 9% higher). On average, timber extraction was 136% more expensive with the winch than with a horse (Table 3).

The calculations indicate that the cost price of extracting a 1 m3 bundle of timber is 70% higher with the experimental winch than with a horse (4.93 vs. 7.73 EUR/m3).

The cost price of horse extraction was not calculated

The results of our tests indicate that the mobile winch described here is practical for extraction of timber from small forest plots. In addition to its manoeuvrability, the main advantage of the winch is that it has sufficient pulling force to extract logs from difficult-toaccess sites and to extract large logs. The tests confirmed that the winch can extract both individual logs and bundles of logs.

Under extreme conditions (when the lumber is large, the total volume of extracted timber is small, or the plot has very steep slopes), it is usually necessary to involve winches for extraction. The construction and use of an aerial cable system is not practical for the extraction of small volumes of timber, and the use of animal (horse) power is limited by the small pulling force (Neruda and Zemánek 2013), which is 25–50 times lower than that of the experimental winch described here.

Due to its pulling force, the experimental winch used in our tests can extract bundles of timber uphill, although doing so it could damage roots and disturb the soil (Conway 1976). Soil disturbance can be eliminated by using a sulky (Spinelli and Magagnotti 2012).

The cost of extracting a bundle of timber is only 9% higher with the mobile winch than with a horse. On average, extraction of logs is 136% more expensive with the winch than with a horse. One reason why costs are higher with the winch is that the winching speed is low. For winching only 10.3 m, for example, required 11.5 minutes in test A (Table 3). Winching

Table 4 Direct costs and the cost price of the experimental winch per cubic meter of extracted timber and comparison with logging by horse Extraction distance m

Volume m3

Time total min

Volume per work shift m3

Direct cost EUR/m3

Cost price EUR/m3

Cost price using a horse EUR/m3

A Únětice

10.3

0.62

11:30

22.78

4.97

7.43

4.93

B Únětice

16.0

0.69

15:12

19.09

5.93

8.86

4.70

D Svatý Jan pod Skalou

15.0

0.40

10:47

15.58

7.26

10.86

2.70

E Liboc

25.0

0.60

13:23

18.83

6.01

8.98

2.70

Mean

16.6

0.6

12:43

19.1

6.0

9.0

3.8

Test and locality

210

Croat. j. for. eng. 39(2018)2

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

could be more efficient if winch speed was increased by the use of a different gearbox. Another reason why costs are higher with winching is that two persons are generally needed for practical and safety reasons. Although only one worker is needed to operate the winch, two workers are needed for preparing the winch and power unit.

A steel rope was used in the current study. The time and effect spent unreeling the rope could probably be reduced, and the distance of extraction increased, by replacing the steel rope with a synthetic rope (Magagnotti and Spinelli 2012). A DynaForce®plastic rope, for example, is lighter than a steel rope but has a higher load capacity (https://www.grube.eu/forestry/timber-harvesting/heavy-timber-felling/1596/ dynaforce-treehoist-rope). It is also safer than a steel rope. On the other hand, the purchase price is higher for a plastic rope and the working life may be shorter.

According to Spinelli et al. (2010), the purchasing cost for a tractor, a tractor-powered cable system (like the Savall 1500), and a self-powered cable system is about € 35,000, € 20,000, and € 16,000, respectively. The purchasing cost of the winch and power unit used in the present study was only about 2000 €.

According to Spinelli et al. (2016), managers of small forest plots are increasingly purchasing smaller, less expensive, and more versatile machines rather than heavy industrial equipment (Spinelli et al. 2016).

In addition to being useful for owners of small forest plots, the winch described here would also be useful for arborists. The arborist is usually transporting only a few logs and only for short distances but the logs are often large and, therefore, require a high pulling force (Neruda and Zemánek 2013). Farmers would also find the winch useful for both forestry and nonforestry jobs.

The advantages of the experimental winch and power unit tested here are their small size, low weight, and high pulling force. The power unit alone can be used to operate a small cable system or as a source of hydraulic pressure for a number of hydraulically powered tools and adapters such as chain saws, delimbing devices, woodsplitters, hydromanipulators, etc.

5. Conclusions The results indicate that the experimental winch described here will be useful for extraction of timber from small forest plots. Because of its small size and low weight, the unit can be easily handled by two workers (tree feller and winch operator), it can be easily moved at short distances in small plots with rough terrain, and easily transported among plots. The winch has sufficient power to extract logs substantially larg-

V. Štícha et al.

er than those that can be extracted with conventional portable winches. It can be used for extraction at distances up to 50 m without moving the winch.

For owners of small forest plots, the described winch should be useful not only for timber extraction but also for the transport of other loads. Relative to a tractor, the winch costs less to purchase and can be used in plots with steep slopes. Compared to other mobile winches, the described winch has considerably greater power but remains easy to handle and transport.

The wages of the two workers present over 70% of the direct costs in operating the experimental winch.

At the test sites, the expense of lumber removal was on average 140% greater with the winch than with a horse. When the winch is not being used, however, it does not require any additional costs (except for depreciation), while the horse must be fed, stabled, and cared for every day. The winch is, therefore, an excellent choice for small cooperatives with irregular work in the woods.

Acknowledgements This work was supported by research project NAZV QJ1520006 »Hodnocení rizika poškození lesních porostů větrem: vývoj a kalibrace národního prediktivního modelu« / »Assessment of the risk of damage to forest stands: development and calibration of the national predictive model«. The authors thank Dr. Bruce Jaffee (USA) for linguistic and editorial improvements.

6. References Akay, A.E., 2005: Using farm tractors in small-scale forest harvesting operations. Journal of Applied Sciences Research 1(2): 196–199.

Balcar, V., 2004: Mula do probírek. Lesnická práce 83(11): 46.

Conway, S., 1976: Logging Practices: Principles of timber harvesting systems. San Francisco. Miller Freeman, Inc., 432 p.

EUROSTAT, 2017: Labour Market. http://ec.europa.eu/eurostat/web/labour-market/statistics-illustrated. Accessed September 20, 2016.

Gardiner, B., Peltola, H., Kellomaki, S., 2000: Comparison of two models for predicting the critical wind speeds required to damage coniferous trees. Ecological Modelling 129(1): 1–23.

Gellerstedt, S., 1997: Mechanised cleaning of young forest – The strain on the operator. International Journal of Industrial Ergonomics 20(2): 137–143.

Horek, P., 1993: Technologické využití vyklizovacího navijáku Alpmobil. Lesnická práce 72(8–9): 245–248.

Jánský, S., 2000: Jak se staví současná lesnická politika k problemům drobných vlastníků lesa? Lesnická práce 79(5): 209–210.

Laurier, J.P., Baraton, M., Capelier, J., 2002: Machines de bûcheronnage: Panorama du parc français des matériels et

Croat. j. for. eng. 39(2018)2 211

V. Štícha et al.

A Mobile Hydraulic Winch for Use in Small-Scale Forestry (205–212)

examen de son évolution de 1980 à 2002. Projet SY55. Convention DGFAR/AFOCEL, France.

Macků, J., Štícha, V., Holuša, J., Trombik, J., 2016: Sestava hydraulického navijáku a pohonné jednotky pro přibližování dřeva, který byl v červnu 2016 zapsán na Úřadu průmyslového vlastnictví v ČR v rejstříku užitných vzorů pod číslem 29595, číslo spisu PUV 2016-32328, Czech Republic.

Magagnotti, N., Spinelli, R., 2012: Replacing Steel Cable with Synthetic Rope to Reduce Operator Workload During Log Winching Operations. Small-scale Forestry 11(2): 223–236.

Mitchell-Banks, P., 2001: Small-Scale Forestry in Canada: or Mammals Living Amongst Governments and Dinosaurs. In: Niskanen, A., Väyrynen, J., (Ed.): Economic sustainability of small-scale forestry: (EFI Proceedings No. 36), European Forest Institute, Joensuu, Finland, 41–50.

Miyata, E.S., 1980: Determining fixed and operating costs of logging equipment. General technical report NC-55. Saint Paul. Forest Service North Central Forest Experiment Station, 14 p.

Nařízení vlády, 2002: Nařízení vlády 28/2002, kterým se stanoví způsob organizace práce a pracovních postupů, které je zaměstnavatel povinen zajistit při práci v lese a na pracovištích obdobného charakteru. https://portal.gov.cz/ app/zakony/zakonPar.jsp?idBiblio=52785&nr=28~2F2002&r pp=15. Accessed December 10, 2011.

Neruda, J., Zemánek, T., 2013: Soustřeďování dříví těžební stroje. Mendelova univerzita v Brně, Czech Republic. http:// docplayer.cz/17029593-Soustredovani-drivi-tezebni-stroje. html. Accessed January 28, 2013.

Ottaviani Aalmo, G., Magagnotti, N., Spinelli, R., 2016: Forest Workers and Steep Terrain Winching: The Impact of Environmental and Anthropometric Parameters on Performance.

Croatian Journal of Forest Engineering 37(1): 97–105.

Peltola, H., Kellomaki, S., Vaisanen, H., Ikonen, V.P., 1999: A mechanistic model for assessing the risk of wind and snow damage to single trees and stands of scots pine, Norway spruce, and birch. Canadian Journal of Forest Research 29(6): 647–661.

Russell, F., Mortimer, D., 2005: A review of small-scale harvesting systems in use worldwide and their potential application in Irish forestry. Ireland, National Council for Forest Research and Development, 48 p.

Spinelli, R., Cacot, E., Mihelic, M., Nestorovski, L., Mederski, P., Tolosana, E., 2016: Techniques and productivity of coppice harvesting operations in Europe: a meta-analysis of available data. Annals of Forest Science 73(4): 1125–1139.

Spinelli, R., Magagnotti, N., 2012: Wood extraction with farm tractor and sulky: estimating productivity, cost and energy consumption. Small-scale Forestry 11(1): 73–85.

Spinelli, R., Magagnotti, N., Lombardini, C., 2010: Performance, capability and costs of small-scale cable yarding technology. Small-scale Forestry 9(1): 123–135.

Visser, R., Stampfer, K., 2015: Expanding ground-based harvesting onto steep terrain: A review. Croatian Journal of Forest Engineering 36(2): 321–331.

Zloch, S., 1971: Lesní těžba: učebnice pro žáky středních lesnických technických škol a lesnických mistrovských škol.

Prague, Státní zemědělské nakladatelství, 430 p.

Authors’ addresses:

Received: October 10, 2017 Accepted: January 26, 2018

212

Václav Štícha, PhD. e-mail: sticha@fld.czu.cz Asst. prof. Jaroslav Holuša, PhD. * e-mail: holusa@fld.czu.cz Asst. prof. Roman Sloup, PhD. e-mail: sloup@fld.czu.cz Jan Macků, PhD. e-mail: macku@fld.czu.cz Jiří Trombik, MSc. e-mail: trombik@fld.czu.cz Czech University of Life Sciences Prague Faculty of Forestry and Wood Sciences Kamýcká 129 165 00 Praha 6 – Suchdol CZECH REPUBLIC * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Productivity and Cost Analysis of Three Timber Extraction Methods on Steep Terrain in Thailand Nopparat Kaakkurivaara, Tomi Kaakkurivaara Abstract Steep terrain harvesting in Thailand has low productivity because of the shortage of suitable logging extraction methods. Common methods involve extraction using manpower on steep slopes where machines cannot operate. This study compared the utilization of log chutes against manpower and mule methods with regard to productivity and cost-efficiency in the same logging compartment in Northern Thailand. The extraction methods were divided into work elements and data were collected based on described work cycles. The log chutes clearly had the highest productivity (2.29 m3/h) compared to the other methods. The hourly cost was lowest using manpower and the highest cost was using the log chute. However, the unit cost indicated the most economic method was the log chute (THB 72.40/m3) and the least was using mule extraction. From a logging contractor point of view, the log chute method helps reduce the number of working days during the harvesting season and provides a higher profit for business.

Keywords: manpower, mule, log chute, cost-efficiency

1. Introduction Substantial amounts of Thai forests are located in mountainous areas due to the growing population causing flat and moderate grade land to be taken up for agricultural purposes. In the mountainous areas, forestry or nature conservation are primary types of land use. As a consequence of mechanization, the development of timber harvesting has been successful in recent decades when productivity has raised and costs have decreased. However, despite the general development of forest machinery technology, there is no harvesting method for steep terrain that could be applied in Thailand. Manual harvesting and extraction are still common when operations are executed on challenging steep terrain, where mechanized methods such as a skidder or a farm tractor cannot operate. In light of the above, it is important to investigate other solutions to improve the productivity and cost-efficiency of harvesting. Suitable harvesting methods for use in the Tropics and developing countries are characterized by many similar features (Sessions 2007, Heinrich 1987). There are four factors that influence the determination of suitable harvesting methods in

Thai forestry: low labour costs, limited investment willingness, lack of professional workers and seasonal harvesting.

Log harvesting in Thai forestry uses both tree length (TL) and cut-to-length (CTL) methods. TL is used with valuable teak and CTL with other low value tree species, which provide raw material sources for sawmills, pulp mills and power plants. Bucking is based on short dimensions for CTL, with only 1â&#x20AC;&#x201C;2 m log lengths (Manavakun 2014). The short dimensions of logs and pulpwood in Thai harvesting methods is due to the fact that felling, delimbing, cross cutting, extraction and truck loading are mainly carried out manually, which restricts the maximum weight of logs. With steep terrain harvesting, the slope gradient helps in moving the logs downhill with the assistance of gravity, making extraction using manpower easier.

The term Âťball hootingÂŤ has been used to describe rolling and sliding logs and pulpwood using manpower down the hillside to the landing area (Wackerman 1949). This work method is arduous and time consuming, especially when applied to a large amount of timber or where there is a long distance to the roadside.

Croat. j. for. eng. 39(2018)2 213

N. Kaakkurivaara and T. Kaakkurivaara

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

Manpower extraction has been studied on flat land in Thailand, where productivity was 3.28 m3/h (log size = 0.011 m3) for a working group of eight persons. In the study, ready cut eucalypt logs were extracted to small piles along the stump line, to then be loaded onto a truck; therefore, the carrying distance was shorter than the length of tree (Manavakun 2014). A similar working method was used in rubberwood plantations, where the productivity was 4.09 m3/h for a working group of four persons, the average log size was 0.038 m3 and the length of trees was 20.4 m (Chuayyok 2014).

Another non-mechanical method uses animals for extraction. Animal extraction is still a reality in mountainous areas in Asia, Latin America and Africa, where it may even be a future solution for a sustainable management of mountainous areas and contribute to the development of the livelihoods of local communities (Rodrigues et al. 2017). In Thailand, the only suitable animal is the elephant (Korwanich 1974). The feasibility of animal power should only be considered after obtaining detailed knowledge about local operation circumstances. It can have higher productivity (3.8 m3/h) than a farm tractor (2.8 m3/h) when the working conditions of extraction are advantageous and the working methods properly implemented (Melemmez et al.

2014). Elephants are used in teak plantations for TL harvesting on terrain where a skidder is not able to operate. The use of mules is not common in forestry in Thailand; instead, they are mainly used for the transportation of agricultural products or for tourism and as such they are only temporarily used in the forest to provide complementary income for owners.

Mule extraction is clearly a minor method compared to manpower extraction in Thailand. The use of mules spread widely before harvesting mechanization and they are well suited to hot weather, unlike horses (Brown 1950). In Thailand, mule extraction is only used for short logs, not for skidding long length logs to the landing area. A similar method was used with mules in a couple of studies in Iran. Ghaffariyan et al.

(2009) studied the circumstances on 30% and 35% slope gradients and reported productivity of 3.3 m3/h for a V-shaped saddle rack, 2.1 m3/h for a conventional rope tightened technique for extracting firewood, and 1.2 m3/h for a both-side-tightened technique for extracting pulpwood. Jourgholami (2012) studied the working cycle using mules and defined regression equations for lumber, pulpwood and fuelwood extractions, which predicted productivities of 0.84, 0.54 and 0.42 m3/h, respectively, when the extraction distance was 500 m. The determining factor in productivity was mainly the extraction distance and the loading time,

214

which were higher with pulpwood and fuelwood than with lumber. Productivity did not depend on the volume per hectare or stems size as much as on the distance and slope gradient (Sessions 2007).

A key factor in achieving cost-efficient and high productivity harvesting on steep terrain is the reduction of manual and labour-intensive work and the use of tools to achieve greater output. Not all harvesting methods are adaptable to Thailand. For example, cable logging would need a high level of professional skills to install and operate the system correctly, its high cost also being a hindrance (Studier and Binkley 1974, Sessions 2007). A considerable option for improvement in Thai logging is the log chute. Log chutes, originally made from wood for fixed installations and metal for portable chutes, use gravity to transfer logs from the logging area down to the landing area (Brown 1950).

Log chutes made from polyethylene plastic were introduced in the 1970s, and since then, the length and weight of one log chute unit has remained the same.

In FAO’s study (1979), productivity was 5–6 m3/h for a three person work group, when the slope gradient varied between 25% and 35% and the transfer distance was 100 m. Another FAO publication (1985) reported 1.06 m3/h productivity for thinning harvesting in Austria, when the installed log chute was up to 60 m long, the slope gradient was 20% and diameter at breast height of trees 19 cm. Raab et al. (2002) reported productivity rates of 1.8, 2.2, 3.2 and 3.7 m3/h for a three person work group when the average log size was 0.06, 0.12, 0.25 and 0.45 m3 (measured under bark), respectively. The Research Agency of the Forestry Commission from UK recommended minimum 22% and maximum 56% slope gradients. The productivity of two person groups was 2.59 and 0.90 m3/h, in two different studies, where the log size was 0.057 and 0.019 m3, respectively. (UK Forestry Commission 1994 and 2002). The reason for different productivity rates in transferring timber in 10 m log chutes was mentioned above. The log chutes were observed to be working well with firewood log lengths of 1–2 m in Turkey, where the recommended minimum and maximum slope gradient was 20% and 60%, respectively.

The productivity study revealed that the length or diameter of the log did not correlate with the transfer time. However, the length or diameter of log influenced the productivity as did the length of the log chute and slope gradient (Eroglu et al. 2007).

The aim of the current research was to investigate three extraction methods – dominant manual method, minor use of mule method and the new log chute method – by comparing their productivity rates, costs and economic profitability in thinning harvesting on Croat. j. for. eng. 39(2018)2

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

steep terrain. There are no other known comparison studies of these three extraction methods. The main task was to define productivity and to investigate whether the mule or the log chute methods were more efficient than manpower extraction.

2. Material and methods 2.1 Study area The study was carried out in Chiang Mai province, Northern Thailand, where the logging compartment was located in the forests of the Royal Agricultural Station Angkhang (19°914′N, 99°048′E). The study was carried out in August 2015. The average temperature was 28 °C during field work, and exceeded 30 °C during the daytime. The slope grade varied between 26.0 and 40.6%. The mean annual precipitation was about 2000 mm. The original vegetation type was determined as hill evergreen forest. The silvicultural treatment was a selection thinning to improve stand quality. The study area was part of larger forestry area, where annual timber removal was defined to be 250 m3. The amount of annual removal was used also for calculations of the annual work load of timber extraction in the case study, although the collected data did not cover 250 m3 of annual cutting.

N. Kaakkurivaara and T. Kaakkurivaara

2.2 Extraction methods In manpower extraction, the logs were moved using three different techniques. The simplest technique involved carrying the logs on a padded shoulder (Fig. 1).

With the two other techniques, gravity was exploited to move the logs. Ball hooting involved rolling a log downhill by kicking using a foot or by dragging them with a rope tied around the log. The work elements are described more precisely in Table 1. The average extraction distance was 25 m to the landing area.

Mules (Fig. 1) can work five days per week for a few hours per working day in the Tropics. During hot weather, full working days are not possible; the work rate needs to be calm to avoid overstress. In this study, three forest workers and a mule made up the team.

Three workers were required for the loading and unloading methods using a mule. In this method, the logs were loaded to a rack saddle, which was lifted onto the mule’s back by manpower. The workers also manually lifted down the rack saddle when unloading. Because of these work elements, the workers walked between the logging area and landing area along with the mule. The average extraction distance was 25 m for the mule. The work elements of this method are described more precisely in Table 1.

Table 1 Description of work elements for three extraction methods Method

Manual

Mule

Work element Walking

Begins when the worker starts walking towards the log to be moved and ends when the worker reaches the log

Hooking

Begins when the worker bends down to lift the log from the ground to the shoulder or attaches a rope around the log and ends when the worker is in standing position

Moving

Begins when the worker starts to walk with the log or drags the log or kicks the log in order to roll it down slope and ends when the worker stops at the landing area

Piling

Begins when the worker drops the log down or bends down to remove the rope from the log or move the log by hand and ends when the worker gets the log into the stack

Walking

Begins when the workers and mule walk towards the log and ends when they reach the log

Loading

Begins when the workers take down the rack and ends when the rack is loaded and tightened on the mule’s back

Carrying

Begins when the workers and mule start to move and ends when they stop at the landing area

Unloading Walking Lifting Chute

Description

Begins when the workers start to take the rack down and ends when the empty rack is placed on the mule’s back Begins when the worker starts walking towards the log and ends when the worker reaches the log Begins when the worker bends down to lift the log to the shoulder and ends when the worker is in a standing position

Carrying

Begins when the worker starts to walk with the log and ends when the worker stops next to the log chute

Sending

Begins when the worker starts to send the log to chute and ends when the log is in the chute

Stacking

Begins when the worker starts to take log from the end of the chute and ends when the log gets into the stack

Croat. j. for. eng. 39(2018)2 215

N. Kaakkurivaara and T. Kaakkurivaara

Productivity and Cost analysis of Three Timber Extraction Methods ... (213â&#x20AC;&#x201C;221)

Fig. 1 Three different timber extraction methods: a) Manual method, b) Timber extraction by mule, and c) Log chutes

With the log chute method, the study included the installation and take off time of the chutes. The additional time was only roughly estimated for this study based on field experiment, because this was the first time that log chutes were used in Thailand and hence experienced staffs were not available. The log chutes were made from split polyethylene pipes by sawing and drilling adequate amount of holes on both sides for attachment ropes. The total length of the installed

216

log chute was 48 m, where the average distance was 24 m. Two workers dispatched the logs and one worker stacked the received logs (Table 1).

2.3 Productivity The productivity of extraction work was studied based on a work cycle method. Every extracted log was measured on the landing area and the log volume was calculated. The time for one work cycle was meaCroat. j. for. eng. 39(2018)2

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

P= ∑

60x tadd +t Nlogs tot

Table 2 Cost factors for three extraction methods Cost factor

Fixed costs

sured using a stop watch (cmin) during the field experiment. Hourly productivity was determined using Eq. (1) for each extraction method. The productivity of manpower extraction was calculated from the time study data of the four different forest workers. Productivity rates for mule and log chute extraction were calculated for a three person working team. Additional time for log chutes was installation and de-installation times, which were divided to every log and allocated to work cycles.

(1)

The basic components of the cost calculation are presented in Table 2. The cost structure for manpower did not require detailed labour cost calculations. The direct salary cost is the only actualized cost factor without any indirect social security contribution or daily allowances. The forest workers were local, and their work contract ended after the logging operations were finished. The labour cost was also only a variable cost and was uniformly defined for all extraction methods. Relevant fixed costs were allocated to the mule and the log chute methods. Purchase price included necessary ropes for the mule and log chutes and a rack saddle for the mule. Salvage price was for recycling the plastic used in the log chutes. Overhead cost was calculated only for the mule extraction and covered normal pasture, feed pellets and shelter only when the mule was used for forest work, not for the whole year. Utilization time was based on the annual work load and productivity rate of this study. Duration of working day was restricted to be shorter for the mule than for humans.

The common machine cost model was used for cost calculations, which was adapted to the mule and log chute extraction methods with certain changes. The fixed cost calculation was determined using the following formulas. Eq. 2 included the average value of annual investment formula, which is generally estab-

Log chute

Unit

Purchase price

–

18,500

37,070

THB

Service life

–

10

10

a

Salvage value

–

15

1

%

Salvage value, SV

–

2755

371

THB

–

6

6

%

–

*

–

%

37.5

37.5

37.5

THB/hour

1

3

3

person/ method

7

5

7

hour/day

250

250

250

m3/a

Interest for capital

(Fixed+Variable) Hourly wage Variable costs

log volume, m

2.4 Cost calculation formulas

Mule

Overheads

3

tadd additional time for log chutes installation and take off, min number of logs in each installation Nlog ttot total time for one work cycle, minutes/work cycle.

Manpower

Int

Where: P productivity, m3/h x

N. Kaakkurivaara and T. Kaakkurivaara

Amount of workers Duration of working day Annual workload

* see equation 7.

lished for forestry (Miyata 1980, FAO 1992). It was used to determine cost of annual interest (Eq. 3). Annual straight-line depreciation was calculated as described in Eq. 4 (Kaakkurivaara and Korpunen 2017), whereas medicine costs of mule were estimated based on Eq. 5 (Rodriquez and Fellow 1986).

AVI =

Where: AVI PP SV SL

( PP − SV )(SL + 1) + SV 2SL

(2)

average value of annual investment purchase price salvage value service life.

C= AVI × Int

Int

100

(3)

Where: cost of annual interest CInt Int Interest percentage for capital average value of annual investment.

C Dep =

PP − SV

SL

(4)

Where: CDep cost of annual depreciation.

Croat. j. for. eng. 39(2018)2 217

N. Kaakkurivaara and T. Kaakkurivaara

CMed =

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

PP ×0.05 AW

(5)

Where: CMed medicine and veterinary service costs of mule per year AW annual workload.

For calculation of variable costs, labour costs were determined based on the working hours and the hourly wage (Eq. 6). The log chutes did not include overhead costs. Instead, mule extraction included an overhead cost, which took into account the annual quantity of working days in the forest (Eq. 7). It was essential to use the above-mentioned annual workload (m3) for calculating the cost per hour (Eq. 8) in order to make the calculation consistent with the case study situation. Furthermore, this cost calculation expressed expenses in Thai baht per cubic meter (Eq. 9).

Where: CLabor Cper Nper DH QD

CLabor = CPer × NPer × DH × QD

(6)

labor costs of working team per hour hourly salary number of workers duration of working hours per day quantity of working days per year.

CTotal= (CInt + C Dep + CMed

CHour = CTotal / AW

(7)

CHour

Where: CCubic cost per cubic meter.

P

Mule

Log chute

Minimum working time min/work cycle

0.14

9.53

0.25

Maximum working time min/work cycle

2.96

17.6

1.78

0.86

12.83

0.31

0.03

0.46

0.01

S.D.

0.48

2.09

0.15

Variance

0.23

4.37

0.02

357

21

790

0.30

0.45

2.29

N 3

Productivity, m /h

(8)

Where: CHour cost per hour.

CCubic =

Manpower

Average working time

Where: CTotal total costs per year OHC overhead cost percentage (%).

Table 3 Productivity of each extraction method, excluding delays (the manpower included one person, while the mule and log chute methods operated by three persons)

 OHC QDmin/work  cycle + CLabor ) ×  1 + ×  100 365   S.E.

 OHC QD  CInt + C Dep + CMed + CLabor ) ×  1 + ×  100 365  

spite of this, the working cycle of the log chute was the fastest. The maximum and average working times were shorter than the corresponding values measured for mule or manpower. The slowest method was clearly mule extraction. Slowness affected even the observed number of work cycles in the field survey. The collected data included only 21 work cycles for mule extraction, but several hundred from the manpower and log chutes. The number of observations on manpower extraction was determined for simultaneous work of the four forest workers in the logging area.

Productivity of the log chute was significantly the highest (2.29 m3/h), being five-fold compared to mule extraction and eight times higher than manpower extraction.

(9)

The operation times for the different extraction methods were calculated based on working hours and days, which was based on the annual work load in the Table 4 Annual operation time for three extraction methods based on annual workload of logging area (250 m3) and productivity rates in case study (the manpower included one person, while the mule and log chute methods operated by three persons)

3. Results 3.1 Productivity The productivity results of extraction are presented in Table 3. The log chute method included half hour for installation and ten minutes for taking it off. De-

218

Manpower

Mule

Log chute

Quantity of working hours hours/a

847.5

555.6

109.2

Quantity of working days days/a

121.1

111.1

15.6

Croat. j. for. eng. 39(2018)2

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

forestry area (Table 4). The log chute method would need about 109 working hours, which means about 16 working days. In contrast, the mule and manpower methods would need roughly seven times more working days to carry out extraction. The difference in the required working days between these two methods compensated for the two hour shorter working day of the mule even though mule productivity was higher than that of manpower.

3.2 Cost calculation The results of the hourly cost and unit cost for the manual, mule and log chute methods are presented in Table 5. As expected, the cheapest hourly cost was for manpower (THB 37.5/h), because it did not include any costs other than the direct salary cost. Mule extraction was over three times more expensive, as it involved the hourly wage of three workers and the costs for the mule, which together were THB 12.92/h. The hourly cost of the log chute was 4.5 times higher than for the manpower method. Nevertheless, the cheapest unit cost was recorded for the log chute (THB 72.40/m3).

The manpower method was almost double and the mule method was almost four times more expensive compared to the unit cost of the log chute.

Table 5 Extraction costs of three methods based on productivity study and cost calculation Manpower

Mule

Log chute

Hourly cost, THB/h

37.50

125.42

165.80

Unit cost, THB/m3

127.12

278.70

72.40

The results of the case study are presented in Table 6, which was used as an example to study the economic viability of the extraction methods. The felling cost and income were the same for all methods. These Table 6 Case study results (THB) of annual logging operation (250 m3) Manpower

Mule

Log chute

Felling cost

25,893

25,893

25,893

Extraction cost

31,780

69,675

18,000

Total cost

57,672

95,568

43,993

Income

75,000

75,000

75,000

Benefit-loss

17,328

–20,568

31,007

1.30

0.78

1.70

Benefit cost ratio

N. Kaakkurivaara and T. Kaakkurivaara

values were actualized values based on practice. The high cost of mule extraction resulted in an operating loss (THB –20,568), which means that even the income from timber selling cannot cover the costs. In contrast, manpower extraction produced almost the same amount as profit (THB 17,328). The log chute provided cost savings and the log chute method had the highest benefit-loss value (THB 31,007), which underpinned the excellent 1.70 benefit cost ratio.

4. Discussion This study collected for the first time accurate productivity data on manpower extraction in steep terrain in a Thai forest. The low productivity of manpower was a surprise. The extraction productivity of short logs using manpower was 0.30 m3/h in this study, which is about ten times lower than the manual method used with eucalypt logs on flat terrain (Manavakun 2014). Ball hooting did not seem to increase productivity, even though the physical stress was reduced by kicking and dragging logs downhill.

On the other hand, walking back uphill certainly did not help workers to recover before moving the next log down to the landing area. It seems that ball hooting was not an efficient method to extract short logs in tropical conditions.

In our study, mule extraction productivity was 0.45 m3/h, which was similar to that reported by Jourgholami (2012), whose results were between 0.42 and 0.84 m3/h. However, our results were less than reported in the studies by Melemez et al. (2014) and Ghaffariyan et al. (2008), who reported the productivity of mule extraction as 3.80 and 2.14 m3/h, respectively.

These higher rates may have been dependent on the distance, log size and extraction equipment. In our study, the loading and unloading of the saddle racks required plenty of time in every work cycle, even when three forest workers were involved in extraction. Two other factors influencing low productivity were the maximum carrying capacity and the lack of experience in forest work. An average mule can carry approximately 100 kg per work cycle. The mule used had been trained to transport agricultural products; therefore, it was not familiar with the work elements in log extraction.

The study revealed high productivity rates for the log chute method, even though the log chutes were used for the first time in Thailand during this study. The productivity of the log chute method was 2.29 m3/h, which was about the same as reported in the studies by Ghaffariyan (2014) and UK Forestry Commission (1994). The forest workers were mostly seasonal workers coming from rural villages, but this new method did

Croat. j. for. eng. 39(2018)2 219

N. Kaakkurivaara and T. Kaakkurivaara

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

not cause any trouble for them. On the contrary, the log chute method was the most favored by forest workers, because it helped to reduce work stress. The workers did not need to carry the logs at long distances.

The cost analysis showed clear differences between the extraction methods. Generally, the manual method was not the most cost-efficient when compared with the log chute method. The extraction operation could be carried out at half the cost by using log chutes. The literature review showed that Ghaffariyan (2014) reported USD 1.99/m3 (approx. THB 66/m3) unit cost for the log chutes, which was very close to our calculation of THB 72.40/m3. The very high cost for mule extraction (THB 278.70/m3) corresponded to the finding by Jourgholami (2012), where the extraction cost was USD 15/m3 (THB 495/m3).

This case study highlighted well the economic feasibility of extraction methods for Thai forestry using logging contract work. The normal use of manpower was not the poorest choice. It did not need any investment, and it provided quite a good profit. Mule extraction was not an option because the income did not cover the minimum salary for the workers. The working method should be developed to be more efficient either by reducing the necessary number of workers or by speeding up the work cycle. The mule itself was not costly; hence it may be reasonable to keep it when not working and when it is not needed for agricultural or tourism activities. The log chute was the fastest and most cost-efficient option. The low investment in log chutes allows a contractor to extract more cubic meters during the harvesting season and, with this method, the highest profit is achieved per cubic meter.

5. Conclusions The log chute method was the best choice for harvesting short logs in steep Thai terrain. Productivity and cost-effectiveness were significantly better than with manpower or mule extraction. The field study showed other aspects of log chutes. There are several advantages in using log chutes: easy installation, reduced work stress, reasonable price investment, expandable total length and minimal environmental disturbance. However, there are also some limitations in using log chutes, which should be borne in mind: danger zone around end point for workers, need of a rubber mat to protect the log when it stops at the end point, suitable only for short logs, cannot be used in flat terrain and is limited to downhill extraction if a winch is not supplied. Future studies should determine the maximum length of logs and implement log chutes in teak plantation logging.

220

Acknowledgements The study was financed by the National Research Council of Thailand. The authors would like to thank staffs at the Royal Agricultural Station Angkhang for providing the study area and conducting the field work.

6. References Brown, N., 1950: Logging, The principles and methods of harvesting timber in the United States and Canada. New York, USA. 417 p.

Chuayyok, S., 2014: Productivity and cost of para rubber logging operations in Surat Thani province. Masters thesis.

Department of Forest Engineering, Faculty of Forestry, Kasetsart University, Thailand, 137 p.

Eroglu, H., Acar, H., Ozkaya, S., Tilki, F., 2007: Using plastic chutes for extracting small logs and short pieces of wood, from forests in Artvin, Turkey. Building and Environment 42(10): 3461–3465.

FAO, 1979: Mountain forest roads and harvesting. Technical report of the second FAO/Austria training course on forest roads and harvesting in mountainous forest, Ort and Ossiach, Austria, 3 June – 2 July 1978. FAO Forestry paper No.14, Rome, Italy, 174 p.

FAO, 1985: Logging and transport in steep terrain. Report of the Fourth FAO/AUSTRIA Training Course on Mountain Forest Roads and Harvesting, Ossiach and Ort, Austria, 30 May to 26 June 1983. FAO Forestry paper No.14, Rev 1.

Rome, Italy.

FAO, 1992: Cost control in forest harvesting and road construction. FAO Forestry Paper No. 99. FAO, Rome, Italy, 106 p.

Ghaffariyan, M.R., Durston, T., Sobhani, H., Mohadjer, M.R.M., 2009: Mule logging in Northern Forests of Iran: A study of productivity, cost and danger to soil and seedlings.

Croatian Journal of Forest Engineering 30(1): 67–75.

Ghaffariyan, M.R., 2014: A short review of efficient groundbased harvesting systems for steep mountainous areas. Bulletin of the Transilvanija University of Brasow, Series II 7(2): 11.

Heinrich, R., 1988: Introduction to appropriate forest operations in supporting rural development. Appropriate Forest Operations, Proceedings of FAO/Finland training course Philippines 23.11.–11.12.1987. FTP Publications: Forest Harvesting 24. FINNIDA, Helsinki, Finland, 88–103 p.

Jourgholami, M., 2012: Small-scale timber harvesting: Mule logging in Hyrcanian forest. Small-scale Forestry 11(2): 255– 262.

Kaakkurivaara, T., Korpunen, H., 2017: Increased fly ash utilization – value addition through forest road reconstruction. Canadian Journal of Civil Engineering 44(3): 223–231.

Korwanich, A., 1974: Logging. Kasetsart University.

Croat. j. for. eng. 39(2018)2

Productivity and Cost analysis of Three Timber Extraction Methods ... (213–221)

N. Kaakkurivaara and T. Kaakkurivaara

Manavakun, N., 2014: Harvesting operations in eucalyptus plantations in Thailand. Dissertationes Forestales 177. Dissertation, University of Helsinki, Finland, 111 p.

Rodriguez, O., Fellow, A., 1986: Wood extraction with oxen and agricultural tractors. Forestry paper 49. FAO. Rome, Italy, 92 p.

Melemez, K., Tunay, M., Emir, T., 2014: A comparison of productivity in five small-scale harvesting systems. Smallscale Forestry 13(1): 35–45.

Sessions, J., 2007: Harvesting Operation in the Tropics.

Springer, Germany, 170 p.

Miyata, E., 1980: Determining fixed and operating costs of logging equipment. Forest Service, North Central Forest Experiment Station, St. Paul, MN, General Technical Report NC-55, 14 p. von Raab, S., Feller, S., Uhl, E., Schäfer, A., Ohrner, G., 2002: Aktuelle Holzernteverfahren am Hang. LWF aktuell, Magazin für Wald, Wissenschaft und Praxis. Nr 36, Bayerische Staatsforstverwaltung, Munich, Germany.

Rodrigues, J.B., Schlechter, P., Spychiger, H., Spinelli, R., Oliveira, N., Figueiredo, T., 2017: The XXI century mountains: sustainable management of mountainous areas based on animal traction. Open Agriculture 2(1): 300–307.

Studier, D.D., Binkley, V.W., 1974: Cable logging systems.

Division of timber management forest service, U.S. Department of Agriculture, Portland, Oregon, USA, 204 p.

UK Forestry Commission, 1994: Glenfinnan log chute. Forest Research Technical Note 6/94. The Research Agency of the Forestry Commission, 12 p.

UK Forestry Commission, 2002: Log chute extraction of a broadleaved crop. Information Note ODW 9.10, The Research Agency of the Forestry Commission, 4 p.

Wackerman, A., 1949: Harvesting timber crops. The American Forestry Series, McGraw-Hill Book Company, inc. New York, USA, 437 p.

Authors’ addresses: Nopparat Kaakkurivaara, D.Sc.* e-mail: ffornrm@ku.ac.th Department of Forest Engineering Faculty of Forestry, Kasetsart University 50 Ngamwongwan Rd., Chatuchak 10900 Bangkok THAILAND

Received: August 21, 2017 Accepted: January 19, 2018

Tomi Kaakkurivaara, PhD. e-mail: tomi.kaakkurivaara@luke.fi Natural Resources Institute Finland Kaironiementie 15 39700 Parkano FINLAND * Corresponding author

Croat. j. for. eng. 39(2018)2 221

Original scientific paper

Production of Wood Chips from Logging Residue under Space-Constrained Conditions Matevž Mihelič, Raffaele Spinelli, Anton Poje Abstract A study was conducted on chip production from logging residue left after a cable yarder operation. The logistics were managed with tractor and trailer units (shuttles). The study specifically dealt with a very difficult case of space constrained operations, further expanding the knowledge about chip supply in extreme work conditions. The focus of the investigation was also extended to the shuttles. The study tested a production chain, in which only 3 machines (1 chipper, 2 shuttles) were used to minimize operational costs. The use of 2 shuttles was decisive, reducing shuttle delays. The chips produced had an average moisture content of 40.2 ±3.1%.

Particle size distribution shows an unfavorable composition. The content of accepts is as low as 72%, while oversized particles get up to 5.4% and fines rise to a maximum of 24%. The estimated net productivity of the whole system was 11.5 t PMH-1, corresponding to a gross productivity of 11.1 t SMH-1. The cost of the whole operation amounted to 21.2 €t-1.

Keywords: Biomass; forestry; logistics; mountain; harvesting

1. Introduction Recent years have witnessed a global drive for sustainable energy. In European policy, measures have been adopted aiming to increase the use of renewable resources (Lindstad et al. 2015). The industry and markets have adapted to these new demands (Mihelič et al. 2015, Stupak et al. 2007). That explains the renewed interest for the recovery of logging residues as a renewable energy medium. The downside of logging residue is the fact that they are scattered on very large areas (Karpachev et al. 2017), which makes collection very expensive, especially given the low market value of the end product.

Logging residues left after whole tree extraction with cable yarders are a good option for wood biomass production, as the residues are pre-concentrated by the cable yarder (Spinelli et al. 2007). Here the problem is the terrain and the condition of forest roads in the alpine region. Poor road standards (steep gradient, narrow width, small turning radius) pose a serious challenge to the economy of the logistic operation.

Trucks with trailers are often unable to reach the

worksite, or there is not enough space on the road for parking a chipper and a truck with trailer. An option would be to use trucks with containers, but that is not a very popular option with contractors, as the tare weight of the truck is increased (Spinelli et al. 2014a).

Theoretically, the most efficient solution is to upgrade the road network, but that incurs significant costs and must be assessed very carefully due to its potential for environmental impact (Petković and Potočnik 2017).

That is especially true for hydro-geological impacts, because larger roads mean steeper banks, which are the origin point of much sediment production (Hernández-Díaz et al. 2015, Pičman et al. 2007).

There is one more option, however, which is popular with private contractors in Southern Europe, namely: using a farm tractor and a trailer for the intermediate transportation of chips from the chipping pad to the roadside landing. Once there, chips are dumped onto the ground and loaded onto truck and trailer units with a separate loader, or with a front-end loader mounted on a tractor. The advantage of this system is flexibility, as tractors with trailers are very maneuverable and can easily navigate steep and narrow

Croat. j. for. eng. 39(2018)2 223

M. Mihelič et al.

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

roads. Furthermore, tractors with trailers are not special machinery and they are widely available at local farms. The system is also less dependent on the arrival of trucks because the chip pile built at the roadside landing serves as a buffer, so when the truck arrives it can be readily loaded (Marchi et al. 2011).

logging residues from a spruce-dominated stand. The final objective was to provide recommendations to operation managers, for deciding if and when this operational mode is economically feasible in their specific working conditions.

However, this system also has its drawbacks, and namely: additional persons have to be present; there is a need to store the piles of chips; additional work phases are included (load transfer, i.e. dumping and reloading). Furthermore, product losses occur almost inevitably, because the landings are usually unpaved and the front end loaders are incapable of retrieving all of the chips dumped onto the terrain, without incurring severe product contamination. In fact, some level of product contamination is most likely to occur, especially if the ground on the landing is rough.

2. Materials

At any rate, this system is often used, but the studies about it are still relatively rare. Therefore, the productivity and the cost of such operation were investigated, including all main processes, from chipping to loading on road trucks. The study also included an analysis of the quality of chips (moisture content, particle size distribution and bulk density) derived from Table 1 Description of the site Height above sea level, m Latitude (WGS84)

1350 46°43’42”

Longitude (WGS84)

14°26’24”

Species

Spruce, beech, maple, larch

Average inclination, %

89

Association

Rhodothamno-Rhododendretum hirsuti

Rockiness, %

45

Total study time, SMH

4

Productive time, PMH

2

Loads produced, n°

5 3

Volume produced, m

126

Chips produced, t

47.5 -3

Bulk density, kg m Feedstock type

Moisture content, % Site organization

377.5±20.93 Top & lop, slash 40.2±3.11 Large pile, very limited space

Notes: m3 – cubic meters of loose chips t – fresh tones PMH – productive machine hours excluding all delay time SMH – scheduled machine hours, including all delay time

224

2.1 Study area The study consists of a case study in the Northern Slovenia. The attributes of terrain, landing characteristics and feedstock type were considered representative for the conditions commonly encountered by wood chipping contractors working with forest residue chipping in mountainous conditions. The characteristics of worksite and feedstock are shown in Table 1.

The chipper used the same work settings during the entire study. In particular, cut length and screen size remained the same (Eliasson et al. 2015, Facello et al.

2013).

2.2 Supply process Piles of forest residue located at cable yarder landing were used for the production of green chips. The piles of slash were piled using hydraulic crane of the cable yarder, following a whole tree cable yarder operation. The operation was a cold deck operation, meaning that residue recovery occurred after the yarder had completed its work and had been dismantled, in order to minimize interruptions and delays of yarding and chipping operations. The truck-mounted chipper was located next to the side of the pile. Two tractor and trailer units acted as shuttles, which were directly fed by the chipper and the chips where transported to the roadside landing. The tractor and trailer units are referred to as shuttles. When chipping was finished, one tractor remained on the road side landing waiting for the trucks to arrive, and then proceeding to load them with a front-end loader attachment. All tests were conducted in early December 2011.

Three workers were on-site during the trial, because the extreme conditions of the site required one driver to be available for each machine. Too much maneuvering was involved for one single tractor driver to manage both shuttles. The chipper operator was also the machine owner and contracting firm manager.

He was a trained and well experienced professional, who was very proficient with his job and equipment.

The drivers of the shuttles were also experienced professionals, well used to work in such conditions.

Slash piles were located below the forest road next to the landing site. The residues consisted of slash, Croat. j. for. eng. 39(2018)2

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

tops, broken or otherwise damaged parts of logs and branches. Residues had been stacked randomly, as trees were brought to the forest road and processed.

The topping diameter was 10 cm.

M. Mihelič et al.

Table 2 Main chipper characteristics Truck manufacturer

Mercedes-Benz

Truck model

Actros 2640

2.3 Machines used and machine configurations

Truck engine type

The chipper used in the operation was a truckmounted Starchl MK74 600. The machine was a standard model, but had been adapted for work with residue piles, as the contractor’s primary business was chipping the logging residue left at the side of forest roads after cable extraction (Fig. 1). The machine was new, and it had just been acquired by the contractor at the time of the study, featuring a detached cab (Poje et al. 2018). Adaptation consisted in the fitting of a loader with the maximum possible reach for the available truck size. The loader mounted a FG31R grapple, designed for handling contaminated logging residue, with open tines. The chipper was mounted on a truck and powered by its own independent engine with total power of 242 kW, which categorized the machine into the larger family of industrial chippers (Spinelli and Hartsough 2001). Main chipper characteristics are presented in Table 2.

The truck chassis was supported by 3 axles. The power was transmitted to rear two axles of the chassis, with locking differentials. The machine was quite heavy and axle weight was close to the maximum Slovenian

Truck engine power Chipper manufacturer

Mercedes Benz V6, OM 501 LA II/3 298kW at 1800 rpm Fa. Helmut Starchl Hackmaschinenbau

Chipper model

Starchl MK 74 600

Chipper – engine type

CAT C9 inline 6-cylinder, 4 stroke diesel

Chipper – engine power

242 kW, 1450 Nm

Weight

24,500 kg

Length

9.5 m

Width

2.55 m

Height (transport position)

4.0 m

Crane

Epsilon Palfinger M110L

Crane reach

10.7 m

Grapple

Epsilon FG31R

Cabin for machine operator

Epsilon Epscab CAM/CAE Chipping

Closed drum with 12 chipping knives In-feed size

740×450 mm

Screen size

100×100 mm

Drum diameter Drum turning speed

610 mm Up to 500 revolutions min-1

Note: Data provided by the manufacturer and contractor

legal limit for road traffic. However, the 3 axle configuration was deliberately chosen by the contractor, as it maximized maneuverability. The minimum turning radius was 9.9 m, which allowed negotiating winding mountain roads. A 4-axle configuration would have been limited by a much larger turning radius.

Fig. 1 The Starchl MK74 600 mounted on a truck chassis

Traction power was delivered to both rear axles.

The 6×4 configuration is better than the 6×2 option, because driving on steep and often unpaved mountain roads requires higher traction capacity than a singleaxle drive configuration is able to provide. On the other hand 8×8, or 6×6 drive options may result too heavy and expensive for producing a low-value assortment. Therefore, the contractor deemed that a 6×4 configuration offered the right trade-off between mobility and maneuverability in such conditions.

Croat. j. for. eng. 39(2018)2 225

M. Mihelič et al.

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

The blower enabled discharging chips to three machine sides: front, rear and port side. The starboard side was unavailable for chip discharge because it carried the chipper in-feed and it always faced the wood piles, which made it impossible to park a container on that side (Spinelli et al. 2015).

When the study started, the truck had 4500 work hours and the chipper 400 h on its meter.

2.4 Supply chain and logistics layout The route from the main public asphalt road to the cable yarder was gravel road, 550 m long, with no place for trucks to turn. The landing for chips was a widening of the main asphalt road and usually served as a parking lot; therefore, the distance from the slash pile to the landing site was the same as the distance between the slash pile and the main road. The logistics of roundwood transportation were extremely strained, as the road was steep, narrow and several hairpins had to be navigated. Therefore, log trucks had to detach their trailers, back up all the way up to the yarder pad, load the logs and return to the roadside, where the logs were transferred onto the trailers. The lessons learned in the transport of roundwood made it clear that the logistics of chip production could not be based on truck and trailer units. Instead, lighter and more agile tractor-trailer units became the contractor’s ve-

3. Methods

Table 3 Main tractor and trailer characteristics Machine Tractor manufacturer Tractor – model Tractor – engine type Tractor – engine power

Configuration 1

Configuration 2

Lindner G.m.b.H

Deutz Fahr

GeoTrac 103

6160

Perkins

Deutz TCD 4.1

74,5kW at 2200 rpm 120 kW at 2100 rpm

Trailer manufacturer

Stetzl

Fliegl

Trailer – model

TK13

Gigant ASW 160

Weight tractor + trailer empty

3720+2950

5670+4200

Permitted total weight of trailer

13,000 kg

16,000 kg

Length tractor + trailer

4550+3437

5600+4772

Width (maximum)

2150

2380

Height (maximum)

2453

3050

Type of discharge

Tipper

Push off

Volume of trailer

226

3

25 m

hicle of choice for managing the supply chain logistics (Table 3). The shuttle units were representative of two standard types often used in this type of operations.

Configuration 1 (Lindner-Stetzl) was smaller and lighter than configuration 2 (Deutz-Fleigl). The trailer in configuration 2 was a new generation push-off model, with built-in hydraulics that substantially shortened unloading time. There were several advantages to the push-off trailer over a conventional tipper trailer, and namely: better stability and capacity to unload in buildings with low ceilings. Such trailers can also compact the load to increase payload when dealing with low-density material, but this option was not used in the study. The drawbacks of the push-off trailer were its higher price and heavier tare weight.

The propulsion of both tractors was in 4×4 configuration, while trailers were without propulsion. The narrow width of the shuttle units and their short turning radius combined to achieve better maneuverability than any truck and trailer convoy could. Chips were blown into the shuttle units, which shuttled them to the roadside landing, for dumping onto a large pile built on the ground. Chips were then re-loaded onto road convoys using a Quicke Q75 front-end loader attachment installed on one of the tractors. In the case of the study, both chip contamination and losses were minimal, as the roadside landing was asphalt.

27 m3

A typical time study (Magagnotti and Spinelli 2012) was performed using a handheld computer, running the dedicated Laubrass UMT Plus time study software. One researcher has timed the chipper and the other the shuttles. The snap-back timing method was used. The study was designed to evaluate chipper and shuttle productivity and identify the variables that were most likely to affect it. Timing operations can be seen in Table 4.

Timing sessions lasted the entire workday. The purpose was to obtain a good representation of the structure of a typical workday, subdivided into different productive and non-productive activities (Bjorheden and Thompson 2000). Productive time was separated from delay time. All delays were included in the study, and not only the delays that were below a set duration threshold. Such practice could misinterpret the incidence of downtime, especially on comparatively long observation periods (Spinelli and Visser 2008). Delays caused by the study itself were separated and excluded from the data set. The filling of one trailer was considered as one work cycle. Study time was divided into defined time elements, according Croat. j. for. eng. 39(2018)2

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

Table 4 Operations of time studies, their short description and classification

Description of operation

Classification of time according to Bjorheden and Thompson (2000)

Chipping

Chipping of material

WP, WT, PW, MW

Feeding

Adding material on the conveyer

WP, WT, PW, MW

Placing the machine

WP, WT, PW, CW

Organizing the pile and picking up leftover material

WP, WT, PW, CW

Main (meal) break

WP, NT, WD, ME

Delays because of worker

WP, NT, WD, RP

Operational delay

Delays because of organization

WP, NT, WD, IT

Mechanical delay

Delays because of the machine

WP, NT, SW, ST

Preparation time

Time it takes to set the machine WP, WT, SW, PT, CO up and take down time

Operation

Positioning Sorting Meal break Personell delay

Driving on road Driving on landing

Driving on road

WP, WT, SW, PT, RL

Driving on skid trail/landing

WP, WT, SW, PT, OP

to the IUFRO classification (Bjorheden and Thompson 2000) and following the harmonized European guidelines (Magagnotti and Spinelli 2012). The operator performance was not normalized by means of productivity rating. All observed cycles were included into the master database.

Mass output was determined by taking loads to a certified weighbridge. Volume output was estimated by measuring the internal volume of trailers, and visually assessing the volume of mounds or voids on trailer tops. Two 500 g samples were collected from each trailer in order to determine particle size distribution and moisture content. Each sample was composed by mixing subsamples collected at different points from the container top. Sample size for determination of particle size distribution was determined according to standard (EN15149-1:2010). Moisture content was determined with the gravimetric method, according to European standards (EN14774-2, 2010). Fresh sample weight was determined immediately after sample collection with a portable scale, to avoid the bias caused by moisture loss during storage and transport to the laboratory. Particle size distribution was determined with the oscillating screen method, according to European standards (EN15149-1, 2011).

Fuel input was estimated by measuring the quantity of fuel used after the machine was refueled. This was the total fuel consumption for the mix of produc-

M. Mihelič et al.

tive work, delays and relocation occurred during the study.

Machine costs were calculated with the method developed within the scope of COST Action FP0902 (Ackerman et al. 2014). It was assumed that the machine worked 2000 scheduled machine hours (SMH) per year, over a depreciation period of 5 years (amounting to a service life of 10,000 scheduled machine hours). Labor cost was calculated at 20€ SMH-1, inclusive of indirect salary costs. The costs of insurance, repair and service were obtained from the contractor. The investment price includes 22% VAT. The calculated operational costs were increased by 25% in order to include relocation, overheads and administration costs (Table 5).

Study data were analyzed with MS Excel in order to extract descriptive statistics and check for the statistical significance. Because of the interdependent nature of particle size distributions, basic approaches of compositional data analysis were used (Aitchison 1986). Compositional data were transformed using Isometric Log Table 5 Estimated cost of machines involved in the operation Chipper

Configuration 1

Configuration 2

Loader

260,000

80,000

92,449

92,449

Investment – attachment EUR

–

16,000

24,278

–

Resale value (20%), EUR

52,000

19,200

23,345.4

18,489.8

Number of working days, Days per year

200

200

200

200

5

5

5

5

1600

1600

1600

1600

4

4

4

4

Investment – base machine, EUR

Service life, Years Utilization, SMH/year Interest rate, % Depreciation, EUR/year

32,640

8693.15 11,707.08 9221,02

Interests, EUR/year

6888

2247.42

3032.13

2399.74

Insurance, EUR/year

5772

1400

1500

1500

Diesel, EUR/year

34,560

12,288

13,824

13,824

Lube, EUR/year

6912

2457.60

2764.80

2764.80

Maintenance, EUR/year

20,800

5535.62

7470.53

5916.74

Total, EUR/year

107,572 32,621.78 40,298.55 35,626.29

Total, EUR/SMH

67,23

20,38

25,18

22,27

Crew, n

1

1

1

1

Labor, EUR/SMH

20

20

20

20

Overheads (25%), EUR/ SMH

23,05

11,34

12,54

11,81

Machine rate, EUR/SMH

115,29

56,73

62,73

59,07

Croat. j. for. eng. 39(2018)2 227

M. Mihelič et al.

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

Ratio transformation in the CODAPAC software (Comas-Cufí M. and Thio-Henestrosa S. 2011), and the ternary diagram was plotted in R (Team 2008).

Shuttle utilization amounted to 76%, with very little interaction delays. This included driving and unloading. The utilization of shuttle 1 was 4.6% lower than that of shuttle 2, due to a different unloading system.

4. Results

Loading of chips, dumped on the ground (asphalt landing) by shuttles, into container trucks required moving chips over an average distance of 7.5 m, or a maximum distance of 15 m. The distance depends on the distribution of dumped chips on the landing.

Loading productivity was very high and averaged 112.9 t PMH-1, or 89.9 t SMH-1. The time it took to clean the roadside landing is included in this figure.

The mechanical availability (i.e. total worksite time minus mechanical delays, divided by total worksite time) of the chipper was very high and amounted to 99.6 . Utilization (the ratio between productive time and total worksite time) of the chipper was 50.4%, which corresponded to a 74% delay factor (Spinelli and Visser 2009).

A detailed overview of worksite time (Fig. 2) shows that waiting accounted for 46% of the total time on site.

Waiting generally occurred because no shuttles were available on site, or they were being maneuvered into place. Apparently, two shuttles were not enough to keep the chipper busy, despite the relatively short extraction distance. However, constrained space and the absence of any space for exchanging would have made it very difficult to introduce a third shuttle.

Chipping time represented 33% of total worksite time. The average chipping time was 37.5 ±3.45 s m-3 (loose volume). Feeding time represented 16% of total worksite time, or 17.8 ±5.61 s m-3 (loose volume).

Pure chipping productivity, defined as productivity achieved during the time of chipping, varied between 87 and 106 m3 of loose chips per hour and was the highest in the last cycle (Table 6). Differences were leveled out by inclusion of supportive (positioning and sorting) and non-productive time (delays), which were an integral part of worktime.

Calculating productivity on the basis of fresh weight did not change the general picture of productivity. Pure chipping productivity ranged from 31 to 41 t h-1.

Fuel consumption varied between 0.4 and 0.62 l m-3 loose volume, or between 0.8 and 1.9 liter per ton of chips.

Table 6 Productivity, fuel use and cost Chipper Pure chipping productivity – chipping only

m3 h-1 chipping 3

96.3

Net chipping productivity – excl. delays

-1

m PMH

65.1

Gross chipping productivity – incl. delays

m3 SMH-1

36.9

-3

Fuel use

0.51

lm

-3

Chipping unit cost

EUR m

3.5

Shuttle 1 m3 PMH-1

Net productivity – excl. delays

3

-1

57.4

Gross productivity – incl. delays

m SMH

18.0

Unit cost

EUR m-3

3.0

m3 PMH-1

59.5

Shuttle 2 Net productivity – excl. delays

3

-1

Gross productivity – incl. delays

m SMH

16.7

Unit cost

EUR m-3

3.5

m3 PMH-1

274.5

Loader Net productivity – excl. delays

3

-1

Gross productivity – incl. delays

m SMH

218.3

Unit cost

EUR m-3

0.2

m3 PMH-1

30.8

Whole system Net productivity – excl. delays

Fig. 2 Breakdown of worksite time for chipper

228

3

-1

Gross productivity – incl. delays

m SMH

29.4

Unit cost

EUR m-3

10.7

Notes: m3 – cubic meters of loose chips Net productivity is calculated on the basis of net time, excluding preparation and delays Gross productivity is calculated on the basis of total time, including preparation and delays

Croat. j. for. eng. 39(2018)2

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

Fig. 3 Ternary diagram of particle size distribution and moisture content of samples The estimated chipping cost was 3.5 € m-3 for loose chips, or 8.8 € t-1 for fresh chips. The cost of chip shuttling was 3.0 € m-3 or 7.8 € t-1 for shuttle 1, and 3.5 € m-3 or 9.5 € t-1 for shuttle 2. Therefore, shuttling with the push off trailer was 1.7 € t-1 more expensive than with the tipper trailer. The cost of re-loading on road convoys amounted to 0.6 € t-1. These figures include transfer to the site, which was accounted for in the overheads. The estimated net productivity of the whole system (one truck-mounted chipper and two shuttles) was 11.6 t PMH-1. Corresponding gross productivity was 11.1 t SMH-1. The estimated cost for the whole operation amounted to 26.7 €t-1.

Wood chip moisture content was different between cycles and ranged from 36.2 to 43.6 percent.

The ternary diagram in Fig. 3 shows that the largest proportion of the chip mass was represented by accepts (particles of size between 100 and 3.15 mm), which averaged 76.2% of the total, with a minimum of 72.2% and a maximum of 80.4%. The average incidence of oversize particles (above 100 mm) was 4.1% (maximum 5.4% and minimum 3.18%). Fines (particles smaller than 3.15 mm) accounted for 19.8% of the total mass as an average, with a minimum at 14.2% and a maximum at 24.5%.

5. Discussion Other studies have already explored the subject of chipping operation in space-constrained mountain

M. Mihelič et al.

areas (Spinelli et al. 2007), but this study deals specifically with a very difficult case, where very limited space is available for maneuvering. Therefore, it further expands the knowledge about chip supply in extreme work conditions, and offers a glimpse of an operation caught in an awkward but not uncommon situation. Furthermore, the study enlarges the focus of the investigation to the shuttles, which previous studies kept in background (Spinelli et al. 2015). Unfortunately, inference is based on a very limited number of observations, due to the second typical limitation of mountain operations, after poor road standards, and namely: small lot size. Nevertheless, the small number of observations obtained from the study is still large enough for the proficient application of statistical analysis, which has allowed checking the quality of our data with scientific methods.

Transportation is a main component of chip supply cost, and requires careful planning, especially in mountainous areas (Kühmaier and Stampfer 2012).

Furthermore, interaction between the chipper and the transport fleet generally results in low chipper utilization (Spinelli et al. 2014a). The alpine work conditions push the economics one step further, as conventional road trucks are often unable to reach the worksite.

That can occur because of a number of reasons, all related to the poor quality of the road infrastructure (e.g. narrow width, steep gradient, small turning radius). When road trucks are unable to reach the worksite, operators are forced to consider alternative transport options. One of them is chipping on site and shuttling the chips with tractor and trailer units to a location suitable for trucks. This additional step will naturally incur additional costs, potentially causing a negative economical outcome for the contractor. However, transportation distance is not the only limiting factor. Chip quality also plays an important role, as it has a major effect on the price of the end product. The study operation produced green residue chips, with high moisture content and a high proportion of leaves and needles, which qualified the chips into the lowest grade and negatively impacted the price. Taken together, transportation and quality constraints contributed to push the economics of this operation quite close to the limits of financial viability.

On the other hand, the study has tested a production chain in which only 3 machines (chipper, 2 tractors and trailers) were used, in order to minimize operational costs (Rawlings et al. 2004). In that regard, the use of 2 shuttles was decisive, because it allowed reducing shuttle delays.

Such conditions are especially detrimental for chipper efficiency, as shown by a very high delay factor,

Croat. j. for. eng. 39(2018)2 229

M. Mihelič et al.

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

which exceeded the already high 50% benchmark reported for truck-mounted chippers (Spinelli and Visser 2009). For the same reason, the incidence of pure chipping time was much lower than in previous studies (Spinelli and Hartsough 2001). Fortunately, feeding was relatively fast, due to the good cooperation between the yarding team and the chipping team. The chipper operator was confident that there were no stones or metal in the pile of residue. Therefore, the chipper operator never checked the grapple for contaminants and always collected full grapple loads. The operator took care to keep the road clean and free of residue, and to make sure that the position of the truck was as near to the edge of the road as possible. This way the length of the crane was fully exploited and the pile of feedstock easily reached.

The machine takes a lot of time to re-position, so it is not ideal for poorly organized sites, which require several positioning maneuvers (Spinelli and Hartsough 2001). Each time the chipper is re-positioned, the operator has to lower the crane and cabin, exit, retract the stabilizers, climb into the truck cabin, move the truck and repeat the whole process in reverse. Fortunately, frequent re-positioning was not a problem in this trial, as all feedstock was easily available in one large pile.

The utilization of the shuttles was high, compared with the figures reported in other studies (Spinelli et al. 2014a). The reason for that was that the distance from the landing to the chipping site was relatively short, and the number of shuttles involved in the system was well-balanced. It is still debatable whether adding a third shuttle would have increased chipper utilization. The forest road had very few enlargements, which prevented the easy exchange between incoming and outgoing shuttles. Therefore, the incoming shuttle had to wait at the nearest enlargement and let the outgoing trailer drive past, before it could drive to the turning point (up the road and past the chipper), turn and return back to the chipper for loading. Under such conditions, a third trailer would only add to the overall waiting time of the system, as it would hinder traffic and add to the interaction delays.

The moisture content of chips obtained from logging residue is usually very high. The reason for that is that branches and tops normally have higher moisture content than stem wood. Residue chips also have a high content of bark and needles (Spinelli et al.

2011a), which detracts from quality. Average moisture contents of fresh coniferous forest residue reported so far are 38% (Rawlings et al. 2004), 35.4 ±3.9% (Spinelli et al. 2014a), and 34.9% (Spinelli et al. 2011a). The chips produced in our study have the highest average moisture content with 40.2 ±3.1%. Some reduction of mois-

230

ture content may have been achieved by leaving the residue sit for a longer time at the landing, so that natural drying may occur. However, that is not always possible, because residues left at the edge of the forest may represent a source of infestation by insect, and forest managers often prescribe rapid removal.

The particle size distribution is also unfavorable.

The content of oversized particles gets as high as 5.4%, while fines rise to a maximum of 24%. These figures are worse than reported in other studies for the same feedstock type, especially for what concerns the presence of fine particles (Spinelli et al. 2011b). A high incidence of fine particles is typical of chips produced from softwood branches, and derives from the abundant presence of needles (Spinelli et al. 2011b). However, the amounts of fines in the study samples are close to the maximum values recorded in all previous studies, which may hint at some problems with machine settings – possibly excessive knife wear (Spinelli et al. 2014b) and/or an excessively small screen size (Nati et al. 2010). Adjusting machine settings and replacing knives more frequently is likely to lead to an improvement in chip quality, although the unfavorable characteristics of the original feedstock will still have their effect. Replacing the blower with a belt conveyor may lead to a further reduction in the incidence of fines, but it would increase machine size and weight and it would be very impractical for a machine used in space-constrained work environments (Spinelli and Magagnotti 2012).

Fortunately, chipping and extraction cost is within the 20–27 € t-1 range found in previous studies (Spinelli et al. 2014a), despite the extreme access constraints. In that regard, the short shuttling distance certainly helped: it is doubtful whether the same results could have been achieved on substantially longer distances, exceeding one or two kilometers.

6. Conclusions Comminution of forest residues is not just an economic activity, but also an important forest tending measure. Unfortunately, the effects of climate change include increased natural disturbances and drought (Dale et al. 2001), which result in a higher risk for forest fires and insect infestations. That calls for stricter fire mitigation measures, including a reduction of forest fuel build ups (Hartsough et al. 2008). Slash piles must be removed whenever possible because they represent a substantial fuel accumulation, and an ideal breeding place for noxious insects (Kacprzyk and Bednarz 2015).

Croat. j. for. eng. 39(2018)2

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

Chipping seems an ideal solution, but the question arises whether there is another, more cost-effective option for biomass chipping than that presented in this study. Recent simulation studies show that extracting chipped biomass is often more efficient than extracting uncomminuted residues and chipping them at the main landing (Spinelli et al. 2014a). While it lacks the comparative element of previous simulation experiments, the present study shows that chipping before extraction is still a viable option even when road network conditions are extremely poor, and may reassert the benefits of chipping as close to the source as possible (Björheden 2008).

Acknowledgements The research was supported by project CRP L42244, Protective forests. Special thanks to prof. Boštjan Košir PhD. for his guidance and help. Authors also wish to thank the Pahernik foundation for supporting the publishing of results.

7. References Ackerman, P., Belbo, H., Eliasson, L., De Jong, A., Lazdins, A., Lyons, J., 2014: The COST model for calculation of forest operations costs. International Journal of Forest Engineering 25(1): 75–81.

Aitchison, J., 1986: The Statistical Analysis of Compositional Data. London – New York: Chapman and Hall.

Björheden, R., 2008: Optimal point of comminution in the biomass supply chain. In: Suadicani, K., Talbot, B., editors.

Proceedings of the Nordic-Baltic Conference on Forest Operations. Universtiy of Copenhagen, Faculty of Life Sciences, Copenhagen, 9 p.

Bjorheden, R., Thompson, M.A., 2000: An international nomenclature for forest work study. In: Field, D.B., editor. Proceedings of IUFRO 1995 S3.04 Subject Area: 20th World Congress. University of Maine, Tampere, Finland, 190–215.

Comas-Cufí, M., Thio-Henestrosa, S., 2011: CoDaPack 2.0: a stand-alone, multi-platform compositional software. In: Egozcue J.J., Tolosana-Delgado R., Ortego M.I., editors. CoDaWork’11: 4th International Workshop on Compositional Data Analysis. Universitat de Girona, Sant Feliu de Guíxols, 1–10.

Dale, V.H., Joyce L.A., Mcnulty, S., Neilson, R.P., Ayres, M.P., Flannigan, M.D., Hanson, P.J., Irland, L.C., Lugo, A.E., Peterson, C.J., Simberloff, D., Swanson, F.J., Stocks, B.J., Wotton, B.M,. 2001: Climate Change and Forest Disturbances Climate change can affect forests by altering the frequency, intensity, duration, and timing of fire, drought, introduced species, insect and pathogen outbreaks, hurricanes, windstorms, ice storms, or landslides. BioScience 51(9): 723–734.

Eliasson, L., von Hofsten, H., Johannesson, T., Spinelli, R., Thierfelder, T., 2015: Effects of sieve size on chipper produc-

M. Mihelič et al.

tivity, fuel consumption and chip size distribution for open drum chippers. Croatian Journal of Forest Engineering 36(1): 11–17.

EN14774-2, 2010: Solid biofuels – Methods for the determination of moisture content – Oven dry method – Part 2: Total moisture – Simplified method.

EN15149-1, 2011: Solid biofuels – Determination of particle size distribution – Part 1: Oscillating screen method using sieve apertures of 1 mm and above.

Facello, A., Cavallo, E., Magagnotti, N., Paletto, G., Spinelli, R., 2013: The effect of chipper cut length on wood fuel processing performance. Fuel Processing Technology 116: 228– 233.

Hartsough, B.R., Abrams, S., Barbour, R.J., Drews, E.S., McIver, J.D., Moghaddas, J.J., Schwilkf, D.W., Stephense, S.L., 2008: The economics of alternative fuel reduction treatments in western United States dry forests: Financial and policy implications from the National Fire and Fire Surrogate Study.

Forest Policy and Economics 10(6): 344–354.

Hernández-Díaz, C., Soto-Cervantes, J., Corral-Rivas, J., Montiel-Antuna, E., Alvarado, R., Goche-Télles, R., 2015: Impacts of Forest Roads on Soil and Timber Harvesting Area in Northwestern Mexico (a Case Study). Croatian Journal of Forest Engineering 36(2): 259–267.

Kacprzyk, M., Bednarz, B., 2015: The occurrence of bark beetles on cut Norway spruce branches left in managed stands relative to the foliage and bark area of the branch.

Journal of Forest Research 20(1): 143–150.

Karpachev, S.P., Zaprudnov, V.I., Bykovskiy, M.A., Scherbakov, E.N., 2017: Quantitative Estimation of Logging Residues

by Line-Intersect Method. Croatian Journal of Forest Engineering 38(1): 33–45.

Kühmaier, M., Stampfer, K., 2012: Development of a Multicriteria Decision Support Tool for Energy Wood Supply Management. Croational Journal of Forest Engineering 33(2): 181–198.

Lindstad, B.H., Pistorius, T., Ferranti, F., Dominguez, G., Gorriz-Mifsud, E., Kurttila, M., Leban, V., Navarro, P., Peters, D.M., Pezdevsek Malovrh, S., Prokofieva, I., Schuck, A., Solberg, B., Viiri, H., Zadnik Stirn, L., Krc, J., 2015: Forest-based bioenergy policies in five European countries: An explorative study of interactions with national and EU policies. Biomass and Bioenergy 80: 102–113.

Magagnotti, N., Spinelli, R., 2012: Good practice guidelines for biomass production studies; WG2 Operations research and measurement methodologies. COST Action FP-0902 and CNR Ivalsa, Sesto Fiorentino, Italy, 52 p.

Marchi, E., Magagnotti, N., Beretti, L., Neri, F., Spinelli, R., 2011: Comparing Terrain and Roadside Chipping in Mediterranean Pine Salvage Cuts. Croatian Journal of Forest Engineering 32(2): 587–598.

Mihelič, M., Spinelli, R., Magagnotti, N., Poje, A,. 2015: Performance of a new industrial chipper for rural contractors.

Biomass and Bioenergy 83: 152–158.

Croat. j. for. eng. 39(2018)2 231

M. Mihelič et al.

Production of Wood Chips from Logging Residue under Space-Constrained Conditions (223–232)

Nati, C., Spinelli, R., Fabbri, P., 2010: Wood chips size distribution in relation to blade wear and screen use. Biomass and Bioenergy 34(5): 583–587.

Spinelli, R., Magagnotti, N., 2012: The Effect of Raw Material, Cut Length, and Chip Discharge on the Performance of an Industrial Chipper. Forest Products Journal 62(7): 584–589.

Petković, V., Potočnik, I., 2018: Planning Forest Road Network in Natural Forest Areas: a Case Study in Northern Bosnia and Herzegovina. Croatian Journal of Forest Engineering 39(1): 45–56.

Spinelli, R., Magagnotti, N., Paletto, G., Preti, C., 2011a: Determining the Impact of Some Wood Characteristics on the Performance of a Mobile Chipper. Silva Fennica 45(1): 85–95.

Pičman, D., Rubil, I., Pentek, T., Pičman, O., 2007: Soil Stabilisation by Powercem and the Possibility of Implementation in Forest Road Building. Šumarski list 131(9–10): 453–463.

Poje, A., Spinelli, R., Magagnotti, N., Mihelič, M., 2018: The effect of feedstock, knife wear and work station on the exposure to noise and vibrations in wood chipping operations.

Silva Fennica 52(1): 1–14.

Rawlings, C., Rummer, B., Seeley, C., Thomas, C., Morrison, D., Han, H., Cheff, L., Atkins, D., Graham, D., Windell, K., 2004: A study of how to decrease the costs of collecting, processing and transporting slash. Montana Community Development Corporation, Missoula (MT), 21 p.

Spinelli, R., De Francesco, F., Eliasson, L., Jessup, E., Magagnotti, N., 2015: An agile chipper truck for space-constrained operations. Biomass and Bioenergy 81: 137–143.

Spinelli, R., Di Gironimo, G., Esposito, G., Magagnotti, N., 2014a: Alternative supply chains for logging residues under access constraints. Scandinavian Journal of Forest Research 29(3): 266–274.

Spinelli, R., Glushkov, S., Markov, I., 2014b: Managing chipper knife wear to increase chip quality and reduce chipping cost. Biomass and Bioenergy 62: 117–122.

Spinelli, R., Hartsough, B., 2001: A survey of Italian chipping operations. Biomass and Bioenergy 21(6): 433–444.

Spinelli, R., Nati, C., Magagnotti, N., 2007: Recovering logging residue: experiences from the Italian Eastern Alps.

Croatian Journal of Forest Engineering 28(1): 1–9.

Spinelli, R., Nati, C., Sozzi, L., Magagnotti, N., Picchi, G., 2011b: Physical characterization of commercial woodchips on the Italian energy market. Fuel 90(6): 2198–2202.

Spinelli, R., Visser, R., 2008: Analyzing and Estimating Delays in Harvester Operations. International journal of forest engineering 19(1): 36–41.

Spinelli, R., Visser, R., 2008: Analyzing and estimating delays in wood chipping operations. Biomass and Bioenergy 33(3): 429–433.

Stupak, I., Asikainen, A., Jonsell, M., Karltun, E., Lunnan, A., Mizaraite, D., Pasanen, K., Pärn, H., Raulund-Rasmussen, K., Röser, D., Schroeder, M., Varnagirytė, I., Vilkriste, L., Callesen, I., Clarke, N., Gaitnieks, T., Ingerslev, M., Mandre, M., Ozolincius, R., Saarsalmi, A., Armolaitis, K., Helmisaari, H.S., Indriksons, A., Kairiukstis, L., Katzensteiner, K., Kukkola, M., Ots, K., Ravn, H.P., Tamminen, P., 2007: Sustainable utilisation of forest biomass for energy—Possibilities and problems: Policy, legislation, certification, and recommendations and guidelines in the Nordic, Baltic, and other European countries. Biomass and Bioenergy 31(10): 666–684.

Team RDC. R, 2008: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Authors’ addresses: Matevž Mihelič, PhD. * e-mail: matevz.mihelic@bf.uni-lj.si Assist. prof. Anton Poje, PhD. e-mail: anton.poje@bf.uni-lj.si University of Ljubljana Biotechnical Faculty Department of Forestry and Renewable Forest Resources Večna pot 83 Ljubljana SLOVENIJA

Received: September 01, 2016 Accepted: January 31, 2017

232

Raffaele Spinelli, PhD. e-mail: spinelli@ivalsa.cnr.it CNR IVALSA Via Madonna del Piano 10 Sesto Fiorentino ITALY * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Dynamic Soil Pressures Caused

by Travelling Forest Machines Milan Marusiak, Jindřich Neruda Abstract Machines travelling in forest stands cause dynamic loading of soil, the size of which depends on a multitude of factors such as terrain ruggedness, machine speed, axle load and tyre inflation pressure. To decide on harvesting and transport machines suitable for specific field conditions, it is necessary to have at least some awareness about their dynamic effects on the soil, which sometimes considerably differ from static values measured on standing machines. The paper deals with the method of determining dynamic ground pressures according to the given parameters of vehicle weight and speed. At the same time, it compares dynamic pressures calculated by using this method with actually measured values.

Keywords: dynamic soil pressures, tyre deformation, contact area

1. Introduction The travel of harvesters and forwarders across forest stands results in the contact occurring between the vehicle chassis and the ground surface. Machine weight and traction effects of the machine chassis induce ground pressures, which spread to sides and into depth. In wheeled machines, the ground pressures are influenced primarily by tyre characteristics such as diameter, width, rate (stiffness) and inflation pressure, other important parameters being adhesion load on the tyre and components of traction forces acting on the wheel. The vehicle contact pressure is the ratio between the weight and contact surface of the vehicle with the ground (soil), and it expresses the environmental suitability of a specific forest vehicle (Poršinsky et al. 2011).

Ground pressure is also considerably affected by characteristics of the soil surface across which the machine travels – elasticity and plasticity in particular.

The effects of elasticity and plasticity especially show when the machine moves, the elastic soil returns to the original condition after temporary compression, and the wheel is also propped in the space behind the axis.

The machine travelling across the plastic soil causes permanent deformation of the latter and the wheel is only supported by a part of the contact area. The value of instantaneous wheel load is also affected by the dy-

namics of machine travel including the dynamic effects of travelling across surface irregularities, where individual wheels experience higher load for a short time, and the wheel engagement may for a short time increase its power load (Neruda et al. 2013). The dependence of the tyre and soil contact area on elastic deformations of the loaded wheel (tyre characteristics, air pressure) and plastic-elastic soil deformations (granulometric content, moisture) is considered as problem when calculating the vehicle contact pressures for forest off-road travel (Poršinsky et al. 2011).

The soil profile, which generates the vibration of the vehicle, is significantly modified by the vehicle itself. The soil is compressed by the vehicle and part of the kinetic energy of the vibrating vehicle is absorbed by the soil, while the dynamic normal and shear forces affect the interaction. The dynamic load created by the moving tractor modifies soil cohesion significantly, while internal friction is not affected. The amount of change in cohesion is a function of travel velocity, the mass of the vehicle and moisture content (Laib 1999).

The total mass of the vehicle, the wheel load and dynamic shear forces are important factors affecting the degree of soil deformation. During soil compaction, strength and bulk density increase, porosity is reduced, the pore size distribution changes, and the infiltration rate decreases (Greacen and Sands 1980).

Croat. j. for. eng. 39(2018)2 233

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

p pd pi r rs S v U Ur z

The extent of soil degradation by farm tractor forwarding operations along the upper 20 cm of the soil profile increase with traffic frequency, but these increases vary with slope gradient and direction of forwarding (Jourgholami et al. 2014).

Actual ground pressures, particularly in the upper soil layers, up to a depth of 15 cm, where a majority of spruce root systems occur, presumably differ from static pressures established by calculation. Therefore, it is necessary to know the actual values of ground pressures in order to be able to compare the impact of travelling forest machines on the condition of the root system. Saarilahti and Antilla (1999) consider the soil layer at a depth of 15 cm as the most representative in terms of the agreement of measured and calculated data of penetration soil resistance.

The paper presents a method of calculating dynamic ground pressures according to the given parameters of vehicle weight and speed, which is based on an assumption that the dynamic wheel load is build on the law of energy conservation and transformation on impact.

The calculated dynamic pressures are compared with the results from measuring the direct ground pressures of eight-wheeled forwarder L511 with max. payload of 5 t.

1.1 Used symbols b tyre width, m bc tyre contact width, m bw tyre tread pattern width, m b dynamic coefficient d tyre diameter, m δ tyre deformation, m δst static tyre deformation, m δr tyre deformation at static load, m Fd dynamic load on the wheel, N Fk static load on the wheel, N Fn nominal load on the wheel, N Gv vehicle weight load, kN g gravitational acceleration, m×s-2 h tyre profile height, m k tyre rate, stiffness, N×m-1 k2 constant taking into account the number of driving axles 2.05–4 axles, 1.95–3 axles and 1.83–2 axles lc tyre contact length, m K kinetic energy, J m vehicle weight proportion per tyre, kg n number of axles,

234

contact ground pressure, kPa contact ground pressure at dynamic load, kPa tyre inflation pressure, kPa unloaded tyre radius, m tyre static radius, m tyre contact surface area (print), m2 vehicle speed, m×s-1 potential energy, J potential energy of stress, J wheel sinking (rut depth), m

2. Theoretical Analysis 2.1 Contact ground pressure There are many authors dealing with the determination of contact ground pressure. The simplest calculation of contact ground pressure is given by its definition as a ratio of the wheel load and the tyre contact area – see for example Pacas et al. (1990):

p=

Fk S

(1)

In practice, a simple estimate of vehicle crossing ability in certain soil conditions is often made by using the NGP (Nominal Ground Pressure) method (Partington and Ryans 2010), according to which contact ground pressure is expressed by the following relation:

p=

Fk

b×r

(2)

According to this method, the length of the wheel contact with the ground equals the wheel radius, which presumes a wheel sinking in the ground at a depth of about 15% of its diameter. As compared with the measured contact pressure, the calculated contact pressure is too low and the sinking of a larger wheel represents a depth, which is not suggested for ecological reasons. Another disadvantage of this method is the negligence of important parameters of wheel contact with the ground such as tyre deformation and inflation pressure (Saarilahti 2002).

To compare the crossing capability of military tracked and wheeled vehicles on soils with low carrying capacity, a method was developed in the early 1970s for the determination of mean maximum contact ground pressure MMP (Rowland 1972). MMP (Mean Maximum Pressure) represents the mean value of maximum contact pressures by which the vehicle affects the soil when travelling. To provide vehicle moCroat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

bility, the soil carrying capacity must be greater than the MMP value for the given vehicle. This method was later worked out by more authors for various chassis configurations (wheeled, tracked, tracked on tandem axles) and soil conditions (cohesive soils and frictional soils).

The original formula for the calculation of contact pressure by using the MMP method (Rowland 1972) for off-road tyres is as follows:

p=

1.18 × Gv 2× n× b× d× h

(3)

Larminie (1992) elaborated modified formulas for calculating MMP, into which he also included tyre deformation as an important parameter. Contact pressure for wheeled vehicle on fine grained cohesive soils is calculated according to the following relation:

k 2 × Gv

p= 2 × nb

0.85

×d

1.15

×

δ

(4)

d

Building on the methodology of measuring soil carrying capacity by means of cone penetrometer, in which soil resistance is measured against cone penetration at a depth of 15 cm, Maclaurin (1997) suggested replacing the MMP value by the expression of Limiting Cone Index (CIL). The measuring method is governed by the norm ASAE EP542 1999 and the gauged value of soil resistance is called Cone Index.

Limiting cone index expresses the lowest load with respect to soil carrying capacity, at which the vehicle with a certain MMP is still mobile. According to this methodology, the contact pressure is calculated as follows:

p=

1.18 × Gv 2× n× b

0.8

×d

0.8

×d

0.4

(5)

The above methods for determining contact pressure use static load of the vehicle as an input value.

However, in actual conditions of machines moving within the forest stand, there are dynamic effects on the soil caused by machines passing over terrain irregularities that, in some cases, may be several times higher than their static values. A simple calculation of the static ground contact pressure of forest harvesting machines is not a good indicator of the dynamic pressure exerted on soil during skidding (Lysne and Burditt 1983). The ground contact pressure is not uniformly distributed over the contact area, and its distribution beneath the wheel is complex due to a number of variables, such as tyre lug pattern, tyre load distribution,

M. Marusiak and J. Neruda

and tyre carcass stiffness (Peng et al. 1994). The maximum ground contact pressure under lugs or stiff tyre sidewalls may be several (even ten) times higher than the estimated average ground contact pressure (Burt et al. 1992).

To be able to assess the impact of forest machines on the soil surface, it is therefore important to identify not only static but also dynamic forces acting in the mutual interaction of machine chassis and soil surface (Neruda et al. 2013). The mode by which the dynamic forces manifest themselves in specific conditions of given soil profile can be documented for example by measuring dynamic ground pressures.

Table 1 brings calculation formulas for establishing contact pressure, presented by several authors. None of the formulas mentioned in the literature is universally applicable for estimating the suitability of tyre use in soils with low carrying capacity. Saarilahti (2002) recommends using formulas with tyre deformation entered as an input variable because their results lead to a better choice of the tyre from the environmental point of view. Eq. (2) through to (6) use static wheel load as an input value. Eq. (1) expresses the value of contact pressure at dynamic load calculated according to the procedure given in Material and Methods, Fig. 3.

Table 1 Contact pressure – calculation formulas Calculation model

Source

Equation

Number

Pacas (1990)

pd = Fd/S

(1)

NGP

p = Fk/r×b

(2)

MMP Rowland (1972) p = 1.18×Fk/(2×b×(d×h)0.5) 0.85

MMP Larminie (1988)

1.15

p = kL×Fk/2×b ×d ×(d/d) kL= 2.05

MMP Maclaurin (1997) p = 1.85×Fk/2×b0.8×d0.8×d0.4 Dwyer (1984)

0.5

(3)

0.5

p = (Fk/b×d)×(h/d) ×(1+b/2×d)

(4) (5) (6)

2.2 Static Tyre Deformation One of important characteristics of the tyre is its rate (stiffness), which is expressed by the load/deformation ratio. Tyre rate depends on tyre design (diagonal or radial and number of layers), size and inflation pressure. If tyre rate data are not available from the manufacturer, the value can be determined based on the known load and deformation calculated in depen-

Croat. j. for. eng. 39(2018)2 235

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

dence on tyre size and type, using empirical formulas published by several authors.

Table 2 brings empirical formulas for the calculation of tyre deformation presented by several authors, obtained from the processing of tyre deformations measured during the tests of forest and agricultural machines. The data express the dependence between tyre deformation, inflation pressure and load (possibly also size) and apply only to the dimensions of tyres used in the tests (within a certain range).

Kinetic energy (K) is expressed by the following formula (Zahradníček and Semrád 2007):

Table 2 Tyre deformation – calculation formulas Source

Calculation formula

Number

Wulfsohn et. al (1988)

d = 0.02+0.006×Fd–1.35×Fd×pi×10–5

(7)

Nokia, model 1

d = 0.121×Fd0.476/pi0.570

(8)

Nokia, model 2

d = 0.008+0.001×(0.365+170/pi)×Fd

(9)

Godbole (1993)

d = h×0.67×(pi×d×b/Fd)

(10)

Schmidt (1988)

d = 0.01+(0.0007+0.302/pi)×Fd

(11)

-0.8

tem under the influence of shock in order to simplify the calculation. In a system stressed in this way, the load size would change within a short time interval and its maximum value can be several times higher than the static load of the standing machine. The size of dynamic wheel load can be established based on the law of energy conservation and transformation on impact.

Kinetic energy of the moving wheel, represented by its load and velocity, would change into potential energy of the flexible system stress, represented by the rate and deformation of the tyre (Fig. 1)

2.3 Determination of Dynamic Load Passing through a rugged terrain, the forest machine chassis is exposed to the load the size of which constantly changes depending on the size of obstacles and terrain irregularities the machine has to cross. Since the terrain in the forest stand is not homogeneous, the forest machine tyre behaviour can be considered as a sys-

m × v2 , J 2

K=

(12)

Initial potential energy (U) is expressed by the following formula:

 v2 U =× Fk  ×δ  2× g r 

  , J  

(13)

This is changing into potential stress energy caused by tyre deformation upon the wheel contact with the base:

Ur =

k × δ r2 2

, J

(14)

Tyre rate can also be expressed as a ratio of load and deformation at static load as follows:

k=

Fk

δ st

, Nm-1

(15)

Modifying the Eg. (12) through to (14) and taking into account a angle (Fig. 3), a formula for the calculation of dynamic deformation of the tyre is obtained:

δ r = δ st × (1 + 1 +

v 2 × sinα ), m 2 g × cosα × δ st

(16)

The expression in brackets is called a dynamic coefficient (b) and gives the ratio of the increase of dynamic deformation δr compared to static deformation δst. It can be also expressed as:

Fig. 1 Energies and forces on the rolling tyre

236

b =1 + 1 +

K

U

(17)

Pursuant to Hooke’s Law, the force and stress are directly proportional to strain (deformation), and the below formula holds for the size of dynamic load Fd: Croat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

Fd= β × Fk , N

M. Marusiak and J. Neruda

(18)

The following conclusions can be made based on the above relations (16), (17) and (18): Þ Increased static load and increased travel speed result in increased dynamic load Þ Increased elasticity of the tyre (e.g. due to decreased inflation pressure), which is expressed by increased static deformation and increased angle a, results in decreased dynamic coefficient b and hence in decreased dynamic load.

2.4 Tyre contact area The forest machine chassis makes contact with the soil surface in the stand and the size of the contact area depends on several factors such as tyre dimensions, tyre deformation, inflation pressure, wheel load, and soil characteristics (soil texture, carrying capacity, moisture content) (Hallonborg 1996, Neruda et al.

2013). On the hard rigid basement (concrete, asphalt, paved road), the tyre contact area is smaller, its shape ranging from circular to rectangular in dependence on the tyre type and inflation. If the tyre sinks into the soil of low carrying capacity, tyre tread sides carry some load too, and the contact area is represented by the general area (Pacas et al. 1983). In practical calculations, the vertical projection of this general area is most frequently used as a contact area, or the tyre print area on the solid base.

The size of the tyre contact area is also affected by the machine dynamics, i.e. by the machine movement and by the engagement of its wheels. Low inflation pressure, high tyre load, and soft soils contribute to large contact areas. In forests, vehicles move on a plastic matrix composed of soil, thus producing an asymmetric contact area that is perpendicular to the tyre. If vehicles move laterally on a slope, the contact area of the wheels is asymmetrical with respect to the longitudinal axis. The size of the contact area changes continuously due accelerating/braking, changing payload, and uneven soil surface (Alakukku 1999).

The determination of the contact area has been dealt with by several authors, who developed calculation formulas based on observations and measured values.

Reviewing the shape of the contact area, Grecenko (1995) recommended multiplying the length and the width of the contact area by a coefficient, c, varying between 0.8 and 0.9.

Hallonborg (1996) proposed a super ellipse model to describe the geometry of tyre-terrain contact with half-axes, a and b, as well as a positive variable expo-

Fig. 2 Elastic tyre on soft ground nent to determine the shape of the ellipse. Nevertheless, they require input data that are not easily acquired, and do not consider the rapid dynamic variation during machine trips.

Analytical calculation formulas are based on the definition of loaded wheel geometry in interaction with the soil environment. The best simulation of the travel of forest machines in the terrain is the model, which describes elastic tyres on soft ground (Fig. 2). In this case, the deformation occurs of both the tyre and the ground.

The calculation formula for this case was elaborated by Schwanghart (1990) based on wheel geometry (Fig. 1). The tyre print area is calculated according to the following formula:

A =c × bc × lc , m2 (19)

Where: c constant, expressing the shape of tyre print. The tyre print shape depends on tyre design, inflation pressure, wheel load and soil characteristics. Its outer contour ranges between the circle and the rectangle. The value recommended for the tyres of forest machines is c = π/4 = 0.785 bc contact width, m – is the maximum width of tyre print lc contact length, m – is the total length of tyre print.

It consists of two contact lengths l1 and l2 (Fig. 2): lc = l1 + l2 = d × ( z + δ ) − ( z − δ ) + d × δ − δ 2 , m (20) 2

One of the input parameters in the calculation of contact length is the depth of tyre sinking in the

Croat. j. for. eng. 39(2018)2 237

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

Table 3 Tyre contact area – calculation formulas Source

Calculation formula

Number

S = 0.785×bc×lc

Schwanghart (1990)

bc = b+c. Fd/Fn [c=0.03...0.05]

(21)

lc = (d×(z+d)–(z+d)2)0.5+(d×d–d2)0.5

Grecenko (1995)

S = 1.57×(d–2×rs)×(d×b)0.5

(22)

S = p/4×lc×bc lc = c3×(d× d–d2)0.5

Lyasko (1994)

c3 = 23/(ABS(d/b–3.5)+11.9)

(23)

bc = 2×((b+h/25)×d–d2)0.5 S = p/4×bc×lc lc = 2×d0.5×d j [j = 0.44]

Febo (1987)

Fig. 3 Procedure for determining dynamic pressures (24)

bc = bw(1–exp–k.d) [k = 36] Komandi(1990) Dwyer (1984)

S = (c2×Fd0.7×(b/d)0.5)/pi0.45 S = Fd /G G = Fd/b×d×(h/d)0.5×(1+b/2×d)

(25) (26)

ground (z). Several ways of its determination, based on the WES method, are mentioned in literature (Maclaurin 1990). For simplification, values used in the following calculation are rut depths measured after the first pass of the forwarder Model L511 during the tests: loaded machine – z = 0.08 m; empty machine – z = 0.07 m.

Empirical calculation formulas were developed based on the observation of the dependence of tyre print size on tyre parameters (rate (stiffness) as expressed by strain (deformation), diameter, width, inflation pressure) and soil environment characteristics (soil texture, penetration resistance, moisture content) (Saarilahti 2002).

Empirical calculation formulas selected for further processing include the following input parameters: tyre dimensions and deformation, inflation pressure, wheel load, possibly also tyre sinking depth. Calculation formulas for the determination of the tyre contact area and their comparison with the measured values are presented in Table 3.

3. Material and methods To be able to determine dynamic ground pressures, we first need to know the size of dynamic load by which the forest machine acts on the soil while travelling at a certain speed. The procedure for determining the dynamic load and dynamic ground pressures consists of the following issues:

238

Þ Determination of tyre deformation under static load. Based on the measured values of deformation, a relation for the calculation is established, which expresses the dependence of tyre deformation on tyre load (tyre rate) and inflation pressure. Þ Determination of dynamic wheel load in the machine moving at a certain speed on flat terrain by using the power conservation law. Þ Calculation of tyre/ground contact area at a given dynamic load. Comparison of calculation formulas developed by several authors and selection of a formula, the results of which are closest to the contact area values, measured during forwarder L511 trials. Þ Calculation of dynamic pressures at a given contact area and dynamic load; Comparison of measured dynamic axle pressures with values calculated according to dynamic load and according to calculation formulas presented by several authors. Determination of static/dynamic pressure ratios.

The correlation between the individual parameters in the procedure of determining the dynamic ground pressures is illustrated in Fig. 3 presented below.

3.1 Measurements of Dynamic Contact Pressures Dynamic contact pressures were measured during tests conducted on the forwarder Model L511 in the locality of Maršov (Czech Republic). Researchers at the Faculty of Forestry and Wood Technology, Mendel University in Brno, developed a method for measuring and registering direct ground pressures by using a pressure probe installed at a shallow depth below the soil surface at the site of machine pass (Zemánek 2015).

Croat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

M. Marusiak and J. Neruda

the pressure probes with a defined burden, by subsequent reading of pressure values and by their comparison with the calculated results. The maximum measurement error was 5%. During the measurements, the pressure probe was installed in the soil at a depth of 20 cm. This depth was chosen because the zone contains a considerable amount of roots of shallow rooting trees that are exposed to damage during the passage of forest machines (Zemánek et al. 2015).

The placement of the probe was followed by the measurement of 10 machine travels in the same rut (5 forward and 5 back) in the loaded and empty version.

Fig. 4 shows the course of contact pressure values recorded by the pressure probe during the second travel of the forwarder.

For comparing the calculated and measured values of dynamic pressure, data from the second through to fifth machine pass were used. The reason is that during the first pass, the measured pressure values are usually lower due to the clearance between the hole drilled in the soil and the diameter of the pressure probe (Zemánek 2015).

Fig. 4 The course of contact pressure during the forwarder pass Dynamic pressures in the ground were measured by using the measuring equipment consisting of pressure sensor, which included a strain gauge connected to a pressure probe, a converter of analogue-digital signal, and a notebook with the measuring programme.

The system of pressure probe and strain gauge was filled with liquid and de-aerated. Thus, a homogeneous hydraulic connection was created, by which the ground pressure on the walls of the pressure probe was transferred to the sensor. The accuracy of the measuring equipment was regularly checked by loading

4. Results and Discussion 4.1 Tyre Deformation In 2015, a series of measurements and tests of parameters of the 8x8 forwarder Model Novotny L511 (max. payload 5 t) were conducted at Mendel University in Brno. The tests also included the measuring of wheel static radius under different values of inflation pressure in tyres and load. Static radius was measured on the rear axle of the machine as a distance from the

Table 4 Measured average values of static radius and tyre deformation Machine weight

Weight proportion per rear axle

Inflation pressure

Measured static radius

Tyre deformation

kg

kg

pi, bar

rs, m

d, m

Dispersion index

6220

2750

3.5

0.412

0.013

0.00024

(Empty machine)

(Empty machine)

2.5

0.403

0.022

0.00027

1.5

0.381

0.044

0.00058

10 920

8570

3.5

0.405

0.020

0.00020

(Loaded machine)

(Loaded machine)

2.5

0.398

0.027

0.00028

1.5

0.318

0.107

0.00041

Tyres Type

Diameter, m

Radius, m

Width, m

Profile height, m

Mitas D FOREST 400/60–15.5

0.85

0.425

0.405

0.24

Croat. j. for. eng. 39(2018)2 239

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

Fig. 5 Measured and calculated values of tyre deformation centre of the wheel hub to the base surface. The difference between the measured static radius and the radius of unloaded tyre then represents the tyre deformation.

Results of measurements and calculated values of tyre deformation of rear axle are presented in Table 4.

According to measured values of static radius the following formula for calculating the tyre deformation was developed through regression:

( −0.008× pi + 0.577 ) , m

= δ 0.001 × Fd

(27)

Fig. 5 presents the comparison of measured and calculated values of tyre deformation.

The agreement of measured and calculated data, expressed by correlation coefficient r2, is presented in Table 5.

Table 5 Tyre deformation – correlation coefficients Calculation equation number

r2

Wulfsohn et. al (1988)

(7)

0.853

Nokia, model 1

(8)

0.854

Nokia, model 2

(9)

0.764

Godbole (1993)

(10)

0.841

Schmidt (1988)

(11)

0.758

Formula developed through regression according to measured values

(27)

0.963

Source

240

Even though the r2 correlation coefficients between the measured and calculated tyre deformation values are relatively high, certain differences can still be observed in Fig. 5, especially in the loaded tyre.

The last equation (27), developed by using regression according to measured values of static wheel radius, where r2 represents a value of 0.963, suits best the deformation of the tyre Mitas D FOREST 400/60–15.5.

This formula is used in the following calculations of dynamic load and contact area of the tyre.

Tyre overloading beyond the tolerable limit can also be seen in Fig. 6, where the calculated curves of tyre deformation are illustrated in dependence on the load. Solid line represents tyre deformation curves at static load, while dashed line represents tyre deformation caused by the dynamic equivalent of static load during the machine travel. Horizontal lines illustrate the degree of tyre strain (deformation) at permitted tyre load and inflation pressure. The rear axle tyre of the loaded machine (Fk=15 900 N) is at the level of the limit deformation already under the static load. During the machine drive with underinflated tyres (150 kPa), the permitted deformation is already exceeded with the empty machine. In the loaded moving machine, the profile height of the rear axle tyre is so low that there is a risk that the tyre casing might come into contact with the tyre bead causing a sidewall collapse.

4.2 Tyre Contact Area The contact surface of the forwarder tyre has been determined experimentally by measuring the external Croat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

M. Marusiak and J. Neruda

Table 6 Measured values of tyre contact area Inflation pressure, kPa Machine

Axle

Wheel load, N

350 Tyre contact area, m2

Empty

Loaded

Front axle

8510

0.145

Rear axle

6744

0.130

Front axle

5812

0.137

Rear axle

21 018

0.182

Table 7 Tyre contact area – data coincidence Source

Calculation equation number

Measured values

Fig. 6 Tyre deformation dimensions of the tyre imprinted in the soil. These external dimensions were embedded in a rectangle, and the area of this rectangle was considered to simulate

Data coincidence, % 100

Schwanghart (1990)

(21)

92

Grechenko (1995)

(22)

50

Lyasko (1994)

(23)

23

Febo (1987)

(24)

49

Komandi(1990)

(25)

86

Dwyer (1984)

(26)

65

the tyre contact area. Measure values of the tyre contact area are presented in Table 6.

Fig. 7 Tyre/ground contact area – calculated and measured values Croat. j. for. eng. 39(2018)2 241

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

Fig. 8 Measured and calculated tyre contact pressures The comparison of the agreement of the calculated and measured values, presented in Table 7 and in Fig. 7, show that the best suiting formulas for the calculation of the ground contact area of the tyre Mitas D FOREST 400/60–15.5 are those by Schwanghart (21) – coincidence of 92% and Komandi (25) – coincidence of 86%.

Eq. (21) is used in the following calculations of tyre contact pressure.

Table 8 Tyre contact pressure – data coincidence Source

Calculation equation number

Measured values

Data coincidence, % 100

Pacas (1990 )

(1)

81

NGP

(2)

15

4.3 Tyre Contact Pressure

MMP Rowland (1972)

(3)

67

Fig. 8 presents the measured and calculated tyre contact pressures. The contact pressures were measured and calculated for the front and rear axle of the empty and loaded machine. The coincidence between measured and calculated contact pressures is presented in Table 8.

MMP Larminie (1988)

(4)

48

MMP Maclaurin (1997)

(5)

71

Dwyer (1984)

(6)

80

The calculation equations give values from 45 to 160 kPa for front axle empty and from 93 to 290 kPa for rear axle loaded. The NGP (2) eq. seems to give very low values, while contact pressure calculated by the Larminie (4) and Rowland (3) eq. are slightly higher.

Eq. (5-Maclaurin) and (6-Dwyer) exhibit a relatively good agreement with the measured data also at the given static load. Eq. (1-Pacas) shows a similarly good agreement (81%) at the given dynamic load.

4.4 Dynamic load The diagram in Fig. 9 presents the measured static load and calculated dynamic load of the wheels of front and rear axles. Horizontal lines represent permitted

242

wheel load at various inflation pressures. It follows from the diagram that the dynamic load represents a multiple value of its static equivalent. Comparing the dynamic and the permitted load values, it can be concluded that, during the machine travel across the rugged terrain, the tyre experiences a short-term overloading beyond the tolerable limit. In addition, there is also the influence of reduced pressure in the tyre, which increases the tyre capacity of absorbing shock loads and hence partly reduces the level of dynamic load.

The static/dynamic load ratio, characterized by the dynamic coefficient b, is an important parameter in the machine design. It represents a load increase, to which the dimensions of all machine parts exposed to load during the machine travel (namely parts of chassis) are Croat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

M. Marusiak and J. Neruda

Fig. 9 Static and dynamic wheel load to be accommodated. It also has great significance in wear calculations and in establishing the service life of machine parts. The size of the dynamic coefficient is determined through estimate according to required

technical conditions of machine operation and based on measuring axle load in the machines of similar class.

For example, in trucks moving on the road, b = 1.2÷1.8, while in military vehicles moving across the terrain at high speed, b = 3.6÷4.0. The results of calculations presented in the diagram (Fig. 8) indicate that the dynamic coefficient of a forest machine working in the field may reach values b = 2.4÷2.6.

The diagram in Fig. 10 shows the size of contact pressures at static and dynamic loads calculated according to the procedure for establishing dynamic pressures presented in Fig. 3. Horizontal lines in the diagram represent the contact pressure at the maximum tyre load for 3 levels of inflation pressure. Vertical lines in the diagram represent the level of calculated dynamic load on the front and rear axles in the empty and loaded machine. Values of contact pressures measured under the passing machine with tyres inflated to 350 kPa are illustrated as dots at the vertical lines of the diagram.

The diagram shows that dynamic contact pressures are by 1.6–2 times higher than contact ground pressures caused by the static load. Since the difference is decreasing with the decreasing tyre inflation pressure (2 times at 350 kPa, 1.6 times at 150 kPa), softer tyres have a higher capacity for reducing the shock load.

Fig. 10 Contact ground pressure caused by the tyre

The reduced air pressure in tyres affects the size of dynamic pressures at several levels. The reduced pressure increases the tyre capacity to reduce the shock

Croat. j. for. eng. 39(2018)2 243

M. Marusiak and J. Neruda

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

load, which reduces the dynamic impact of the machine on the soil environment. The tyre with a lower pressure has a larger contact area, which reflects in a lower contact ground pressure.

In spite of the favourable effects of machine travel with the reduced tyre pressure on the soil, it is necessary to take into account the permitted tyre load in selecting the tyre pressure. This particularly applies to forwarders, where the load on the rear axle changes depending on whether the machine is loaded or empty. To prevent damage to tyres and to extend their service life, machine operators usually inflate them to the maximum prescribed pressure without considering environmental consequences.

In practice, it often happens that damage occurs to the sidewall of the tyre during the travel of machine with underinflated tyres, or a branch gets between the tyre bead and the wheel rim edge. Internal pressure in the tyre is too low to press the tyre to the wheel rim, which can even result in tyre spinning on the wheel rim at a powerful wheel engagement or in tyre slipping under the influence of lateral forces on the slope and at turning.

One of the methods to solve the problem is using a split rim with the BEAD LOCK ring in the wheel assembly (Fig. 11), which can provide additional thrust of tyre bead to the wheel rim even at a lower pressure. A similar principle with the RUN FLAT insert in wheels is used in military vehicles with the standard built-in central tyre inflation system or in

vehicles transporting VIPs. A vehicle with the mounted BEAD LOCK rings is capable of moving over a certain distance even with the deflated tyre, e.g. in a difficult terrain or on low carrying capacity soils.

A disadvantage of this solution is a high price of the complete wheel and complex assembly. The ring is manufactured at a width higher than the inner size of the assembled tyre in order to generate pre-stress between the ring and the tyre walls. At assembling the wheel, it is necessary to press down the locking ring of the wheel rim and lock it in this position by the safety ring. Since a special fixture is used for this purpose, tyre replacement in the field is not possible.

Despite the above-mentioned disadvantages, it should be verified whether it is possible to use wheels with the BEAD LOCK ring in the operation of forest machines and to investigate their loading impact on the soil at driving with underinflated tyres.

5. Conclusion This paper presents the method of determining contact ground pressures based on axle loads and machine travel speed. Calculation results and their comparison with the measured values show that this method of establishing the dynamic pressures can be an alternative to calculation models based on already applied WES and MMP methods. In the future, further tests will be needed to verify this calculation procedure also in the machines of higher load-carrying capacity class with different inflation pressures and travel speeds.

Acknowledgement This paper was created as a part of the project TAČR Alfa TA04020087 »Development and manufacture of variable forwarder with the focus on ecological cleanliness of operations and effective processing of biomass in forestry«.

6. References Alakukku, L., 1999: Subsoil compaction due to wheel traffic.

Agricultural Science in Finland 8(4–5): 333–351.

Burt, E.C., Wood, R.K., Bailey, A.C., 1992: Some comparisons of average to peak soil tyre contact pressures. Trans. ASAE 35: 401–404.

Dwyer, M.J., 1984: Tractive performance of wheeled vehicles. Journal of Terramechanics 21(1):19–34.

Fig. 11 BEAD LOCK ring

244

Greacen, E.L., Sands, R., 1980: Compaction of forest soils. A review. Australian Journal of Soil Research 18(2): 163–189.

Croat. j. for. eng. 39(2018)2

Dynamic Soil Pressures Caused by Travelling Forest Machines (233–245)

M. Marusiak and J. Neruda

Grecenko, A., 1995: Tyre footprint area on hard ground computed from catalogue value. Journal of Terramechanics 32(6): 325–333.

Partington, M., Ryans, M., 2010: Understanding the nominal ground pressure of forestry equipment. FPInnovations, 12(5): 1–8.

Haas, H., 1994: Off-Road Tyres with Emergency Capabilities. Proceedings of the 6th ISTVS Conference, Vienna, 9/1994, 696–706.

Peng, C., Cowell, P.A., Chisholm, C.J., Lines, J.A., 1994: Lateral tyre dynamic characteristics. Journal of Terramechnics 31(6): 395–414.

Hallonborg, U., 1996: Super ellipse as tyre-ground contact area. Journal of Terramechanics 33(3): 125–132.

Poršinsky, T., Stankić, I., Bosner, A., 2011: Ecoefficient Timber Forwarding Based on Nominal Ground Pressure Analysis. Croatian Journal of Forest Engineering 32(1): 345–356.

Jourgholami, M., Soltanpour, S., Etehadi, A.M., Zenner, E.K., 2014: Influence of slope on physical soil disturbance due to farm tractor forwarding in a Hyrcanian forest of northern Iran. iForest 7(5): 342–348.

Larminie, J.C., 1992: Modifications to the mean maximum pressure system. Journal of Terramechanics 29(2): 239–255.

Lysne, D.H., Burditt, A.L., 1983: Theoretical ground pressure distributions of log skidders (forest equipment). Trans.

ASAE 26: 1327–1331.

Lyasko, M.I., 1994: The determination of deflection and contact characteristics of a pneumatic tire on a rigid surface.

Journal of Terramechanics 31(4): 239–242.

Maclaurin, E.B., 1990: The use of mobility numbers to describe the in-field tractive performance of pneumatic tyres.

Proceedings of the 10th ISTVS Conference, Kobe, 8/1990, 177–186.

Neruda, J., Ulrich, R., Kupčák, V., Slodičák, M., Zemánek, T., 2013: Harvestorové technologie lesní těžby Mendelova univerzita v Brně. ISBN 978-80-7375-842-4, 166 p.

Pacas, B., 1983: Teorie stavebních stroju. Vysoké učení technické v Brně. Publication Nr. 411-33363, 244 p.

Rowland, D., 1972: Tracked vehicle ground pressure and its effect on soft ground performance. Proceedings of the 4th ISTVS Conference, April, Stockholm, Sweden, 353–384.

Saarilahti, M., 2002: Soil interaction model. University of Helsinki, Department of Forest Resource Management, Np. QLK5-1999-00991, 86 p.

Saarilahti, M., Anttila, T., 1999: Rut depth model for timber transport on moraine soils. Proceedings of the 13th ISTVS Conference, September 1999, Munich, Germany, 29–37.

Schwanghart, H., 1990: Measurement of contact area, contact pressure and compaction under tires in soft soil. Proceedings of the 10th ISTVS Conference, August, Kobe, Japan, 193–204.

Zahradníček, R., Semrád, K., 2007: Pružnosť a pevnosť II. Košice, TU, LF, 144 p.

Zemánek, T., Neruda. J., Ulrich, R., 2015: Okamžité tlaky v pude vyvozované přejezdy lesní techniky. Lesnictví. In Mobilné energetické prostriedky – hydraulika – životné prostredie – ergonómia mobilných strojov, Zvolen, Technická univerzita vo Zvolene, 171–178.

Authors’ addresses:

Received: June 18, 2017 Accepted: December 17, 2017

Milan Marušiak, MSc. * e-mail: milan.marusiak@gmail.com Prof. Jindřich Neruda, PhD. e-mail: neruda@mendelu.cz University in Brno Zemĕdĕlská 3 613 00 Brno CZECH REPUBLIC *Corresponding author

Croat. j. for. eng. 39(2018)2 245

Original scientific paper

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density Ahmad Solgi, Ramin Naghdi, Eric R. Labelle, Petros A. Tsioras, Ali Salehi Abstract The objective of this study was to compare forest soil bulk density values obtained through conventional sampling methods such as the volumetric ring (VR: diameter 5 cm, length 10 cm) and paraffin sealed clod (PSC), with a variation of the VR, where rectangular boxes (RB) of four different dimensions were used. Sampling transects were established on a machine operating trail located in a beech (Fagus orientalis Lipsky) stand in Northern Iran. At each transect, three soil samples were collected at three different locations. Samples from different methods were spaced by a 50 cm distance to avoid direct interactions. The soil class of our study area was Combisols according to the WRB classification with a clay texture. Soil bulk density differed significantly between the three sampling methods. The lowest values were obtained with the RB (average 1.25 g cm-3), followed by the VR (average 1.40 g cm-3), and lastly the PSC (average 1.52 g cm-3). The values obtained with four variations of the RB method ranged from 1.22 to 1.28 g cm-3 and were not found significantly different. When soil bulk density was calculated after the removal of the weight and volume of roots included in the samples, the values were determined to be higher than before but with the same range of magnitude. The lowest coefficient of variation was found for RB4 (CV=2.3%), while the highest values were observed for VR and RB1 (CV=5.7%).

Keywords: paraffin sealed clod, rectangular box, sampling method, soil compaction, volumetric ring

1. Introduction Forest management practices, such as timber harvesting and off-road timber transportation, have the potential to cause detrimental levels of soil and site disturbances (Kozlowski 1999, Najafi and Solgi 2010, Labelle and Jaeger 2011, Naghdi et al. 2015). The growing mechanization of forest operations combined with higher machine payload increases the magnitude of soil compaction, resulting in a decrease in soil macroporosity (i.e., cavities larger than 0.08 mm in diameter) (Berli et al. 2004, Frey et al. 2009, Solgi et al. 2015a, Solgi et al. 2015b). This decrease, thereby, reduces the rate of exchange of air, water, and solutes (Greacen and Sands 1980, Botta et al. 2007).

In forest operations, the use of dry bulk density as a measure of estimating soil compaction is common (Pires et al. 2005a). Soil dry bulk density is defined as the mass of dry soil particles in a unit volume of soil (Craig 2004). During compaction, solid particles do not

change in volume, but are subject to rearrangement, which may be accompanied by bending of clay platelets, changes in the shape of organic matter, and breaking of bonds (Soane and Van Ouwerkerk 1994).

Many studies presented and compared methods to quantify soil bulk density with both modern and conventional sampling procedures (Pires et al. 2005a, Timm et al. 2005). One of the conventional and standard sampling methods used in the majority of studies is the volumetric ring method (Grossman and Reinsch 2002). Standard dimensions and construction material of the volumetric ring are used by some researchers.

However, the method is highly dependent upon the specific location where the sample is taken (Lestariningsih et al. 2013). This method is susceptible to error arising from compression and vibrations or shattering of the core, while the cylinder is inserted into the soil profile with external force. According to Pires et al.

(2005a), there is a tendency of compaction near the cylinder walls and in the top and the bottom regions

Croat. j. for. eng. 39(2018)2 247

A. Solgi et al.

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247–254)

of samples taken with the volumetric ring method.

Thus, the risk of sample compression increases with decreasing core diameter (Soane and Van Ouwerkerk 1994). Freitag (1971) suggested that the diameter should be selected to give an adequate sample size and that the length should be no more than three times the diameter.

In the forest, trees have high root volumes that occupy the soil space, in particular the topsoil. Under such circumstances, the estimation of bulk density by means of a cylinder is becoming a very difficult process. Parts of tree roots are often included in the core sample, making the sampled soil quantity unsuitable for further analysis. Therefore, a large number of samples must be collected with this method in order to have the required statistical validity of the results, a prerequisite that increases sampling cost.

Another conventional soil sampling method is that of the paraffin sealed clod. In this method, the clod is weighted and its volume is determined by coating it in paraffin wax and immersing it in a volumenometer (Soane and Van Ouwerkerk 1994). The next step includes removal and weighing of the wax. Reliable measurements can be performed especially on cohesive soils, but the method remains time-consuming as care has to be taken to ensure that the wax coats do not penetrate the soil pore system. Van Remortel and Shields (1993) reported that the penetration of some paraffin into the pores of the clods, reduced their measured volume, resulting in higher soil bulk density values.

Apart from the conventional methods, modern methods such as computed tomography and nuclear moisture and density gauge have been introduced to forest soil bulk density measurements (Petrovic et al.

1982, Timm et al. 2005, Labelle and Jaeger 2011). These methods are advantageous in many aspects; the computed tomography provides a detailed analysis of soil bulk density variation along a sample, while the nuclear moisture and density gauge allows for repeated measurements of an identical area without extracting and destructing the soil sample (Timm et al. 2005, Labelle and Jaeger 2011). Despite these important advantages, the cost of this specific equipment can be prohibitive, especially in the case of developing countries and as such the computed tomography and the nuclear moisture and density gauge methods were not targeted during our study. In this view, the rationale behind this research effort was to improve our knowledge on the use of rectangular boxes of different dimensions as an alternative to the widely-used sampling cylinders of the volumetric method. Samples were collected with the volumetric ring, paraffin

248

sealed clod and rectangular box methods, and the respective soil bulk density values were compared with the aim of assisting future research endeavours on the choice of the more appropriate method.

2. Materials and Methods 2.1 Site description The study was conducted in compartment 41 of the third district of Shenrood forest, Guilan Province, northern Iran (between 36°31′56″ N and 36°32′11″ N latitude and 51°47′49″ E and 51°47′56″ E longitude).

The forest is comprised predominantly of oriental beech (Fagus orientals Lipsky) with an average canopy cover of 80%, stand density of 170 trees ha-1, mean tree diameter at breast height of 29.7 cm, and mean tree height of 22.9 m. The study area has an elevation of approximately 800 m above sea level and a northerly aspect. The average annual rainfall recorded at the closest national weather station situated about 20 km away from the research site is 970 mm. The maximum mean monthly rainfall of 120 mm usually occurs in October, while the minimum rainfall of 25 mm occurs in August. The mean annual temperature is 15 °C, with the lowest values recorded in February. At the time of skidding, weather conditions were dry and warm with an average soil moisture content of 210 g kg-1 (21%).

The soil class of our study area in soil classification according to WRB was Combisols. Soil texture in the studied machine operating trail was determined based on particle-size analysis using the Bouyoucos hydrometer method (Kalra and Maynard 1991) and was classified as clay (Table 1).

Table 1 Soil particle size distributions at different depths of the skid trail Horizon

Depth, cm

Sand, %

Silt, %

Clay, %

A

0–15

26

26

48

B

15–55

23

28

49

C

55–85

21

29

50

The range of particle size in Table 1 was 0.05–2 mm, 0.002–0.05 and <0.002, for sand, silt, and clay, respectively. Soil texture in all three depths was clay.

The soil had not been driven on before the experiment. Ground-based skidding operations were performed with a Timberjack 450 C rubber-tired skidder (Table 2). The rubber-tired skidder was used to extract 3 to 4 m long logs from the felling site to the nearest Croat. j. for. eng. 39(2018)2

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247â&#x20AC;&#x201C;254)

Table 2 Technical details of the rubber-tired skidder Timberjack 450 C Specifications Weight, kg Number of wheels Tire size, mm

Timberjack 450C 10,257 4 775x813

Ground pressure, kPa

221

Engine power, hp

177

Year of manufacture

1998

Manufacturing location

Canada

forest road on machine operating trails in flat terrain.

Traffic frequency of the loaded skidder was twelve passes.

2.2 Experimental design and data collection Soil bulk density was measured with three different methods. The first method of the volumetric ring (hereafter VR) has been extensively used and could be considered as the standard method of soil bulk density measurement. The volumetric ring was made from 1.5 mm thick stainless steel with an inside diameter of 5 cm and a length of 10 cm (Fig. 1).

The second method made use of paraffin sealed clods (hereafter PSC). The sampling procedure in this case consisted of excavating the soil surface down to a depth of 10 cm with a spade. The removed soil clod was wrapped in a plastic film directly after sampling to

Fig. 1 Different sampling methods used in the study A: Rectangular Boxes and Volumetric Rings; B: Volumetric Rings; C and D: Rectangular Boxes

A. Solgi et al.

avoid water loss, and care was taken during the transportation to minimize alterations in the clod structure.

The third method referred to the use of a rectangular box (hereafter RB) as a means for collecting soil samples. This method resembles the VR, with the exception of the shape of the sampling device (Fig. 1).

Four rectangular boxes with different dimensions (RB1: length 20 cm, width 20 cm, height 10 cm, RB2: length 20 cm, width 15 cm, height 10 cm, RB3: length 20 cm, width 10 cm, height 10 cm, and RB4: length 15 cm, width 10 cm, height 10 cm) were used in our study. All rectangular boxes were made from 1.5 mm thick steel.

The characteristics of these rectangular boxes are given in Table 3.

Table 3 Inside dimensions of rectangular boxes Thickness[a] mm

Length cm

Width cm

Height cm

RB1

1.5

20

20

10

RB2

1.5

20

15

10

RB3

1.5

20

10

10

RB4

1.5

15

10

10

Rectangular box

In order to compare the above mentioned methods, a 200 m longitudinal segment was chosen inside the beech stand, along which samples were taken from the soil surface layer (0â&#x20AC;&#x201C;10 cm). More specifically, soil samples were collected from eight sampling transects each separated by 25 m and aligned perpendicular to the 200 m long segment. At each transect, 18 samples (three for each method including the different variations of RBs) were collected at three different locations: on the left track, between tracks, and on the right track (Fig. 2). Volumetric ring, clod, and rectangular boxes samples were spaced 50 cm apart to create a buffer zone, thus avoiding possible interactions. A total of 144 soil samples were collected for the analysis.

At the time of sampling (VR and RBs methods), we placed a piece of wood flush on one of the extremity of the rings and boxes and then we inserted those into the soil by hitting the piece of wood with a hammer.

Using a piece of wood prevented damaging the top of rings and boxes as to maintain their shape and volume correctly. After extracting the rings and boxes from the soil with minimal disturbance to their contents, the soil samples were trimmed flush with the ring and box end and extruded into plastic bags for transportation to the laboratory. Immediately after sampling, the samples were brought to the laboratory and were

Croat. j. for. eng. 39(2018)2 249

A. Solgi et al.

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247–254)

Fig. 2 Schematic of the treatment set-up with the location of sampling transects within the research area promptly weighed. Soil samples were dried in an oven at 105 ºC (24 h) to obtain the dry weight value. The water content in the soil samples was measured gravimetrically after drying in an oven. However, due to the large amount of wet soil mass in the rectangular boxes, dry soil weight was counted based on a subsampling method in which approximately 300 g of wet soil contained in a rectangular box was measured in the laboratory and oven-dried at 105 ºC (24 h) to obtain the dry weight value (Lestariningsih et al. 2013). For each sample originating from a rectangular box, wet weight and volume of tree roots were determined in a laboratory.

Uncorrected bulk densities from rectangular boxes were computed with Eq. (2):

Soil bulk density value from volumetric ring was computed using Eq. (1):

Where: ρd dry bulk density, g cm-3 Wt total mass of rectangular box and soil from the field, g Wb mass of rectangular box, g Wr mass of wet root, g Swc soil water content, g g-1 Vt volume of apparatus, cm-3 Vr volume of roots, cm-3.

ρd =

Where: ρd Wd Vc

250

Wd Vc

dry bulk density, g cm-3 mass of dry soil, g volume of soil cores, 196.25 cm3.

(1)

 ( Wt − Wb − Wr )    (1 + Swc )  

ρd = Vt

(2)

Corrected bulk densities (for root correction) from rectangular boxes were computed with Eq. (3):

 ( Wt − Wb − Wr )    (1 + Swc )   ρd =

( Vt − Vr )

(3)

Croat. j. for. eng. 39(2018)2

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247–254)

A. Solgi et al.

Total soil porosity was computed with Eq. (4):

 ρd  AP =− 1  × 100 2.65  

(4)

Where: AP total porosity, % ρd dry bulk density (g cm-3), and 2.65 (g cm-3) is the assumed particle density (Najafi et al. 2009).

2.3 Statistical Analysis Mean values of soil physical properties of each method were compared to those in other methods using Tukey’s HSD test (Zar 1999). One-way ANOVA (significance test criterion a≤0.05), included in the SPSS statistical package version 11.5, was used to compare the soil physical properties among the different methods. Paired t-tests were used to analyze soil bulk density data from the »without« and »after« root correction among RBs methods at an a level of 0.05.

3. Results and Discussion In the context of the study, soil bulk density values were used as indicator of soil compaction in an oriental beech stand. This forest species was chosen because of its importance in the Hyrcanian forest area in Iran, where it covers 16.5% (245 372 ha) of the total forested area and represents approximately 25% of the wood harvested annually (Sagheb Talebi et al. 2014). However, it should be noted that oriental beech has a high density of roots that are located near the soil surface, which increased difficulty during the soil sampling process. The post-hoc Tukey test showed that the average values of PSC, VR, and RB differed statistically. However, this was not the case in the RBs, where measurement differences were not statistically significant (Fig. 3).

The lowest soil bulk density values were obtained with the RBs in the range of 1.22 to 1.28 g cm-3 (average 1.25 g cm-3) and the highest with the PSC method (1.52 g cm-3). The average bulk density value of the VR (1.40 g cm-3) was 11.7% higher than that of RBs and 8.7% lower than that of PSC. This result is in line with Lestariningsih et al.

(2013), who also obtained lower soil bulk density values compared to the VR for box dimensions similar to our RB1 configuration. Pires et al. (2004) showed that the VR method induces changes in soil structure during sampling procedures, mainly for small soil samples, causing under- and over-estimated bulk density values.

These modifications in bulk density occur due to compaction close to the cylinder walls and, in some cases, at the top and bottom regions of the soil sample. The same problem was observed by Pires et al. (2005b) while working with different soil samplers.

Fig. 3 Comparison of soil dry bulk densities from values determined by different sampling methods (means followed by the same letter are not significantly different at p=0.05, error bars indicate standard deviations) Comparing the sample area can lead to some interesting results. The surface area of VR is 19.63 cm2, while in the RB method it increases from 150.00 cm2 (RB4) up to 400.00 cm2 (RB1). Higher bulk density values in VR could be partly explained by the increased vibration and compression forces associated with this sampling technique and the effect of sampler walls on samples of smaller area. However, this trend has not been verified when comparing the various RB samples. RB1 and RB4 had higher average soil density values than RB2 and RB3. Thus, more research is needed to clarify the reason for this and possibly define the optimum dimensions for the RB method.

The PSC samples were found to be more compacted than those of the VR, confirming the observations made by Van Remortel and Shields (1993) and Timm et al. (2005). Both studies suggest that bulk density obtained with the PSC method are, in general, higher than those collected with the VR method, due to some paraffin penetration into the pores of the clod, causing a reduction of the volume measurement.

In rectangular box methods, when root correction is included (the volume and weight of roots included in the samples were subtracted from the total soil sample size and soil compaction was calculated based on the new data), bulk density values measured with volumetric rings and paraffin sealed clods continue to be higher than those obtained with rectangular boxes

Croat. j. for. eng. 39(2018)2 251

A. Solgi et al.

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247â&#x20AC;&#x201C;254)

Table 4 Percentage differences between measurements of various RB dimensions, PSC and VR (comparison of soil bulk densities measured with six sampling methods (N=24)) PSC[a]

VR[b]

15.6%

8.3%

RB2

18.5%

11.5%

RB3

19.6%

12.7%

RB4

16.4%

9.2%

RB1

21.1%

14.3%

RB2

24.3%

17.7%

RB3

26.2%

19.8%

20.8%

13.9%

RB1[c] With root correction

Without root correction

RB4 [a]

[b]

[c]

Paraffin Sealed Clod; Volumetric Ring; Rectangular Boxes

(Table 4). Comparison of the bulk density values by means of the paired sample t-test showed that there was a significant difference between bulk density before and after root correction among the same sampling methodologies (Fig. 4). Similar results have been reported by Lestariningsih et al. (2013).

The spatial variability of soil bulk density values, with an error pattern, along the 200 m longitudinal segment for the various sampling methods is present-

Fig. 5 Distribution of soil dry bulk density values along 200 m longitudinal segment determined by different sampling methods, error bars indicate standard deviations ed in Fig. 5. Table 5 presents average soil bulk density values determined by each method, the respective standard deviations and coefficients of variation (CV).

Soil bulk density values obtained by the RB4 method present the smallest variation in relation to the average value (CV=2.3%), while that obtained by the VR and RB1 methods demonstrated the highest variation (5.7% in both cases). Variation in both methods is at low levels, especially considering the sample size. Statistical analysis revealed that standard deviation value was significantly different between samples from different methods with p<0.05. Standard deviation of soil Table 5 Average soil bulk density values determined by the six different methods and respective standard deviations and coefficients of variation (N=24) Dry bulk density g cm-3

Standard deviation g cm-3

Coefficient of variation %

1.52a

0.04

2.6

VR

b

1.39

0.08

5.7

RB1

1.28c

0.07

5.7

RB2

c

1.23

0.06

4.8

RB3

1.22c

0.06

4.9

RB4

c

0.03

2.3

Methods PSC

Fig. 4 Average soil dry bulk density before and after root correction for rectangular box methods (means followed by a different letter are significantly different at p=0.05, error bars indicate standard deviations)

252

1.27

Croat. j. for. eng. 39(2018)2

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247–254)

bulk density values from VR was higher than PSC and RB1 values indicating a larger heterogeneity in the VR samples. This larger heterogeneity could have been induced during the sampling procedure through modifications in soil structure caused by a limited volume of soil sample inside the cylinders, as shown by Pires et al. (2004).

Also, our results showed that the standard deviation of soil bulk density values from the PSC samples were equal to those obtained with the RB1 method.

Our finding is similar to that of Timm et al. (2005), who reported a coefficient of variation of 5.9% for soil density values obtained with the volumetric ring method.

Soil bulk density values were obtained for each method using Eq. (4), since for the PSC and VR methods the bulk density values were higher, and consequently the total porosity estimates along the segment were smaller (on average 35% and 46%, respectively).

The highest total soil porosity values were obtained with the RBs in the range of 51.76% to 54.27% (on average 52.85%) and the lowest with the PSC method (35.52%), which, according to Kiehl (1979), is within the range of mineral soil porosity variations, usually from 40% to 60%. The average total soil porosity value of the VR (46.13%) was 14.6% lower than that of RBs and 29.8% higher than that of PSC, respectively. This fact shows the importance of the choice of the method for the determination of soil bulk density.

4. Conclusion In our study, we compared the well-established method of volumetric ring to that of paraffin sealed clod and four rectangular boxes of varying dimensions under forest conditions. Statistical differences in soil dry bulk density have been found between the three general methods but not among the four rectangular boxes, indicating substantial differences between the three general methods. While the volumetric ring is the most commonly used methodology and is supported by a large amount of international literature, there is evidence that it can overestimate soil dry bulk densities due to increased compaction and vibrations during the sampling process, in particular with smaller-sized cylinders.

The rectangular box method yielded promising results when used in the conditions (soil texture, water content and root size/density) observed at our study site. However, these results need to be tested in other conditions, such as different soil texture, soil moisture content, soil depth and soil organic matter content, before drawing general conclusions or giving recommendations.

A. Solgi et al.

5. Reference Berli, M., Kulli, B., Attinger, W., Keller, M., Leuenberger, J., Flühler, H., Springman, S.M., Schulin, R., 2004: Compaction of agricultural and forest subsoils by tracked heavy construction machinery. Soil Till Res 75(1): 37–52.

Botta, G., Pozzolo, O., Bomben, M., Rosatto, H., Rivero, D., Ressia, M., Tourn, M., Soza, E., Vazquez, J., 2007: Traffic alternatives in harvest of soybean (Glycine max L.): effect on yields and soil under direct sowing system. Soil Till Res 96(1–2): 145–154.

Craig, R.F., 2004: Soil Mechanics. Seventh edition. Spon press.

British Library Cataloguing in Publication Data, 95 p.

Freitag, D.R., 1971: Methods of measuring soil compaction.

In: K.K. Barnes, W.M. Carleton, H.M. Taylor, R.I. Throckmorton and G.E. Vanden Berg (Editors), Compaction of Agricultural Soils. Am. Soc. Agric. Eng., St. Joseph, MI, USA, 47–103.

Frey, B., Kremer, J., Rüdt, A., Sciacca, S., Matthies, D., Lüscher, P., 2009: Compaction of forest soils with heavy logging machinery affects soil bacterial community structure. Eur J Soil Biol 45(4): 312–320.

Greacen, E.L., Sands, R., 1980: Compaction of forest soils: a review. Aus J Soil Res 18(2): 163–189.

Grossman, R.B., Reinsch, T.G., 2002: Bulk density and linear extensibility. In: Dane JH, Topp C, Co-editors. Method of soil analysis part 4 physical methods, Madison, Wisconsin: Soil Sci Soc Am Inc, 201–228.

Jourdan, C., Rey, H., 1997: Architecture and development of the oil-palm (Elae isguineensis Jacq.) root system. Plant Soil 189(1): 33–48.

Kalra, Y.P., Maynard, D.G., 1991: Methods and manual for forest soil and plant analysis. Forestry Canada, Re NORX-319. Northern Forestry Center, 125 p.

Kiehl, E.J., 1979: Manual de edafologia. São Paulo: Agronômica Ceres, 262 p.

Kozlowski, T.T., 1999: Soil compaction and growth of woody plants. Scand J For Res 14(6): 596–619.

Labelle, E.R., Jaeger, D., 2011: Soil compaction caused by cutto-length forest operations and possible short-term natural rehabilitation of soil density. Soil Sci Soc Am J 75(6): 2314– 2329.

Lestariningsih, I.D., Widianto, K.H., 2013: Assessing soil compaction with two different methods of soil bulk density measurement in oil palm plantation soil. Proc Environ Sci 17: 172–178.

Naghdi, R., Solgi, A., Zenner, E.K., 2015: Soil Disturbance Caused by Different Skidding Methods in North Mountainous Forests of Iran. Inter J For Eng 26(3): 212–224.

Najafi, A., Solgi, A., Sadeghi, S.H., 2009: Soil disturbance following four-wheel rubber skidder logging on the steep trail in the north mountainous forest of Iran. Soil Till Res 103(1): 165–169.

Najafi, A., Solgi, A., 2010: Assessing site disturbance using two ground survey methods in a mountain forest. Cro J For Eng 31(1): 47–55.

Croat. j. for. eng. 39(2018)2 253

A. Solgi et al.

Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density (247–254)

Pires, L.F., Arthur, R.C.J., Camponez Do Brasil, R.P., Correchel, V., Bacchi, O.O.S., Reichardt, K., 2004: The use of gamma ray computed tomography to investigate soil compaction due to core sampling devices. Braz J Phys 34(3A): 728–731.

Pires, L.F., Pilotto, J.E., Timm, L.C., Bacchil, O.O.S., Reichardt, K., 2005a: Qualitative and quantitative analysis of soil samples by computerized tomography. Publ UEPG Ci Exatas Terra Ci Agr Eng Ponta Grossa 11(2): 7–15.

Pires, L.F., Arthur, R.C.J., Bacchil, O.O.S., 2005b: Assessment of soil sample quality used for density evaluations through computed tomography, International Nuclear Atlantic Conference – INAC 2005, Santos, SP, Brazil, August 28 to September 2, 5 p.

Sagheb Talebi, K., Sajedi, T., Pourhashemi, M., 2014: Forests of Iran a treasure from the past, a hope for the future. Plant and Vegetation. Volume 10. ISBN 978-94-007-7370-7 (Print) 978-94-007-7371-4 (Online), 143 p.

Soane, B.D., Van Ouwerkerk, C., 1994: Soil Compaction in Crop Production. Developments in Agricultural Engineer-

ing Series, vol. 11. Elsevier Science, Amsterdam, Netherlands, 643 p.

Solgi, A., Naghdi, R., Tsioras, P.A., Nikooy, M., 2015a: Soil compaction and porosity changes caused during the operation of Timberjack 450C skidder in northern Iran. Cro J For Eng 36(2): 77–85.

Solgi, A., Naghdi, R., Nikooy, M., 2015b: Effects of skidder on soil compaction, forest floor removal and rut formation.

Madera y Bosques 21(2): 145–153.

Petrovic, A.H., Siebert, J.E., Lieke, P.E., 1982: Soil bulk density analyses in three dimension by computed tomographic scanning. Soil Sci Soc Am J 46(3): 445–450.

Timm, L.C., Pires, L.F., Reichardt, K., Roveratti, R., Oliveira, J.C.M., Bacchi, O.O.S., 2005: Soil bulk density evaluation by conventional and nuclear methods. Aust J Soil Res 43(1): 97–103.

Van Remortel, R.D., Shields, D.A., 1993: Comparison of clod and core methods for determination of soil bulk density.

Commun Soil Sci Plant Ana 24(17–18): 2517–2528.

Zar, J.H., 1999: Biostatistical analysis, 4th edition, Prentice Hall, Upper Saddle River, NJ, USA, 662 p.

Authors’ addresses: Ahmad Solgi, PhD. e-mail: aforestsolgi@gmail.com Prof. Ramin Naghdi, PhD.* e-mail: rnaghdi@guilan.ac.ir Assoc. prof. Ali Saleh, PhD. e-mail: asalehi@guilan.ac.ir Department of Forestry Faculty of Natural Resources University of Guilan P.O. Box 1144 Sowmeh Sara Guilan IRAN Assist. prof. Eric R. Labelle, PhD. e-mail: eric.labelle@tum.de Department of Ecology and Ecosystem Management Technische Universität München Hans-Carl-Von-Carlowitz-Platz 2 85354 Freising GERMANY

Received: June 27, 2017 Accepted: December 12, 2017

254

Tsorias A. Petros e-mail: ptsioras@for.auth.gr Laboratory of Forest Utilization Faculty of Forestry and Natural Environment Aristotle University of Thessaloniki GR-54124 Thessaloniki GREECE *Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran Ehsan Abdi Abstract Shallow landslides are a frequently recurring problem in some parts of Iran, including the Hyrcanian forest. In addition to traditional civil engineering measures, a potential solution for this problem is the application of soil bioengineering techniques. The mechanical reinforcement effect of plant roots is one of the major contributions of vegetation to the mitigation of shallow landslides. Given the lack of information on the mechanical properties of common Hyrcanian forest species, the present study assessed the root strength of 10 common species of this forest. Eight tree species occurring in natural regeneration sites (Carpinus betulus, Fagus orientalis, Parrotia persica and Quercus castaneifolia) and plantations (Acer velutinum, Alnus glutinosa, Fraxinus excelsior and Picea abies) and two shrub species (Crataegus microphylla and Mespilus germanica) were selected. Fresh roots were collected and mechanical tests were carried out on 487 root samples. The ranges of root diameter, tensile force, and root resistance were 0.29â&#x20AC;&#x201C;5.90 mm, 3.80â&#x20AC;&#x201C;487.20 N, and 2.41â&#x20AC;&#x201C;224.35 MPa, respectively. Two different algorithms, including the nonlinear least square method and logtransformation, were used to obtain power regressions for diameter-force and diameter-resistance relationships. The results of the two algorithms were compared statistically to choose the optimal approach for soil bioengineering applications. The nonlinear least square method resulted in lower Akaike information criteria and higher adjusted R2 values for all species, which means that this model can more efficiently predict tensile force and resistance based on root diameter. Log-transformation regressions generally underestimate tensile force and resistance. Significant differences were found among mean root tensile force (ANCOVA; F=37.36, p<0.001) and resistance (ANCOVA; F=34.87, p<0.001) of different species. Also, root diameter was significant as a covariate factor in tensile force (F=1453.77, p<0. 001) and resistance (F=274.26, p<0.001). Shrub species and trees in natural regeneration sites had higher tensile force and resistance values, while trees from plantation stands had lower values.

The results of this study contribute to the knowledge on the root force and resistance characteristics of several shrub and tree species of the Hyrcanian forest and can be used in evaluating the efficiency of different species for bioengineering purposes.

Keywords: landslides, log-transformation, nonlinear least square, power regression, soil bioengineering, stability

1. Introduction Shallow landslides are a frequently recurring problem in some parts of Iran, including the Hyrcanian forest. In Iran, the total estimated losses as a result of landslides are about $ 50 million per year (Ebrahimi et al. 2015). Landslides are defined as processes that result in the downward and outward movement of

slope-forming materials, with gravity and water as the primary triggers (Stokes et al. 2014). In the Hyrcanian forest, landslides pose a severe threat to access infrastructure, including forest roads, resulting in severe economical consequences. Among the different landslide triggering mechanisms in this forest environment, rainfall-induced (Abedi et al. 2010) and humaninduced (Savadkoohi and Hosseini 2013) landslides

Croat. j. for. eng. 39(2018)2 255

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

are the most frequently reported. Human-induced landslides usually occur after disturbances such as road construction, modifying the shape of the slope, removing vegetation cover, and decreasing both the density and resistance of roots (Vergani et al. 2016); such processes might also contribute to the concentration of flow and increase pore-water pressure, resulting in local instabilities (Schwarz et al. 2010b). In contrast, rainfall-induced landslides are generally triggered by high-intensity rainfall, which causes a sudden increase in soil moisture content and a decrease in soil suction, leading to soil strength reduction and possible failure (Cislaghi et al. 2017, Hayati et al.

2017). In general, high soil moisture contents (or low suction) lead to weaker apparent soil cohesion and higher landslide risks (Stokes et al. 2014).

Traditional civil engineering measures (i.e. grey solutions) with high initial costs and increasing maintenance needs over time are unsuitable in the long term (Morgan and Rickson 1995), especially in natural resources with extensive areas. Potential solutions for this problem are soil bioengineering measures or green solutions, which are characterized by the use of any form of vegetation (grass, shrubs, or trees) as materials to perform engineering functions (Morgan and Rickson 1995) such as soil reinforcement, erosion control, and the prevention of shallow instability. In recent years, using plants for soil bioengineering measures has been recognized as an eco-friendly and low CO2 emission solution for soil stabilization, as compared to the existing traditional grey or ÂťhardÂŤ engineering solutions (Boldrin et al. 2017). Plants are selfregenerating and respond dynamically to changes of site conditions without losing their engineering properties (Morgan and Rickson 1995); in addition, they can improve the stability of hillslopes and the ecological conditions (Bischetti et al. 2010).

Vegetation can increase slope stability by protecting and holding soil particles together, mechanically reinforcing the soil and increasing soil matric suction (Capilleri et al. 2016) through both interception of rainfall and depletion of soil water content via transpiration (Hayati et al. 2017). This is particularly the case for forest environments, where mechanical and hydrological modifications by trees enhance the stability of hillslopes (Moos et al. 2016). The mechanical effect of root systems has been recognized as one of the major contributions of vegetation to the mitigation of shallow landslides (Vergani et al. 2016). Some researchers have reported a negative correlation between the magnitude of root reinforcement and landslide susceptibility (Hubble et al. 2013, Roering et al. 2003, Schmidt et al. 2001, Moos et al. 2016), and it is now clear that

256

plants positively influence the triggering mechanisms via root strength, hydrological regulation, and root anchorage (Cislaghi et al. 2017). Thick roots act like soil nails on slopes, reinforcing soil in the same way that steel reinforces concrete. Thin and fine roots act in tension during failure on slopes; if they cross the slip surface, they reinforce the soil by adding additional cohesion to the soil cohesion (Stokes et al. 2014). The efficiency of roots in reinforcing soil depends on root strength resistance, root distribution, and morphology. The greater the strength and the wider the distribution, the better the plants will reinforce the soil (Stokes 2002). With increasing root strength, larger masses of soil are needed to overcome resisting forces and, therefore, the critical landslide area increases (Moos et al.

2016). It is worth mentioning that the effectiveness of root reinforcement is limited to shallow landslides with a volume of less than about 1000 m3 (Giadrossich et al. 2017), which includes most of the shallow landslides in the Hyrcanian forest.

Regarding the important role of plant roots in soil bioengineering, many studies have assessed root tensile resistance, which varies widely from thousands to millions of Pascal, depending on the species and the environment (Nilaweera and Nutalaya 1999, Abernethy and Rutherfurd 2001, Schmidt et al. 2001, Tosi 2007, Genet et al. 2008, Schwarz et al. 2010a, Vergani et al. 2012, Giadrossich et al. 2016). Tensile resistance depends on a variety of factors such as plant species (Stokes 2002), root diameter (Watson et al. 1999, Bischetti et al. 2005), soil environment (Goodman and Ennos 1999), time of year (Abernethy and Rutherfurd 2000), management type (Coppin and Richards 1990), test speed, sample length and diameter, root moisture and storage (De Baets et al. 2009, Hales and Miniat 2016), chemical composition (Genet et al. 2005), orientation along the slope (Abdi et al. 2010), plant age and altitude (Vergani et al. 2014), and the mechanical role of the root (Stokes 2002). Although some researchers have reported that variations in root tensile resistance are dependent on the species (e.g. Bischetti et al. 2009, Abdi et al. 2010, Vergani et al. 2012), many previous studies have focused on one or a few species (e.g. Watson et al. 1999, Tosi 2007, Genet et al. 2008, Abdi et al. 2009, Abdi et al. 2010). However, assessing and comparing the mechanical properties of several species provide valuable data for ranking species in terms of their potential role in bioengineering applications (Watson and Marden 2004).

The magnitude of soil reinforcement due to the presence of roots (Cr) can be modeled using different methods, e.g. the Wu method (Wu et al. 1979), the fiber bundle model or FBM (Pollen and Simon 2005), the Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

root bundle model or RBM (Schwarz et al. 2010a), and the root bundle model Weibull or RBMw (Schwarz et al. 2013). Apart from the model type, all models consider Cr as a function of root tensile resistance (in Wu and FBM models) or tensile force (in RBM and RBMw models) and of root distribution within the soil.

Concerning the engineering applications, the quantification of the tensile force and the resistance of roots are key parameters for several fields of application, including slope stability (Vergani et al. 2012); root reinforcement estimation (Vergani et al. 2014); erosion control measures (Giadrossich et al. 2016), and soil bioengineering technique design (Bischetti et al. 2010).

For example, in soil bioengineering applications in forest engineering, such as brush layering for the stabilization of road cuts and fill-slopes (Bischetti et al. 2010), wattle fences for stabilizing uphill cut-slopes, and brush wattles for roadside slope stabilization (Schiechtl 1980), plant roots play an important engineering role, and root tensile force and resistance values are needed to estimate the magnitude of the bioengineering effectiveness (Bischetti et al. 2010).

In the current study, root tensile force and resistance of 10 typical species (two shrub and eight tree species) of the Hyrcanian forest were investigated to expand our knowledge of the values typical of this environment and to compare and assess the variability among species. Although some studies have reported the tensile resistance of Hyrcanian forest species (e.g., Abdi et al. 2009, Abdi et al. 2010), so far, no study has assessed different species, especially trees with different stand origin (i.e. natural or artificial regeneration) and shrubs. Also, most of the previous studies have used log-transformation regression as one of the most common patterns in biology (Xiao et al. 2011), which was introduced approximately a century ago (Packard 2012) to express the relation between root tensile force and tensile resistance as a function of root diameter (e.g. Bischetti et al. 2005, Bischetti et al. 2009, Vergani et al. 2012, Vergani et al.

2014). However, Schwarz et al. (2013) and Giadrossich et al. (2017) reported that different regression methods (log-transformation and nonlinear least square) for fitting of the root diameter–force and resistance curve lead to quite different coefficients of the equation, and these changes lead to variations in the estimated reinforcement effect of vegetation. Also, in several biological studies, the use of log-transformation power regression has been criticized, suggesting that the analysis on logarithmic scales is flawed and that instead, analyses should be carried out on the original measurement scale, using nonlinear regression (e.g.

Fattorini 2007, Packard 2009, Packard 2011, Packard

Ehsan Abdi

and Birchard 2008, Packard et al. 2011). Therefore, we used two power regression methods, the Akaike information criteria (AIC) and adjusted R2, as statistical criteria for selecting the optimal model (Zuur et al.

2007, Xiao et al. 2011, Lai et al. 2013).

In this context, the objectives of this study were as follows: i) to investigate to which extent root tensile force and resistance depend upon the species and ii) to investigate the effects of different regression methods (nonlinear least square and log-transformation) on the power regression coefficients for the relationship between root tensile force and tensile resistance as a function of root diameter.

2. Material and Methods 2.1 Characteristics of the study site The Hyrcanian vegetation zone, also called the »Caspian forest«, is a green belt stretching across the northern slopes of the Alborz Mountain Ranges and covering about 1.9 million hectares. The area is rich in hardwood species, with about 50 tree and 80 shrub species. Broadleaved species are dominant, and some small stands of softwoods have been artificially introduced to this forest about 40 years ago. The main tree species are Fagus orientalis, Carpinus betulus, Parrotia persica, Acer cappadocicum, Acer velutinum, Alnus glutinosa, Ulmus glabra and Quercus castaneifolia.

The study was conducted in the educational and experimental forest of the University of Tehran (Kheyrud Forest), with a total area of about 7000 ha. The first district, named Patom (Fig. 1), covers an area of 900 ha and was chosen as the study area because of the high occurrence of instabilities compared to other districts.

Elevation ranges from 40 to 930 m above sea level, while the gradient ranges from 0 to 70 degrees. The parent rock is composed of hard calcareous layers with a large number of cracks. According to the Unified Soil Classification System (USCS), the most frequent soil types are CH (clay with high plasticity), CL (clay with low plasticity), and ML (silt with low plasticity). Average annual precipitation at the site is about 1200 mm, with average summer and winter temperatures of 22.5 and 10 °C, respectively. The management system is the »selection system«, which is followed to ensure sustainable management and yields. The future of the forest highly depends on natural regeneration; in some areas, trees are planted in gaps.

Shallow landslides occur in some parts of this forest (Fig. 2) and are more frequent in areas where vegetation has been cleared for the construction of roads.

These slides involve the shallow layers of the slopes,

Croat. j. for. eng. 39(2018)2 257

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

Fig. 1 Location of the study area in the Kheyrud forest, northern Iran (the black point) where vegetation can have a beneficial effect on the stability through the reinforcing action of lateral roots and the action of coarse taproots. In 1994 and 2004, a

series of landslides have occurred that caused the closure of the road network in the Patom district, resulting in costs of about $ 47 000 for the maintenance of a

Fig. 2 A natural shallow landslide; lateral root reinforcement can be seen along the tension crack (a) and failure of forest road cut slope (b) in the study area

258

Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

Ehsan Abdi

damaged segment and the construction of gabion walls in some parts of the road network. Also, gully erosion by concentrated flow from road drainage can be seen along some parts of the road (Fig. 3).

2.2 Sampling

Fig. 3 Gully erosion formed by concentrated flow from a road side ditch Table 1 List of studied plant species Tree species ID

Botanical name

English name

Shrub species ID

Botanical name

English name Hawthorn Medlar

1

Acer velutinum

Persian maple

9

Crataegus microphylla

2

Alnus glutinosa

Black alder

10

Mespilus germanica

3

Carpinus betulus

Common hornbeam

4

Fagus orientalis

Oriental beech

5

Fraxinus excelsior

Common ash

6

Parrotia persica

Persian ironwood

7

Picea abies

Norway spruce

8

Quercus castaneifolia

Chestnutleaved oak

Eight tree and two shrub species were selected for resistance investigations due to their dominance and frequent distribution in the road edge zone (Table 1).

Six specimens were selected randomly from each species to consider intra-species variability, and live root samples were collected randomly from the soil by excavating pits beside the trees at a depth of about 30 cm below the soil surface (Cofie and Koolen 2001, Abdi et al. 2010). To prevent pre-stress effects, roots were cut with sharp scissors and placed in plastic bags.

In most previous studies, the root samples were treated with a 15% alcohol solution (e.g. Bischetti et al.

2005, Bischetti et al. 2016) to preserve them from deterioration prior to the tensile tests. Therefore, their moisture content is higher than that of field-tested roots (Vergani et al. 2016), and the root is slightly swollen, resulting in a higher root diameter (Boldrin et al.

2017) and possible error sources. Hales and Miniat (2016) found that roots with 50% less moisture were more than twice as strong as fresh roots. To overcome this problem and prevent severe changes of root moisture content, we only sprayed the roots with a 15% alcohol solution instead of adding the solution to the root bags. Roots were preserved at 4 °C for a few days to avoid an impact on the measured parameters (Bischetti et al. 2005).

2.3 Resistance tests In the laboratory, roots were carefully inspected for possible damage. Prior to the experiment, root diameter was measured at three different positions along the middle length of the root to obtain a representative value (Bischetti et al. 2005, Vergani et al. 2012, Abdi et al. 2014). Tensile tests were carried out using a Floor Model 4486 computer-controlled Instron Universal Testing Machine (UK), equipped with a 5 kN maximum-capacity reversible load cell.

The root ends were clamped and a strain rate of 10 mm/min (Bischetti et al. 2005, Mattia et al. 2005, Pollen 2007, Abdi et al. 2014) was selected, similar to the approach used in previous studies, to allow comparison. De Baets et al. (2008) reported velocities ranging between 1 and 300 mm/min for rapid landslides.

As most shallow instabilities in the study area are classified as rapid and occur during and after heavy rainfalls, this test speed was considered adequate. Only samples ruptured near the middle of the root between

Croat. j. for. eng. 39(2018)2 259

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

the clamps were considered. A total of 487 root samples were tested.

The tensile force (N) at the point of rupture was taken as the peak load (FMax), and the related tensile resistance (MPa) was calculated by dividing the breaking force by the cross-sectional area of each tested root (mm2), see (Eq. 1):

TR =

FMax

π  2  ×d 4

(1)

selection (Zuur et al. 2007, Xiao et al. 2011, Lai et al.

2013). To compare root tensile force and resistance values between species and to take diameter into consideration as a covariate factor, ANCOVA was used (Abdi et al. 2010, Vergani et al. 2014, Vergani et al. 2016). The Kolmogorov–Smirnov test was used to check the normality of the data before proceeding the ANCOVA; due to the non-normality of the data (force and resistance values), the values were log-transformed to ensure homogeneous residual variance and normality.

Tukey´s test was used to compare mean root tensile force and resistance of different species.

here: TR FMax d

3. Results

tensile resistance maximum force to break the root root diameter.

The diameters of the root samples ranged from 0.29 to 5.90 mm. Thicker roots were difficult to test because of clamping constraints (De Baets et al. 2008).

2.4 Statistical analyses The relationship between root tensile force, F (N), and tensile resistance, TR (MPa), as a function of root diameter d (mm), was interpreted through power regressions. To obtain the power regression coefficients (i.e., a and b), two different methods were used: nonlinear least square and log-transformation methods, using R software. The suitability of the regressions and goodness of fit (efficiency of the model) were evaluated using the Akaike information criterion (AIC) and the adjusted R2 values as statistical criteria for model

Descriptive statistics of tested roots and their corresponding force and resistance values are shown in Table 2. Regarding Table 2, the number of valid tensile tests ranged between 30 and 64, based on the species.

Root diameter ranged from 0.29 to 5.90 mm, and mean root diameter for each species varied between 1.53 and 2.45 mm.

3.1 Tensile Force As shown in Table 2, the variability of force and resistance among and even within a given species was high. The minimum force values ranged between 3.80 and 12.10 N for hardwood tree species, 4.30 N for the only softwood tree species, and between 8.90 N and 15.30 N for the two shrubs. Considering the maximum values, the ranges were 203.80 o 398.10 N for hardwood tree species, 198.70 N for the only softwood tree

Table 2 Descriptive statistics of tested roots including diameter, force, and resistance Species

Diameter, mm

n

Force, N

Resistance, MPa

Mean

SD

Max.

Min.

Mean

SD

Max.

Min.

Mean

SD

Max.

Min.

Acer velutinum

56

1.72

1.26

4.45

0.29

64.15

75.11

291.30

7.11

30.77

25.03

135.87

3.97

Alnus glutinosa

59

1.71

1.25

4.68

0.38

60.25

64.84

251.80

7.20

26.12

16.79

108.43

4.52

Carpinus betulus

32

1.63

0.80

3.17

0.35

95.36

83.23

349.50

8.30

43.31

23.55

124.39

13.65

Fagus orientalis

33

1.69

0.94

4.00

0.52

74.92

61.68

237.40

8.30

30.47

12.34

66.47

12.92

Fraxinus excelsior

50

2.39

1.12

4.71

0.52

54.29

42.62

203.80

3.80

12.74

6.61

30.60

3.32

Parrotia persica

58

1.72

1.15

4.77

0.49

84.59

91.29

398.10

12.10

36.41

24.12

123.76

10.91

Picea abies

47

2.41

1.18

4.85

0.40

66.51

51.97

198.70

4.30

15.75

15.51

108.10

2.41

Quercus castaneifolia

30

1.53

0.93

4.02

0.60

83.59

69.87

249.00

9.90

42.67

21.79

104.47

15.83

Crataegus microphylla

64

2.08

1.08

5.90

0.40

135.86 100.06 441.90

8.90

44.94

29.77

224.35

11.01

Mespilus germanica

58

2.45

1.34

5.00

0.50

143.25 123.94 487.20

15.30

32.69

18.28

89.95

7.74

260

Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

Ehsan Abdi

Fig. 4 Tensile force as a function of root diameter. Nonlinear least squares approximation (dashed line) and log-transformation method (continues line) Croat. j. for. eng. 39(2018)2 261

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

Fig. 5 Root tensile force (mean±SE). Means with different letters are statistically different (p<0.05) species, and between 441.90 N and 487.20 N for the two shrubs. Mean tensile force values for hardwoods, softwood, and shrub species were 54.29–95.36, 66.51, and 135.86–143.25 N, respectively.

The relationship between root tensile force F (N) and root diameter d (mm) through power regressions (nonlinear least square and log-transformation) is presented in Fig. (4).

As shown in Fig. (4), log-transformation regressions generally underestimate the situation. Exceptions were found for F. orientalis, Q. castaneifolia, and

C. microphylla, for which the nonlinear least square regressions were below the curves for the log-transformation. Generally, the main differences occurred at the top of the curves (thicker root diameters).

The regression coefficients (a and b), AIC, and adjusted R2 for tensile force regression models are presented in Table (3).

The ranges of a and b were 13.02<a<45.41 and 1.26<b<1.55 for log-transformation and 14.16<a<61.83 and 1.07<b<1.66 for nonlinear least square power regressions regarding all species (Table 3).

Table 3 Regression coefficients, adjusted R2 and AIC for tensile force of different species Log-transformation

Nonlinear least square

a

b

Adj. R2

AIC

a

b

Adj. R2

AIC

Acer velutinum

25.02

1.26

0.70

575.96

20.71

1.61

0.79

556.85

Alnus glutinosa

22.98

1.41

0.85

547.63

25.48

1.38

0.86

543.35

Carpinus betulus

40.64

1.40

0.60

346.39

42.17

1.49

0.63

344.30

Fagus orientalis

29.77

1.52

0.92

280.37

32.96

1.41

0.93

278.59

Fraxinus excelsior

13.02

1.45

0.62

470.92

14.16

1.45

0.63

468.79

Parrotia persica

33.44

1.34

0.87

573.50

26.95

1.66

0.93

527.34

Picea abies

13.62

1.55

0.64

458.97

19.90

1.32

0.66

455.64

Quercus castaneifolia

38.01

1.50

0.73

302.22

50.77

1.15

0.80

293.58

Crataegus microphylla

45.41

1.33

0.61

711.91

61.83

1.07

0.65

705.95

Mespilus germanica

34.38

1.39

0.65

663.95

43.50

1.28

0.67

661.17

Species

262

Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

Ehsan Abdi

Fig. 6 Root resistance as a function of root diameter. Nonlinear least squares approximation (dashed line) and log-transformation method (continues line) Croat. j. for. eng. 39(2018)2 263

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

Table 4 Regression coefficients, adjusted R2, and AIC for resistance of different species Log-transformation

Nonlinear least square

a

b

Adj. R2

AIC

a

b

Adj. R2

AIC

Acer velutinum

28.59

–0.73

0.66

461.68

28.34

–1.08

0.79

434.86

Alnus glutinosa

26.27

–0.59

0.53

458.25

26.87

–0.82

0.58

450.79

Carpinus betulus

46.45

–0.59

0.37

280.30

50.05

–0.55

0.39

279.16

Fagus orientalis

34.02

–0.48

0.41

244.15

35.47

–0.48

0.42

243.52

Fraxinus excelsior

16.58

–0.54

0.35

310.97

17.56

–0.49

0.38

308.93

Parrotia persica

38.22

–0.65

0.58

484.82

40.63

–0.94

0.67

470.90

Picea abies

17.34

–0.44

0.08

389.31

23.49

–0.62

0.15

385.61

Quercus castaneifolia

43.44

–0.49

0.17

266.45

46.79

–0.43

0.20

265.49

Crataegus microphylla

57.85

–0.66

0.33

591.87

62.44

–0.65

0.35

590.32

Mespilus germanica

43.80

–0.60

0.44

470.42

48.38

–0.64

0.47

467.32

Species

As shown in Table (3), nonlinear least square regression resulted in lower AIC and higher adjusted R2 values in all cases; therefore, lower residuals and better goodness of fit, indicating the advantage of nonlinear least square models and their coefficients for the relationship between force and diameter.

The results of the ANCOVA showed that mean root tensile forces were significantly different among species (F=37.36, p<0.001) with regards to root diameter as covariate factor (F=1453.77, p<0.000). The results of Tukey’s test for mean comparisons are presented in Fig. 5.

The two shrub species, along with C. betulus, P. persica, and Q. castaneifolia, are categorized as the strongest species or as the species with the highest tensile force among the studied species (group A in Fig. 5).

The species A. velutinum and A. glutinosa are intermediate (group B), while P. abies and F. excelsior are the weakest species regarding tensile force (group C in Fig. 5). The exception was F. orientalis, which, although the samples were obtained from a natural regeneration stand, was not among the strongest species regarding tensile force.

3.2 Root resistance As shown in Table 2, minimum root resistance ranged between 3.32 and 15.83 MPa for hardwood tree species, 2.41 MPa for the only softwood tree species, and between 7.74 and 11.01 MPa for the two hardwood shrubs. Considering the maximum values, the ranges were 30.60 to 135.87 MPa for hardwood tree species, 108.10 MPa for the only softwood tree species,

264

and 89.95 to 224.35 MPa for the two hardwood shrubs. Mean tensile resistance values for hardwoods, softwood, and shrubs were 12.74–43.31, 15.75, and 32.69–44.94 MPa, respectively.

The relationship between root resistance (MPa) and root diameter d (mm) through power regression (nonlinear least square and log-transformation) is presented in Fig. (6).

As shown in Fig. (6), log-transformation regressions generally underestimate the situation, especially in smaller root sizes. The exceptions were A. velutinum, A. glutinosa, and P. persica, where nonlinear least square regressions are below the log-transformation curves in roots greater than 1 mm in diameter.

The regression coefficients (a and b), AIC, and adjusted R2 for root resistance regression models are presented in Table (4).

The ranges of a and b in resistance were 16.58<a<57.85 and –0.66<b<–0.44 for log-transformation and 17.56<a<62.44 and –1.08<b<–0.43 for nonlinear least square power regression.

As shown in Table (4), nonlinear least square regression resulted in lower AIC and higher adjusted R2 values for all species and, therefore, lower residuals and better goodness of fit, indicating the advantage of nonlinear least square models and their coefficients for the relationship between resistance and diameter.

The results of the ANCOVA revealed that mean root tensile resistance was significantly different among the tested species (F=34.87, p<0.001) with regards to root diameter as covariate (F=274.26, p<0.000).

The results of Tukey’s test for mean comparisons are presented in Fig. (7).

Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255â&#x20AC;&#x201C;270)

Ehsan Abdi

Fig. 7 Root resistance (meanÂąSE). Means with different letters are statistically different (p<0.05) Regarding Fig. (7), the species can be classified in three groups (A, B, and C) based on root tensile resistance. The weakest species (classified as group C) was represented by F. excelsior in the 40-year-old plantation. Intermediate species (group B) were A. velutinum, A. glutinosa, and P. abies in plantation stands. Two shrub species as well as C. betulus, F. orientalis, P. persica, and Q. castaneifolia were the strongest species regarding root resistance (group A).

4. Discussion Root tensile force and resistance are important factors that influence soil reinforcement and tree anchorage and provide essential data about using live materials for bioengineering techniques (Bischetti et al. 2010).

Similar to several previous studies (e.g., Bischetti et al.

2005, Mattia et al. 2005, Schwarz et al. 2013, Vergani et al. 2014), we found a large variability in root tensile force and resistance based on species and root diameter. According to previous studies (e.g., Nilaweera and Nutalaya 1999, Bischetti et al. 2005, Abdi et al. 2010, Vergani et al. 2014), this relationship is well described, in terms of both force and resistance, by positive and negative power law regressions, respectively, confirming the strong dependence of root strength on root size.

Genet et al. (2005) justified this relationship by different cellulose to lignin ratios, with smaller roots having higher ratios. Also Ye et al. (2017) attributed this relationship to the chemical composition of root tissues

and showed that tensile force was significantly negatively correlated with cellulose and holocellulose and significantly positively correlated with lignin and the lignin to cellulose ratio, while for tensile resistance, opposite correlations have been reported.

In the current study, two regression methods (nonlinear least square and log-transformation) were used to obtain the coefficients of the power regressions (i.e., a and b). Based on the results, the power equation parameters were different in the two regression types (Tables 3 and 4). Regarding the AIC values and the adjusted R2 values for model selection, the nonlinear least square method resulted in lower AIC and higher adjusted R2 values, making it the preferred model for both force and resistance power regressions. This is consistent with Changyong et al. (2014), who reported that log-transformation may inaccurately estimate model parameters. Zuur et al. (2007) stated that the model with the lowest AIC and the highest adjusted R2 values can be selected as the optimal model, which indicates the improved fit of the model to the data and, therefore, a lower residual sum of square (lack of fit).

This may be due to the basis of the nonlinear least square method, which approximates the model first and then refines the parameters by successive iterations (Hesse 2006). This is consistent with the results of Schwarz et al. (2013) and Giadrossich et al. (2017), who indicated that different algorithms lead to different coefficients of the equation, although they did not report the optimal model. Previous studies (e.g.,

Croat. j. for. eng. 39(2018)2 265

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

Vergani et al. 2014, Giadrossich et al. 2017) reported that small changes in the fitting curve of the root diameter–force or resistance relationship lead to changes in the output of root reinforcement models as an important factor in efficiency assessment of bioengineering measures (Bischetti et al. 2010). It has, therefore, been suggested that root resistance and force are critical factors in slope stability evaluations. Giadrossich et al. (2016) reported that using the log-transformation method results in an underestimation of force. Our results agree with the findings of Giadrossich et al.

(2016) and showed that using the log-transformation method leads to an underestimation of both tensile force and resistance. Therefore, using values of the equation parameters of log-transformation will underestimate the effect of root reinforcement in stability analyses.

From a statistical point of view, Packard et al.

(2011) point out that log-transformed models predict the geometric mean for the response variable, and that log-transformation inherently distorts the relationship between variables. The authors, therefore, recommend that analyses should be performed on the arithmetic scale via nonlinear regression, and this was reported as the advantage of nonlinear regression. Also, Packard (2012) showed that log-transformation of data created new distributions that actually obscured the relationships between predictor and response variables and led to bias. The author concluded that log-transformation is not a generally reliable way to estimate parameters in a simple power function on the original scale.

The other advantage of nonlinear regression is that the use of nonlinear model fitting is facilitated by the availability of easy-to-use advanced statistical packages (Lai et al. 2013), which were not available at the time when log-transformation power regression appeared (Packard 2012).

Regarding the advantages of nonlinear regression and as power model coefficients are important factors in root reinforcement assessments in soil bioengineering measures, it is suggested that log-transformation models be replaced by nonlinear least square models to obtain more realistic estimates based on observations (tested root samples).

Regarding tensile force coefficients, Vergani et al.

(2012) obtained the following a and b ranges for seven common European tree species: 8.31<a<19.66 and 1.49<b<1.85. In the current study, a and b coefficients for the force-diameter relationship are generally out of the range for European tree species in both log-transformation and nonlinear least square methods. In this study, the ranges for eight common tree species of Hyrcanian species were 13.2<a<40.64 and 1.26<b<1.55 for the log-transformation method and 14.16<a<50.77 and 1.15<b<1.66 for the nonlinear least square method. The

266

difference between the coefficients may be a result of different species and varying environmental conditions (Vergani et al. 2012, Boldrin et al. 2017).

Regarding root resistance coefficients, Nilaweera (1994) suggested the following a and b ranges for hardwood tree species roots: 29.1<a<87.0 and –0.8<b<–0.4. In this study, the ranges for eight common tree species of the Hyrcanian forest were 16.58<a<46.45 and –0.73<b<–0.44 for the log-transformation method and 17.56<a<50.05 and –1.08<b<–0.43 for the nonlinear least square method. In the current study, the a and b coefficients were generally in the range obtained for the log-transformation method (except a for F. excelsior and P. abies). However, for F. excelsior and P. abies, the a values were in the range obtained from the nonlinear least square method, while for A. velutinum, A. glutinosa, and P. persica, the b values were outside of this range (below –0.8). As ranges reported in Nilaweera (1994) are based on forests in Thailand, they may need to be reconsidered and modified based on more recent studies in different forest zones.

In the current study, the measured mean tensile forces for the Hyrcanian forest species (A. velutinum 64.15 N, A. glutinosa 60.25 N, C. betulus 93.36, F. orientalis 74.92 N, F. excelsior 54.29 N, P. abies 66.51 N) were similar to those obtained by Vergani et al. (2016) for some European species (Acer pseudoplatanus 65 N, Ostrya carpinifolia 56 N, Fagus sylvatica 84 N, Fraxinus excelsior 47 N, Picea abies 46 N). However, they were lower than those reported by Chiaradia et al. (2016) for Fagus sylvatica (122.46 N) and Picea abies (70.68 N). The differences between the values presented in this study and those in the literature may be explained by the different responses of plants to different environmental conditions (plasticity) to minimize abiotic and biotic stresses (Boldrin et al. 2017).

The measured mean tensile resistance values in the current study (A. velutinum 30.77 MPa, A. glutinosa 26.12 MPa, C. betulus 43.31 MPa, F. orientalis 30.47 MPa, F. excelsior 12.74 MPa, P. persica 36.41 MPa, P. abies 15.75 MPa and Q. castanefolia 42.67 MPa) are comparable to those reported in Stokes (2002), including Alnus incana (22 MPa), Fraxinus excelsior (26 MPa), Acer platanoides (27 MPa), Picea abies (28 MPa), Quercus rubra (32 MPa) and Alnus japonica (41 MPa). However, our values were larger than the mean resistance values reported by Boldrin et al. (2017) for re-established small trees (7.1–23.2 MPa). This may be explained by the report of Genet et al. (2006), who showed that tensile resistance was lower in the early growth stage and increased in older plants.

Comparisons of force–diameter relationships for different species (ANCOVA, Fig. 5) confirmed that Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

there are statistically significant differences in root tensile force between species. In this regard, F. excelsior and P. abies can be considered as the weakest species, while C. microphylla and M. germanica (shrubs) are among the strongest ones. Broadleaved species generally have a higher tensile force than conifer species (except F. excelsior species), which is consistent with the results of Vergani et al. (2012). Stokes and Mattheck (1996) justified this with the different root anatomy of broadleaves and conifers, as broadleaves generally have larger cells and thinner cell walls.

We found statistically significant differences in root tensile resistance between species, with mean values ranging between 12.74 and 44.49 MPa. Several authors attributed these differences to genetic and environmental factors and the root system tissue composition (Genet et al. 2005, De Baets et al. 2008, Chiaradia et al.

2016, Ye et al. 2017).

In this regard, F. excelsior was the weakest species, while C. microphylla and M. germanica (shrubs) were among the strongest ones. This is in contrast with Morgan and Rickson (1995), who stated that the range of root resistance of shrub species is not significantly different from that of trees, although our results showed that they may even have higher values than some tree species. This is in agreement with Burylo et al. (2011) and Boldrin et al. (2017), who revealed that the roots of shrubs were more resistant than those of tree species. The mean resistance values of the two shrub species in our study (i.e., 44.94 and 32.69 MPa) are comparable to the results of Mattia et al. (2005) (Pistacia lentiscus 55.0, Atriplex halimus 57.2 MPa), and are higher than those found by Tosi (2007) (Rosa canina 22.95, Inula viscosa 18.72, and Spartium junceum 29.93 MPa) and Comino and Marengo (2010) (Rosa canina 42.9 Cotoneaster dammeri 18.7 and Juniperus horizontalis 14.8 MPa). Concerning the results of Boldrin et al.

(2017), Crataegus monogyna had the highest tensile resistance (23.2 MPa) among the 10 shrubs and small trees of Europe, but in our study, the resistance of Crataegus microphylla (44.94 MPa) was about twice as high as that reported by Boldrin et al. (2017). This may partly be explained by the different environments and the fact that the species assessed by Boldrin et al. (2017) were in the early stage of establishment. Watson et al.

(1999) reported a direct relationship between plant growth and increased reinforcement effect; similarly, Genet et al. (2006) reported increasing tensile resistance with plant growth and attributed this phenomenon to higher quantities of cellulose in older plants.

Previous studies have reported ranges for coefficients of root resistance power regressions in shrub species (e.g., Mattia et al. 2005, De Baets et al. 2008, Comino and Marengo 2010, Burylo et al. 2011). They

Ehsan Abdi

reported the following ranges: 73.0<a<91.2 and –0.60<b<–0.45 for two shrub species (Atriplex halimus and Pistacia lentiscus) in Mattia et al. (2005); 4.41<a<45.59 and –1.77<b<–0.45 for 14 Mediterranean shrub species in De Baets et al. (2008); 14.79<a<37.77 and −1.28<b<−0.83 for three species (Rosa canina, Cotoneaster dammeri and Juniperus horizontalis) in Comino and Marengo (2010); and 4.4<a<91.2 and −1.75<b<−0.52 for two shrubby species (Genista cinerea and Thymus serpyllum) in Burylo et al. (2011). The results of the current study are consistent with those of Burylo et al.

(2011), but do not fall into the range of De Baets et al.

(2008) for a coefficients and De Baets et al. (2008) and Comino and Marengo (2010) for both a and b coefficients. Again, the difference between coefficients may be the result of different species and environmental conditions (Vergani et al. 2012, Boldrin et al. 2017).

The weakest species, based on resistance, were species in 40-year-old plantations (Fig. 7). This is consistent with the results of Watson and Marden (2004), who reported lower resistance values for plantations of radiate pine (17 MPa) and Douglas fir (25 MPa) compared to 11 New Zealand indigenous riparian plant species. The mean resistance values of plantation species in our study (ranging from 54–66 MPa) were significantly higher than those reported by Watson and Marden (2004) for radiate pine (17 MPa) and Douglas fir (25 MPa) in plantations and by Genet et al. (2008) for three different age stages of Cryptomeria japonica plantations, i.e., 22.6, 25.3, and 31.7 MPa for the juvenile, intermediate, and mature stands, respectively.

Considering that most landslides in the study area are rainfall-induced, hydrological effects of vegetation might not significantly affect soil stability in seasons with heavy rainfall (autumn and winter). In these seasons, the mechanical effects of vegetation or root reinforcement can play an important role in soil stabilization. Roots can mobilize their tensile resistance during failures along tension cracks (Vergani et al. 2017) and at the lateral surface of the landslide (Fig. 2a) and increase the resisting force against the driving force, thereby improving the stability of the slope. For this reason, species with higher root resistance (group A in Fig. 7) are preferred (Stokes 2002) in soil bioengineering systems (Bischetti et al. 2010), and the magnitude of the root resistance can influence the performance of these species. The results of our study should, therefore, be considered by forest managers when selecting suitable species for the reestablishment of vegetation on cut and fill slopes after road construction. In addition to tensile resistance, surcharge is an important criterion for soil bioengineering measures, especially on forest road cut and fill slopes that usually have higher slope angles than natural slopes. In this study,

Croat. j. for. eng. 39(2018)2 267

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

shrubs showed high root resistance, and as the negative effect of surcharge on slope stability (Chiaradia et al. 2016) is negligible for them compared to tree species (Morgan and Rickson 1995), they can be good choices for forest road stabilization. This may represent an advantage of shrubs compared to trees in soil bioengineering applications, especially on forest road cut and fill slopes.

5. Conclusions We investigated the root mechanical behavior of eight common tree and two shrub species of the Hyrcanian forest. Two algorithms (nonlinear least square and log-transformation) were used to estimate the coefficients of the power regressions for force-diameter and resistance-diameter.

Root mechanical behavior is dependent on root diameter and can be well described by power law relationships as a function of the root diameter for both force and resistance functions. Our results showed that a and b not only depend on the species, but also on the statistical method applied. The nonlinear least square method was selected as the optimal model and can better explain the relationship between diameter-force and diameter-resistance. Also, using the log-transformation model underestimates power regressions of root force and resistance and therefore will underestimate the root reinforcement magnitude. Root tensile force and resistance differed significantly among species (ANCOVA) and were grouped into three strength classes; in terms of both force and resistance, shrubs constituted the strongest class. Furthermore, this study showed that trees in plantation stands had a lower resistance than trees in natural stands. Root tensile force and resistance are important inputs of root reinforcement models to estimate the quantity of increased soil cohesion and calculate slope stability, considering the presence of plant roots. These data may be used for the reestablishment of vegetation in cut and fill slopes, with the aim to reduce the risk of instabilities.

Further studies on the use of plants in bioengineering strategies should consider additional factors that might influence root mechanical behavior, such as site type, soil type, plant age, and elevation.

Acknowledgements The author would like to acknowledge the financial support of the »Iran National Science Foundation (INSF)« under the project number 93022486. Also I would like to thank the two anonymous reviewers for their detailed comments and suggestions that led to an improvement of the work.

268

6. References Abdi, E., Majnounian, B., Rahimi, H., Zobeiri, M., 2009: Distribution and tensile strength of Hornbeam (Carpinus betulus) roots growing on slopes of Caspian Forests, Iran. Journal of Forestry Research 20(2): 105–110.

Abdi, E., Majnounian, B., Genet, M., Rahimi, H., 2010: Quantifying the effects of root reinforcement of Persian Ironwood (Parrotia persica) on slope stability; a case study: Hillslope of Hyrcanian forests, northern Iran. Ecological Engineering 36(10): 1409–1416.

Abdi, E., Azhdari, F., Abdulkhani, A., Soofi, H., 2014: Tensile strength and cellulose content of Persian ironwood (Parrotia persica) roots as bioengineering material. Journal of Forest Science 60(10): 425–430.

Abedi, T., Hoseini, S.A., Naghdi, R., 2010: Study of relationship between soil mechanical characteristic and landslide in forest road route. Journal of Watershed Management Research 1(1): 17–29.

Abernethy, B., Rutherfurd, I.D., 2000: The effect of riparian tree roots on the mass-stability of riverbanks. Earth Surface Processes and Landforms 25(9): 921–937.

Abernethy, B., Rutherfurd, I.D., 2001: The distribution and strength of riparian tree roots in relation to riverbank reinforcement. Hydrological Processes 15(1): 63–79.

Bischetti, G.B., Chiaradia, E.A., Simonato, T., Speziali, B., Vitali, B., Vullo, P., Zocco, A., 2005: Root strength and root area ratio of forest species in Lombardy (Northern Italy). Plant and Soil 278 (1–2): 11–22.

Bischetti, G.B., Chiaradia, E.A., Epis, T., Morlotti, E., 2009: Root cohesion of forest species in the Italian Alps. Plant and Soil 324 (1–2): 71–89.

Bischetti, G.B., Chiaradia, E.A., D’Agostino, V., Simonato, T., 2010: Quantifying the effect of brush layering on slope stability. Ecological Engineering 36(3): 258–264.

Bischetti, G.B., Bassanelli, C., Chiaradia, E.A., Minotta, G., Vergani, C., 2016: The effect of gap openings on soil reinforcement in two conifer stands in northern Italy. Forest Ecology and Management 359: 286–299.

Boldrin, D., Leung, A.K., Bengough, A.G., 2017: Root biomechanical properties during establishment of woody perennials.

Ecological Engineering 109: 196–206.

Burylo, M., Hudek, C., Rey, F., 2011: Soil reinforcement by the roots of six dominant species on eroded mountainous marly slopes (Southern Alps, France). Catena 84 (1): 70–78.

Capilleri, P.P., Motta, E., Raciti, E., 2016: Experimental study on native plant root tensile strength for slope stabilization.

Procedia Engineering 158: 116–121.

Changyong, F.E.N.G., Hongyue, W.A.N.G., Naiji, L.U., Tian, C.H.E.N., Hua, H.E., Ying, L.U., 2014: Log-transformation and its implications for data analysis. Shanghai Archives of Psychiatry 26(2): 105–109.

Chiaradia, E.A., Vergani, C., Bischetti, G.B., 2016: Evaluation of the effects of three European forest types on slope stability by field and probabilistic analyses and their implications for

Croat. j. for. eng. 39(2018)2

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270) forest management. Forest Ecology and Management 370: 114–129.

Cislaghi, A., Bordoni, M., Meisina, C., Bischetti, G.B., 2017: Soil reinforcement provided by the root system of grapevines: Quantification and spatial variability. Ecological Engineering 109: 169–185.

Cofie, P., Koolen, A.J., 2001: Test speed and other factors affecting the measurements of tree root properties used in soil reinforcement models. Soil and Tillage Research 63(1): 51–56.

Comino, E., Marengo, P., 2010: Root tensile strength of three shrub species: Rosa canina, Cotoneaster dammeri and Juniperus horizontalis: soil reinforcement estimation by laboratory tests.

Catena 82(3): 227–235.

Coppin, N.J., Richards, I.G., 1990: Use of vegetation in Civil Engineering. Butterworth publication London 272 p.

De Baets, S., Poesen, J., Reubens, B., Wemans, K., De Baerdemaeker, J., Muys, B., 2008: Root tensile strength and root distribution of typical Mediterranean plant species and their contribution to soil shear strength. Plant and Soil 305(1–2): 207–226.

De Baets, S., Poesen, J., Reubens, B., Muys, B., Baerdemaeker, J. De, Meersmans, J., 2009. Methodological framework to select plant species for controlling rill and gully erosion: application to a Mediterranean ecosystem. Earth Surface Processes and Landforms 34(10): 1374–1392.

Ebrahimi, M., Habibolahian, M., Amir-Ahmadi, A., Zangeneh Asadi, M., Nezhadsoleimani, H., 2015: Considering the road building effect on the occurrence of surface landslides by using the slope stability model. Geographical Planning of Space 5(15): 149–162.

Fattorini, S., 2007: To fit or not to fit? A poorly fitting procedure produces inconsistent results when the species-area relationship is used to locate hotspots. Biodiversity and Conservation 16(9): 2531–2538.

Genet, M., Stokes, A., Salin, F., Mickovski, S.B., Fourcaud, T., Dumail, J.F., Van Beek, R. 2005: The influence of cellulose content on tensile strength in tree roots. Plant and soil 278(1–2): 1–9.

Genet, M., Stokes, A., Fourcaud, T., Hu, X., Lu, Y., 2006: Soil fixation by tree roots: changes in root reinforcement parameters with age in Cryptomeria japonica D. Don. plantations. In Interpraevent: 535–542.

Genet, M., Kokutse, N., Stokes, A., Fourcaud, T., Cai, X., Ji, J., Mickovski, S., 2008: Root reinforcement in plantations of Cryptomeria japonica D. Don: effect of tree age and stand structure on slope stability. Forest Ecology and Management 256(8): 1517–1526.

Giadrossich, F., Cohen, D., Schwarz, M., Seddaiu, G., Contran, N., Lubino, M., Valdés-Rodrŕguez, O.A., Niedda, M., 2016: Modeling bio-engineering traits of Jatropha curcas L.. Ecological Engineering 89: 40–48.

Giadrossich, F., Schwarz, M., Cohen, D., Cislaghi, A., Vergani, C., Hubble, T., Phillips, C., Stokes, A., 2017: Methods to measure the mechanical behaviour of tree roots: A review. Ecological Engineering 109: 256–271.

Ehsan Abdi

Goodman, A.M., Ennos, A.R., 1999: The effects of soil bulk density on the morphology and anchorage mechanics of the root systems of sunflower and maize. Annals of Botany 83(3): 293–302.

Hales, T.C., Miniat, C.F., 2016: Soil moisture causes dynamic adjustments to root re-inforcement that reduce slope stability.

Earth Surface Processes and Landforms 42(5): 803–813.

Hayati, E., Abdi, E., Mohseni Saravi, M., Nieber, J., Majnounian, B., Chirico, G., 2017: Temporal pattern of soil matric suction in the unsaturated soil slope under different forest cover.

In EGU General Assembly Conference Abstracts 19: 533 p.

Hesse, R., 2006: Incorrect nonlinear trend curves in Excel. Foresight: The International Journal of Applied Forecasting 3: 39– 43.

Hubble, T.C.T., Airey, D.W., Sealey, H.K., De Carli, E.V., Clarke, S.L., 2013: A little cohesion goes a long way: estimating appropriate values of additional root cohesion for evaluating slope stability in the Eastern Australian highlands. Ecological Engineering 61: 621–632.

Lai, J., Yang, B., Lin, D., Kerkhoff, A.J., Ma, K., 2013: The allometry of coarse root biomass: log-transformed linear regression or nonlinear regression? PloS one 8(10): 1–8.

Mattia, C., Bischetti, G.B., Gentile, F., 2005: Biotechnical characteristics of root systems of typical Mediterranean species.

Plant and soil 278(1–2): 23–32.

Moos, C., Bebi, P., Graf, F., Mattli, J., Rickli, C., Schwarz, M., 2016: How does forest structure affect root reinforcement and susceptibility to shallow landslides? Earth Surface Processes and Landforms 41(7): 951–960.

Morgan, R.P., Rickson, R.J., 1995: Slope stabilization and erosion control: a bioengineering approach. Second edition.

Chapman and Hall, 293 p.

Nilaweera, N.S., 1994: Effects of tree roots on slope stability – The case of Khao Lunag Mountain Area, Southern Thailand.

Doctor Thesis No. GT-93-2, 452 p.

Nilaweera, N.S., Nutalaya, P., 1999: Role of tree roots in slope stabilisation. Bulletin of Engineering Geology and the Environment 57(4): 337–342.

Packard, G.C., 2009: On the use of logarithmic transformations in allometric analyses. Journal of Theoretical Biology 257(3): 515–518.

Packard, G.C., 2011: Rotational distortion in conventional allometric analyses. Comparative Biochemistry and Physiology Part A: Molecular & Integrative Physiology 159(4): 392–400.

Packard, G.C., 2012: Is non-loglinear allometry a statistical artifact? Biological Journal of the Linnean Society 107(4): 764– 773.

Packard, G.C., Birchard, G.F., 2008: Traditional allometric analysis fails to provide a valid predictive model for mammalian metabolic rates. Journal of Experimental Biology 211(22): 3581–3587.

Packard, G.C., Birchard, G.F., Boardman, T.J., 2011: Fitting statistical models in bivariate allometry. Biological Reviews 86(3): 549–563.

Croat. j. for. eng. 39(2018)2 269

Ehsan Abdi

Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran (255–270)

Pollen, N., 2007: Temporal and spatial variability in root reinforcement of streambanks: accounting for soil shear strength and moisture. Catena 69(3): 197–205.

Pollen, N., Simon, A., 2005: Estimating the mechanical effects of riparian vegetation on streambank stability using a fiber bundle model. Water Resources Research 41(7): 1–11.

Roering, J.J., Schmidt, K.M., Stock, J.D., Dietrich, W.E., Montgomery, D.R., 2003: Shallow landsliding, root reinforcement, and the spatial distribution of trees in the Oregon Coast Range.

Canadian Geotechnical Journal 40(2): 237–253.

Savadkoohi, N.M, Hosseini, S.A., 2013: Effect of Physical and Mechanical Soil Properties on Landslide in Forest Road Edge.

Journal of Watershed Management Research 4(8): 28–42.

Schiechtl, H.M., 1980: Bioengineering for land reclamation and conservation. University of Alberta Press, 404 p.

Schmidt, K.M., Roering, J.J., Stock, J.D., Dietrich, W.E., Montgomery, D.R., Schaub, T., 2001: The variability of root cohesion as an influence on shallow landslide susceptibility in the Oregon Coast Range. Canadian Geotechnical Journal 38(5): 995– 1024.

Schwarz, M., Lehmann, P., Or, D., 2010a: Quantifying lateral root reinforcement in steep slopes from a bundle of roots to tree stands. Earth Surface Processes and Landforms 35(3): 354–367.

Schwarz, M., Preti, F., Giadrossich, F., Lehmann, P., Or, D., 2010b: Quantifying the role of vegetation in slope stability: A case study in Tuscany (Italy). Ecological Engineering 36(3): 285–291.

Schwarz, M., Giadrossich, F., Cohen, D., 2013: Modeling root reinforcement using a root-failure Weibull survival function.

Hydrology and Earth System Sciences 17(11): 4367–4377.

Stokes, A., 2002: Biomechanics of tree root anchorage. In Plant roots: The hidden half, Editors: Waisel, Y., Eshel, A., Kafkafi, U.: 175–186.

Stokes, A., Mattheck, C., 1996: Variation of wood strength in tree roots. Journal of Experimental Botany 47(5): 693–699.

Stokes, A., Douglas, G.B., Fourcaud, T., Giadrossich, F., Gillies, C., Hubble, T., Kim, J.H., Loades, K.W., Mao, Z., McIvor, I.R., Mickovski, S.B., Mitchell, S., Osman, N., Phillips, C., Poesen, J., Polster, D., Preti, F., Raymond, P., Rey, F., Schwarz, M., Walker, L.R., 2014: Ecological mitigation of hillslope instabil-

ity: Ten key issues facing researchers and practitioners. Plant and Soil 377(1–2): 1–23.

Tosi, M., 2007: Root tensile strength relationships and their slope stability implications of three shrub species in the Northern Apennines (Italy). Geomorphology 87(4): 268–283.

Vergani, C., Chiaradia, E.A., Bischetti, G.B., 2012: Variability in the tensile resistance of roots in Alpine forest tree species. Ecological Engineering 46: 43–56.

Vergani, C., Schwarz, M., Cohen, D., Thormann, J.J., Bischetti, G.B., 2014: Effects of root tensile force and diameter distribution variability on root reinforcement in the Swiss and Italian Alps. Canadian Journal of Forest Research 44(11): 1426–1440.

Vergani, C., Schwarz, M., Soldati, M., Corda, A., Giadrossich, F., Chiaradia, E.A., Morando, P., Bassanelli, C., 2016: Root reinforcement dynamics in subalpine spruce forests following timber harvest: A case study in Canton Schwyz, Switzerland.

Catena 143: 275–288.

Vergani, C., Giadrossich, F., Schwarz, M., Buckley, P., Conedera, M., Pividori, M., Salbitano, F., Rauch, H.S., Lovreglio, R., Schwarz, M., 2017: Root reinforcement dynamics of European coppice woodlands and their effect on shallow landslides: A review. Earth-Science Reviews 167: 88–102.

Watson, A., Phillips, C., Marden, M., 1999: Root strength, growth, and rates of decay: root reinforcement changes of two tree species and their contribution to slope stability. Plant and Soil 217(1–2): 39–47.

Watson, A.J., Mardern, M., 2004: Live root-wood tensile strengths of some common New Zealand indigenous and plantation tree species. New Zealand Journal of Forestry Science 34(3): 344–353.

Wu, T.H., McKinnell III, W.P., Swanston, D.N., 1979: Strength of tree roots and landslides on Prince of Wales Island, Alaska.

Canadian Geotechnical Journal 16(1): 19–33.

Xiao, X., White, E.P., Hooten, M.B., Durham, S.L., 2011: On the use of log transformation vs. nonlinear regression for analyzing biological power laws. Ecology 92(10): 1887–1894.

Ye, C., Guo, Z., Li, Z., Cai, C., 2017: The effect of Bahiagrass roots on soil erosion resistance of Aquults in subtropical China. Geomorphology 285: 82–93.

Zuur, A., Ieno, E.N., Smith, G.M., 2007: Analyzing ecological data. Springer Science&Business Media, 672 p.

Authors’ address:

Received: November 02, 2016 Accepted: February 06, 2018

270

Assoc. prof. Ehsan Abdi, PhD. * e-mail: abdie@ut.ac.ir University of Tehran Department of Forestry P.O.BOX 31585-4314, Karaj IRAN * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression and Artificial Neural Networks Mohammad Javad Heidari, Akbar Najafi, Seyedjalil Alavi Abstract The accurate prediction of forest road pavement performance is important for efficient management of surface transportation infrastructure and achieves significant savings through timely intervention and accurate planning. The aim of this paper was to introduce a methodology for developing accurate pavement deterioration models to be used primarily for the management of the forest road infrastructure. For this purpose, 19 explanatory and three corresponding response variables were measured in 185 segments of 50 km forest roads. Logistic regression (LR) and artificial neural networks (ANNs) were used to predict forest road pavement deterioration, Pothole, rutting and protrusion, as a function of pavement condition, environmental factors, traffic and road qualify. The results showed ANNs and LR models could classify from 82% to 89% of the current pavement condition correctly. According to the results, LR model and ANNs predicted rutting, pothole and protrusion with 83.5%, 83.00% and 81.75%, 88.65% and 85.20%, 80.00% accuracy. Equivalent single axle load (ESAL), date of repair, thickness of pavement and slope were identified as most significant explanatory variables. Receiver Operating Characteristic Curve (ROC) showed that the results obtained by ANNs and logistic regression are close to each other.

Keywords: forest road maintenance, pavement management system, pavement strength, pothole, protrusion, rutting

1. Introduction The existence of qualitatively and quantitatively optimal forest transportation systems, which can be divided into primary and secondary network, is one of the basic requirements in todayâ&#x20AC;&#x2122;s modern, integrated, technologically advanced, rational, cost-effective, environmentally sound, socially responsible, biodiversity respectful and income sustainable management of forest ecosystems (PotoÄ?nik et al. 2015).

Similar to public roads, forest roads are deteriorated because of excessive load, transportation on negative weather condition, inconvenient drainage construction, planning the forest roads on low bearing capacity soils, using of unsuitable techniques for forest road construction (Eroglu et al. 2003). While it is important to do the right repair at the right place at the

right time, it is cheaper to maintain roads in good shape than to fix broken roads. An excellent pavement maintenance program is usually part of an overall management plan. It can also be used as the starting point to develop such a plan (Ouma et al. 2015). One of the most important keys to successful pavement maintenance is to know what the proper repair is. This can range from doing nothing to reconstruct the entire road. It may be better to do nothing rather than to make a repair that fails prematurely (Santos and Ferreira 2013). Therefore, there must be a detailed plan for forest roads to keep their efficiency and reduce environmental damage and costs.

Pavement management is a program for improving the quality and performance of pavements and minimizing costs through good management practice (Bent et al. 2012, Roberts and Attoh-Okine 1998). For-

Croat. j. for. eng. 39(2018)2 271

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271â&#x20AC;&#x201C;287)

est Roads Pavement Management System (FRPMS) is a set of defined procedures for collecting, analyzing, maintaining, and reporting forest road pavement data, to assist the decision makers in finding optimum strategies for maintaining forest road pavements in serviceable condition over a given period of time for the least cost (Rusu et al. 2015, Sen 2013). It is, moreover, designed to provide objective information and useful data for analysis so that road managers can make more consistent, cost-effective, and defensible decisions related to the preservation of a pavement network (Santos and Ferreira 2013, Yang 2004). The first step for a successful introduction of asset management systems is to develop a reliable deterioration model considering the heterogeneous deterioration process of their road network. Although most forest road pavement experts or researchers already understand the importance of this, the task is never easy due to insufficient data for statistical methods that usually demand a large amount of inspection data to draw characteristics of the deterioration process of their road network (Han et al. 2014).

The impact of various factors on pavement performance is complex. To understand the mechanism and predict the future state of pavement, it is essential to study the factors affecting pavement deterioration (Moreno-Navarro et al. 2015, Schlotjes 2013). Factors affecting pavement condition can be various factors such as the age of the pavement, traffic, environment, materials, thickness of pavement, pavement strength and properties of the substrate that affect the mechanical properties of the pavement (Salour and Erlingsson 2013). The effectiveness of maintenance planning depends on the accuracy of the predicted future performance and observed current condition of the pavement. If the deterioration models used in determining the maintenance policies cannot sufficiently represent the actual deterioration process, the planned maintenance strategies might be far from optimal. Therefore, performance measurement and deterioration models are essential components of the maintenance planning (Lin et al. 2014). Pavement deterioration models actually predict the future of pavement and it is useful for developing models of pavement maintenance management or maintenance priority index (MPI) (Saha et al. 2014). Pavement condition performance models, which simulate the deterioration process of pavement condition, play a pivotal role in FRPMS (Owolabi and Oladapo 2011). The ranking criteria used to prioritize pavement maintenance program are based on the severity of the stress and conditioned by it. The conditions governing the forest roads is different from the main roads maintenance management, and it is more

272

complex (Sundin and Braban-Ledoux 2001). On the other hand, there is no special equipment to check the condition of roads or it is very expensive. For this reason it is recommended to use the techniques of linear and nonlinear models, as they are cheaper and faster (Hahne et al. 2014). Myriads of researches have been done with respect to pavement performance modeling in forest and public roads (Faghih-Imani and Amador Jimenez 2013, Forsyth et al. 2006, Tabatabaee et al.

2013, Tunay 2006). Regression technique is used by researchers as a traditional method to predict pavement deterioration rate (Kaur and Pulugurta 2008).

Logistic regression is used when the target variable is binary or binomial and the independent variables are numerical and (or) categorical (Xu et al. 2014). Most specialists agree that no single prediction model is applicable to all pavements. This is due to the high variability in the manner in which each agency measures its pavement. For example, they may vary in the number, scale, type of pavement characteristics, and in the pavement deterioration indicators used (Roberts and Attoh-Okine 1998). In recent years, predicting the expected pavement deterioration has been the focus of many works (Attoh-Okine 1994) using traffic and time-related models, interactive time, traffic, or distress models. To date, approaches used in forecasting the pavement condition have included: regression models, artificial neural network, empirical model, mechanistic models and deterministic and probabilistic models in public roads. Within these approaches, logistic regression analysis and ANN are used by researchers as a new method to predict pavement deterioration rate in forest roads. Logistic regression is a data mining method that can be used to classify a given dataset. Logistic regression builds a linear model based on a transformed variable (Friedman et al.

2000) often referred to as logit variable (Hosmer Jr. et al. 2013), which is used to assess the relation between one dependent variable (binary, categorical or ordinal) and several predictor variables (continuous or categorical). Among the various methods of regression, according to the nature of data, logistic regression is a good method for pavement modeling and prediction for forest roads (Hosmer et al. 2013).

However, the pavement deterioration process is so complex that it is difficult and sometimes impossible to find an appropriate functional form, as used by traditional modeling (Lee et al. 2013). Hence a new approach, which can be categorized as Âťbiologically-inspiredÂŤ, is taking the territory from its traditional counterpart. A typical model in this category is Artificial Neural Networks (ANNs) (Yang et al. 2003). Neural network abstracts the underlying relationship beCroat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

tween dependent and independent variables from the exemplar data pairs and expresses it as forms of weight matrix (Russell C. Eberhart 1990, Yang 2004, Yang et al. 2003). Among the list of useful features of ANNs, many are favorable for FRPD prediction. The primary feature is that ANNs can represent any arbitrary nonlinear function, while in regression analysis relationships, or at best pre-specified nonlinearity, are needed (Xu et al. 2014). In ANN, the neural net discovers its own function with no limit associated with linearity. The other useful features are its ability to generalize a relationship from only a small subset of data, to remain generally vigorous in the presence of noisy inputs or missing input parameters, and to adapt and continue to learn even with evolving situations (Thube 2012). The main objective of this study was to introduce and develop a pavement performance model to predict Forest Road Pavement Deterioration (FRPD) and prioritize forest roads deterioration by applying artificial neural network and logistic regression model. The models can help forest engineers to define alternative ways of road maintenance, highly cost-effective and environmentally friendly in future. Moreover, the subgoal was to identify and quantify the new explanatory variables on FRPD. To date many models have been developed for forecasting of pavement conditions, most of them focusing on single index and all models relating to forest and public roads (Yang et al.

2003). Pavement deterioration models were developed in the present study to predict the forest roads deterioration based on current pavement conditions such as traffic loads, environmental, design, construction, and maintenance practices.

M.J. Heidari et al.

The roads were divided into a total of 185 road segments; road segments as defined by the road length between road drainage structures, intersections with other roads or trails, or changes in road condition (Coulter et al. 2006) and recorded by GPS. Within each segment, location of the initial sampling line was determined on the road by generating random value of 0–20 m and at 20 m interval away from the last line perpendicular to the wheel track (Fig. 1). Manual surveys were conducted by walking and noting the existing surface distress.

In a given segment, there were two main types of road inventory data, »Section« or »Continuous« data and »Event« or »Discrete« data to be collected, each of which needs a different and individual data collection treatment. Continues data such as grooves, pits and protrusions, checked rut and pothole were measured in linear or square meter. In contrast, event data such as road prism were described by a single change and an off-set from the center-line (Hill 2011).

A visual inspection is the first level of assessment and can be as simple as a walk-through the area. Rutting, protrusion and pothole were measured on crosssection as shown in Fig. 1. A stick marked in cm was used to measure the vertical distance between the road surface across the ruts, protrusion and pothole, and an aluminum bar mounted on a 1 m long rebar was driven into the ground on either side of the variables. The stick was aligned parallel to marks on the bar to ensure

2. Methods 2.1 Data collection The foundation of a successful FRPMS plan is the collection of data according to methods, standards, and protocols to be used in collecting pavement condition data. FRPMS rely on data from a variety of sources (e.g., roadway inventory, traffic data, materials, and construction history). This data maybe available or must be obtained by road inventory and managed so that it can be readily accessed by decision makers at all levels (McQueen and Timm 2005).

2.2 Road inventory Necessary details of all the roads have been taken with the road inventory method. A road inventory (manual survey) was completed on 50 km including primary and secondary roads during Oct–Nov 2014.

Fig. 1 Forest Roads Inventory for selected segments by generating random value of 0–20 m and at 20 m interval away from the last line

Croat. j. for. eng. 39(2018)2 273

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

the vertical measurements were made in the same orientation with the stick. These vertical distances were measured at 2.5 cm horizontally across each rut and pothole unless more closely spaced vertical measurements were necessary to adequately define the shape of the road surface (Gatto 2001).

Within each segment, forest roads pavement condition was assessed based on pavement current condition (thickness of the component layers, materials, grooves, pits and protrusions checked), management history (traffics, maximum load, age of pavement), road prism (road slope, longitudinal and lateral drainage (and road location properties (distance to stream, ground slope, canopy density). The survey data were generally recorded on paper and then entered into the computer every day to create the pavement management system (PMS) database. Survey data was used in this study for both modeling and implementation purposes.

2.2.1 Study area The Chob o kaghaz Mazandaran maintains approximately 400 km of primarily gravel surfaced low volume roads located in three separate forested tracts in Mazandaran province north of Iran, between 36°20’30’’ N and 36°23’58’’ N latitude and 45°17’30’’ and 52°18’35’’ E longitude. The area is part of the Caspian forest in northern Iran, with rough topography and dense vegetation cover (Jaafari et al. 2015). Elevations within the study area range from 150 to 800 meters above sea level. Mean annual precipitation for this area averages 867 millimeters, the primary and secondary roads were mostly under 30–40 years in age, and most have been reconstructed using current management practices in recent years; moreover, maintenance operations are done every six months. Average annual precipitation is about 872 mm and an average annual temperature ranges between 7 and 15 °C. The climate is humid and cold, according to Emberger climagram. The soil in this region includes soil type of brown forest and brown washed with argillic and calcic horizon.

Average monthly traffic was 425. The recentlygraded roads had more traffic because grading was generally a prerequisite to timber hauling. Most of the study sites were on the dry and wet season and average of timber hauling was 1200 m3 with three and two axle trucks.

2.2.2 Response variables Pavement management typically operates at two levels, (1) network level and (2) project level. At the network level, priority program and work schedule are developed within overall budget constraints. On the other hand, at the project level, specific physical improvements are implemented according to network

274

decisions (Shahnazari et al. 2012). Information collected as part of a network-level data collection effort may involve many items, but a standard set of data typically collected as response variables, including rutting, pothole and protrusion, were identified as response variables through the literature, while standards were extracted from the PMS database.

When forest harvesting equipment and other vehicles move across a forest road, rutting can occur.

Ruts are the trenches or furrows created by machine tires or tracks. Rutting displaces forest roads and damages it. Rutting is a normal occurrence for gravel roads.

Ruts are indicators of maintenance need. If ruts exceed 5 cm in depth or direct water down the road, or surface roughness affects the ability to travel on the road, it is time to perform surface maintenance.

Potholes are impressions in the forest roads caused by heavy traffic and often occur at lower slope level.

They are at least 3 cm wide and 3–5 cm long. Two different depth criteria (3 cm, 8 cm and 12 cm) apply, depending on the hazard of the standards being assessed. On sites with a high or very high deterioration hazard, or where the deterioration hazard has not been assessed, both depth criteria apply. On sites with a moderate or low deterioration hazard, only the greater than 12 cm depth criterion applies. This category does not require the survey point to be assessed for evidence of deterioration.

The category repeated machine traffic describes protrusion resulting from repeated heavy machine traffic.

Such protrusion is typically found on roads and especially on repeatedly used skid trails, which are obvious linear features. However, it occurs on heavy traffic areas associated with roadside work areas and in middle slope or upper slope level. This disturbance also occurs on moderate or low compaction pavement logged under dry conditions, where random skidding operations have a limited use of trails – one or two passes.

2.2.3 Explanatory variables Every variable that may affect pavement performance should be considered initially in road inventory. This list will typically be large. For their implementation within a FRPMS, however, predictive models must only use the variables that can be directly measured within acceptable cost and time constraints, retrieved from historical records, or computed or estimated (Zhang et al. 2013).

Previous studies prepared a summary of significant effective variables in FRPMS or highway or rural road (Dong 2011). Table 1 presents the list of variables for the current study.

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

M.J. Heidari et al.

Table 1 Important data elements in FRPMS deterioration Main category

Pavement condition

Environmental factors

*

Traffic (ADT and MADT)

Input variable

Quality of variable

Thickness of pavement, cm

Ordinal

Pavement material

Nominal

Classes

Min.

Max.

80.28

205.8

Mixed, river and mountain

1,2 1,2,3

Age of pavement, year

Ordinal

30

44

1,2,3,4

Maintenance historic, year

Nominal

0.60

1.20

1,2,3,4

Precipitation, mm

Ordinal

433

1910

1,2,3,4,5,6

Canopy, %

Scale

0

85

1,2,3,4

Elevation, m

Ordinal

143

838

1,2,3,4

Slope, %

Ordinal

1

23

1,2,3,4

Aspect

Nominal

0

4

N,S,E,W

ESAL, kN **

Scale

3

24

No

Frequency

0

11

No

Volume of timber, m

Scale

1067

2780

No

Loss of road

Presence or absence

0

1

Yes or No

Number of skids *** 3

Road qualify

Value

Sand in road ****

Presence or absence

0

1

Yes or No

Drainage

Presence or absence

0

1

Yes or No

Type of Road

Nominal

1

3

1,2,3

Intersection

Presence or absence

0

1

Yes or No

Status of ditch

Scale

0

8.79

No

Turn

Frequency

0

4

No

* (ADT/MADT): ADT: Average daily traffic, MADT: Maximum ADT ocurred in this network (Gralund and Puckett 1996) ** Equivalent standard axle loads calculated in accordance with ESAL calculator program (Martin et al. 2000) *** Total number of skids, timber trucks and truck brakes recorded by GPS **** Amount of sand on road surface, as a result poor compaction

The identified variables could be categorized under major topics that are known to affect performance. A preliminary list of important explanatory variables is prepared under four major categories, which affect long-term pavement behavior. These categories include pavement condition; environmental factors, traffic and road qualify as listed in Table 1. This list will be the primary source for explanatory variables.

3. Data Analysis LR and ANNs models were developed to model the overall pavement conditions, encompassing the individual pothole, protrusion and rut ratings. The SPSS version 16 was used for data analysis with logistic regression, and the default ANNs training algorithm of NeuroSolution Infinity software version 1.1.0.1 was used for neural network purpose.

Table 2 Subclasses of response variable (Reid and Dunne 1984, Smith 1993, Yee and Roelofs 1980) Variable

Class 1

Class 2

Class 3

Class 4

Pothole

Depth < 3 cm

Depth = 3–8 cm

Depth 8–12 and < 12 cm and Area > 1 m2

Depth >12 cm

Rutting

Depth < 5 cm

Depth = 5–10 cm

Depth = 10–15 cm

Depth > 15 cm

Height = 5–8 cm

Height > 8 cm or Area > 2 m * 3 m (6 m2) in each class

Protrusion

Height < 3 cm

Height = 3–5 cm

Croat. j. for. eng. 39(2018)2 275

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

3.1 Response Variable Classification Initially, response variables including pothole, rutting and protrusion were divided into four sub-classes (Table. 2).

3.2 Logistic Regression Model In this study, since the classification within the dependent variable has no meaning, ordinal logistic is used for each class. Logistic regression builds a linear model based on a transformed variable using a link function referred to as the Logit function or model, which is the log of the odds that an event occurs. The maximum likelihood estimation procedure is used to obtain the estimates of the regression coefficients by maximizing the value of log-likelihood function through an iterative process, with the aim of making the likelihood of observed data greater (Hosmer et al. 2013). The number of logistic regression equations required is usually lower by one category because one of the prediction categories is chosen as a reference category.

3.3 ANN Model An ANN is a massively parallel distributed information processing system that has certain performance characteristics resembling biological neural networks of the human brain (Suman and Sinha 2012). ANNs have been developed as a generalization of mathematical models of human cognition or neural biology (Izenman 2008, Movagharnejad and Nikzad 2007, Rao 2000). A neural network is characterized by its architecture that represents the pattern of connection between nodes, its method of determining the connection weights and the activation function (Russell C. Eberhart 1990, Si et al. 2015). The basic structure of a network usually consists of three layers: the input layer, where the data are introduced to the network; the hidden layer or layers, where data are processed; and the output layer, where the results for given inputs are produced (Kumar et al. 2013, Suman and Sinha 2012).

3.3.1 Model Architecture According to the database partitioning, the validation dataset has been considered statistically independent from the datasets used for training and testing purposes. Hence, the verification of ANN models through using the validation dataset can be considered a touchstone in examining the performance of the developed ANN models from an implementation point of view (Thube 2012). The selection of ANN architecture is not a decision making process. Most of the time, trial and error, combined with engineering judgment, is used to determine the appropriate architecture for a particular problem (Thube 2012). In the present

276

study, a number of explanatory and response variables are kept constant, and variations are made in the hidden layers and in the neurons per hidden layers with the software default. First, the depth of the experiment search must be determined. The defaults are designed to choose the most commonly used preprocessing functions and neural network topologies, but if the computing resources are limited, the options can be changed to either limited, partial or none for a less thorough search (Abu Jamous 2013).

3.3.2 Data Optimization It is necessary to determine how the data should be allocated for optimization. The database was divided into three datasets, and the first set has been used for training purposes. One set contains 70% (125 segments) of the data that are used for network training, and the remaining set contains 15% (30 segment) of the data used for network testing and 15% (30 segment) for validation.

3.4 The ROC Curve Receiver Operating Characteristic (ROC) is used for evaluating two models. A ROC is a standard technique for summarizing classifier performance over a range of trade-offs between true positive (TP) and false positive (FP) error rates (Phillips et al. 2015). ROC curve is a plot of sensitivity (the ability of the model to predict an event correctly) versus 1-specificity for the possible cut-off classification probability values P0 (Humphrey et al. 2012). For logistic regression, it is necessary to create a 2×2 classification table of predicted values from model for response if y^=0 or 1 versus the true value of y=0 or 1 (Sharma et al. 2011).

A rough guide for classifying the accuracy of a diagnostic test is the traditional academic point system: 0.90–1 = excellent (A), 0.80–0.90 = good (B), 0.70–0.80 = fair (C), 0.60–0.70 = poor (D) and 0.50–0.60 = fail (F) (Hosmer Jr. et al. 2013).

While the ROC curve contains most of the information about the accuracy of a continuous predictor, it is sometimes desirable to produce quantitative summary measures of the ROC curve (Anifah et al. 2013).

The most commonly used such measure is by far the area under the ROC curves (AUC) (Friedman et al.

2000). In an empirical ROC curve, this is usually estimated by the trapezoidal rule, which forms trapezoids using the observed points as corners, computing the areas of these trapezoids and then adding them up (Gonen 2006). This may be quite an effort for a curve with many possible thresholds. Fortunately, AUC is connected to a couple of well-known statistical measures that facilitates comparison and improves interpretation (Hosmer Jr. et al. 2013).

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

M.J. Heidari et al.

Table 3 Model summary of input explanatory response variable at four levels of pothole Variable

Sig.

Wald Test

Standard deviation

Df

Coefficient

Iteration 1

0.300

4.7

1.234

1

–2.76

Iteration 2

0.233

1.4

1.231

1

–1.47

Iteration 3

0.599

0.2

1.224

1

0.643

Slope

0.206**

4.9

0.011

1

0.002

***

Date of repair

0.015

24.5

1.08

1

–3.002

Turn

0.251**

Percent of canopy

1.5

0.086

1

–0.009

***

14.74

0.007

1

–0.332

**

0.000

Thickness of pavement

0.222

1.4

0.006

1

1.320

ESAL

0.000***

75.6

1.616

1

2.279

Drainage

**

0.599

0.3

0.082

1

0.480

Material

0.025***

8.5

0.412

1

1.823

0.806

0.769

0.736

0.804

–

AUC 1,2,3,4 *** Strong relation ** Medium relation

4. Result

4.1 LR Models

To simulate FRPD and evaluate Forest road pavement performance three response variables and 19 explanatory variables were defined based on literature review and field survey.

4.1.1 Pothole Model LR was applied to model pothole at four subclasses (Table 2). The result of pothole analysis and ROC evaluating model are presented in Table 3.

Table 4 Model summary of input explanatory response variable for rutting Variable

Sig.

Wald Test

Standard deviation

Df

Coefficient

Iteration 1

0.492

0.47

1.124

1

0.772

Iteration 2

0.022

5.218

1.135

1

0.594

Iteration 3

0.000

12.45

1.126

1

4.078

Slope

0.06 ***

0.658

0.011

1

0.921

Date of repair

0.382 **

1.237

1.37

1

1.465

Turn

0.000 ***

72.241

0.111

1

–0.004

Percent of canopy

0.534 **

0.386

0.007

1

0.004

Thickness of pavement

0.049 ***

0.976

0.006

1

1.001

ESAL

0.081 ***

3.043

-1.880

1

–3.281

Drainage

0.266 **

0.1

0.078

1

0.012

Material

0.982 **

0.915

0.429

1

0.01

0.946

0.827

0.748

0.965

–

AUC 1,2,3,4 *** Strong relation ** Medium relation

Croat. j. for. eng. 39(2018)2 277

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

Table 5 Model summary of input explanatory response variable for protrusion Variable

Sig.

Wald Test

Standard deviation

Df

Coefficient

Iteration 1

0.000

16.526

1.295

1

–5.265

Iteration 2

0.004

8.281

1.281

1

–3.688

Iteration 3

0.371

0.8

1.249

1

–1.118

Slope

0.02 ***

4.132

0.013

1

0.525

Date of repair

0.05 ***

3.855

1.567

1

3.076

Turn

0.03 ***

Percent of canopy Thickness of pavement ESAL

4.196

0.091

1

0.047

0.375

**

0.856

0.008

1

–0.007

0.328

**

0.957

0.006

1

–0.006

10.406

-2.280

1

5.258

0.000 ***

Drainage

0.909 **

0013

0.083

1

0.009

Material

0.083 **

2.282

0.439

1

–0.064

0.921

0.923

0.806

0.935

–

AUC 1,2,3,4 *** Strong relation ** Medium relation

According to Wald test for all classes of pothole levels, the most important variables were date of repair, ESAL, percent of canopy and material.

According to the results, the maximum and minimum AUC were found in class 1 and class 3, 0.806 and 0.736, respectively (Table 3). Table 3 shows four AUC representing excellent, good, and fair.

4.1.2 Rutting Model The results of rutting analysis and evaluating (AUC) model are presented in Table 4.

Wald test for rutting showed that the most important variable for this responsible variable were turn,

thickness of pavement, ESAL and slope. Similar to the results obtained for pothole, the results showed that the maximum AUC was found in class 4 and minimum in class 3 (Table 4).

4.1.3 Protrusion Model Protrusion LR analysis and evaluating (AUC) are presented in Table 5.

According to Wald test ESAL, date of repair, material, number of turn and slope are the effective factors in protrusion formation. In protrusion, maximum AUC was found in class 4 (Table 5) and minimum in class 3. The best tested plot in protrusion showed that all classes were excellent excluding class 3 (Table 5).

Table 6 Model overview for Pothole Variable

Pothole

278

Model

Experiment

Project

Input

Percent of Issue

Dataset

Optim.

Leave-N-Out

Date of Repair

20.6

Score

98.82

65.75

Canopy

19.6

Percent correct

100

85.23

Material

18.3

Avg. area ROC

1

0.880

ESAL

15.2

Avg. correlation

0.91

0.78

Slope

15.1

Avg. norm. MSE

0.0892

0.0377

Drainage

9.5

Avg. norm. MAE

0.099

0.234

Precipitation

1.7

Max. abs. error

0.044

1

–

–

Training epochs

3

–

–

–

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271â&#x20AC;&#x201C;287)

M.J. Heidari et al.

4.2 ANN Modeling 4.2.1 Pothole Model Based on the prediction of pothole, a satisfactory prediction model has been developed. Given the various ANNs model overview for pothole (Table 6), the weights of links among the neurons are determined through the training process. The training process has been carried out for a fixed number of epochs (10,000) (Semeida 2015).

The model comparisons for different ANN models are carried out by comparing the normal mean square error (NMSE) values during testing stage. The details of NMSE and average correlation variations for different ANN models are shown in Table 6.

The results indicated that the final model has a very good NMSE, NMAE and AUC (Table 6). Based on Table 6, the higher values of Percent of Issue indicate that the variable is relatively more important. The five most influential variables on the failure of the pavement and pothole were the date of repair (20.6), material (18.3), ESAL (15.2), Slope (15.1) and drainage (9.5).

Finally, the ANN models correspond to the ROC curve and average correlation at the testing stage was selected. Another critical step, prior to the actual application of the developed model, is to evaluate the performance of the model. The details of the evaluation for pothole ANN models are shown in Fig. 2 and 3 as an example, while those for other models were removed to reduce the volume of figures.

Fig. 3 Pothole ROC Curve in ANN model The correlation of pothole model showed that maximum correlation was achieved in thirty minutes after running the model, after which there was no performance. However, the disturbances of explanatory variable got better and Maximum performance came to 0.9. AUC was used to compare the evaluation of this model with pothole model generated by LR (Fig 3).

The AUC shows very good disturbance of explanatory variable a few minute after running the model.

The AUC in this pothole ANN model was at first 0.6 and after thirty minutes it reached the maximum performance (0.88).

4.2.2 Rutting Model Similar to LR modeling, to develop the rutting model, rutting data was classified into four classes.

Total rutting results showed that the most important variables were: the thickness of pavement, Elevation, turn and ESAL over timeline (Table 7).

As can be seen from Table 7, the percentage of Thickness (30.3) has a strong impact on the rutting followed by Elevation (22) and ESAL (13.6).

Fig. 2 Correlation between explanatory variable in Pothole ANN model

4.2.3 Protrusion Model To develop the protrusion model, the data was classified similar to those in LR modeling. To analyze the influence of explanatory factors on FRPD, the relatively effective and ineffective maintenance were distinguished. Pavement that received relatively effective treatment can be determined using protrusion condition. These types of response (Correlation and ROC) are illustrated in Table 8.

Croat. j. for. eng. 39(2018)2 279

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

Table 7 Rutting model overview

Rutting

Model

Experiment

Project

Input

Percent of Issue

Dataset

Optim.

Leave-N-Out

Thickness of pavement

30.3

Score

99.985

68.56

Elevation

22

Percent Correct

100

83

Turn

15

Avg. Area ROC

1

0.854

ESAL

13.6

Avg. Correlation

1

0.74

Material

9.6

Avg. Norm. MSE

0

0.043

Slope

3.4

Avg. Norm. MAE

0

0.0198

Date of Repair

5.8

Max. Abs. Error

0

1

Drainage

2.8

Training Epochs

3

–

Precipitation

1.5

Table 8 Protrusion model overview

Protrusion

Model

Experiment

Project

Input

Percent of Issue

Dataset

Optim.

Leave-N-Out

ESAL

36.1

Score

99.275

75.376

Slope

35.7

Percent Correct

100

98.65

Material

12.9

Avg. Area ROC

1

0.921

Drainage

7.5

Avg. Correlation

0.99

0.83

Turn

6

Avg. Norm. MSE

0.0592

0.0283

Canopy

1.7

Avg. Norm. MAE

0.065

0.140

–

–

Max. Abs. Error

0.103

0.92

–

–

Training Epochs

3

–

–

–

As can be seen from Table 8, the percentage of ESAL (36.1), slope (35.7) and material (12.9) were considered the most significant variables influencing the protrusion. It is important to note that more explanatory variables mentioned in Table 8, such as channel and canopy, is included in the functional form of models. Finally, the protrusion models corresponding to the correlation and ROC curve at the testing stage were selected.

The effect of important explanatory variable on response variable is shown in Fig 4. These figures show the effect of management operations in output model and FRPMS.

(PCP) and Root Mean Square Error (RMSE) (Table 10).

The LR and ANNS were able to classify precisely 89% and 82% of the pavement segments, respectively.

Table 9 Comparison of LR and ANNs models in assessing pavement deterioration condition Pavement deterioration Pothole

4.3 Comparison of models

Rutting

The models developed by LR and ANNs were then applied to the data set and their performances were compared by AUC, and Percent of Correct Prediction

Protrusion

280

Model description

AUC

PCP

RMSE

ANN model

0.880

85.2%

0.194

LR model

0.832

81.2%

0.253

ANN model

0.854

83%

0.207

LR model

0.910

83.5%

0.265

ANN model

0.921

88.6%

0.168

LR model

0.817

81.7%

0.244

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271â&#x20AC;&#x201C;287)

M.J. Heidari et al.

Fig. 4 Effect of important explanatory variable on response variable in FRP

Predictions of rutting, pothole and protrusion are carried out by using the trained ANN models of selected architectures, as well as by using the LR model.

The maximum protrusion accuracy (88.6%) was achieved with ANN model. Pothole and rutting achieved the maximum accuracy with ANN and LR model.

5. Discussions and Conclusions Forest Road Pavement Management is a topic of great significance in forest engineering. It is essential to develop reliable pavement management systems, which have the ability to estimate the overall pavement condition and the ability to forecast when and

Croat. j. for. eng. 39(2018)2 281

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271â&#x20AC;&#x201C;287)

what kind of repair will be needed on certain pavements. The models of the pavement performance prediction are developed using the past pavement performance data. Thus Pavement Performance Prediction models are integrated into the decision making process and help to schedule the repairs and estimate the budgets (Kaur and Pulugurta 2008).

ESAL had the most significant effect on FRPD (pothole, ruts and protrusion) in both models. By definition, it removes the effect of pavement design, age, and condition variables. For example, one ESAL on a strong pavement corresponds to a much lower proportion of its fatigue life than one ESAL on a weak pavement (Sun et al. 2007). For this reason, its effect is significant in our forest roads because the average age of the road is 35. In the study area, the transporting machines with maximum capacity of logs are used to reduce the logging and timber transportation costs.

Max ESAL found in forest roads was more than 20 kN and in this segment max deterioration was observed.

ADT and MADT sized pieces resulting from the weight of the truck and friction between the tire and the aggregate (Miller 2014). These smaller, fine particles are then more easily mobilized. During wet weather hauling, the weight of the truck on the road layer may also cause fine sediment from the subgrade to move upward indicate the traffic impact on pavement performance; it merges with ESAL or at least includes the percentage of truck information in our model. It is a well-known fact that roads with high levels of traffic, especially truck traffic, need to be repaired more often than roads with lower levels of traffic. Higher traffic levels increase the ESAL as well as the volume of fine material, and this is a major reason why traffic increases pavement deterioration (Smith 1993). Log truck traffic increases pavement deterioration by increasing the availability of fine sediment on the road surface (Fassman and Blackbourn 2011). Degradation of the surface aggregate into smaller sized pieces is the result of the truck weight and friction between the tire and the aggregate (Miller 2014). These smaller, fine particles are then more easily mobilized.

During hauling in wet weather, the weight of the truck on the road layer may also cause fine sediment from the subgrade to move upward to the surface in a process known as pumping (Schaefer et al. 2008). Larger logs on trucks can cause breaking of the pavementâ&#x20AC;&#x2122;s upper layer providing conditions for the water to get into roadbed and cause deterioration and rutting of pavements (Wang 2011). Heavy vehicles will do far more damage to pavements than lighter vehicles. In the current research, ESAL in LR and ANNs model had a most significant role in FRPD and this is in line

282

with the research of (Adlinge and Gupta 2013, Fassman and Blackbourn 2011, Miller 2014, Peshkin 2011, Schaefer et al. 2008, Smith 1993, Wang and Al-Qadi 2009).

The most common maintenance activity that influences pavement deterioration is the date of the last repair of pavements (Zhang et al. 2010). The results of this research showed that the increased maintenance activity resulted in lower pavement deterioration.

These results indicate that roads that are not adequately maintained become deteriorated, and are more deteriorated than well maintained roads. The mentioned pavements should be blocked to be repaired and maintained after the logging operations (Miller 2014).

With the start of maintenance operations, road traffic increases and hence at the beginning of the road maintenance, deterioration is significant. However, with time, after maintenance operations, the deterioration of the road surface is reduced. We also believe that increasing the thickness of pavements, in the course of maintenance, is a reason to increase FRPD (Fig. 4), because maintenance operations do not improve compaction in these segments.

The composition (mixed, riverine and mountain) and thickness of road surface materials influences the FRPD, as high quality rock will not degrade into smaller, more mobile particle sizes (PĂŠrez and Gallego 2010). Segments covered by the mountain material (40%) deteriorated more rapidly than those paved by mix or riverine materials. Mixed materials were found in 35 percent of segments, and they had lower deterioration. The lowest deterioration was measured in the segment with riverine material pavement (Fig. 4).

Thickness of pavement provides insurance against deterioration from the bottom layer (Giroud and Han 2004). There is correlation between the rate of pavement deterioration and pavement thickness (Giroud and Han 2004). Apparently, low thickness in higher traffic has a much greater effect causing deterioration (protrusion and pothole). The results showed that, with the decrease in thickness, pothole and protrusion increase, while rutting appears in high thickness (Fig. 4). When thickness is low, soil strength is not sufficient to support the applied load from vehicles or equipment traffic (ESAL) and thus potholes and protrusions occur on forest roads and trails (Cambi et al. 2015).

High thickness of forest roads provide the surface for rutting in wet season, and after maintenance operations, traffic increases and tire pressure causes rutting (Fig. 4). The increase of water pressure can make completion material unsuitable and unstable and this may result in permanent deformation of the road surface and cause rutting (Rodgers et al. 2014). With the inCroat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

crease of thickness, rutting decreases. This is quite the contrary with protrusion, because with high thickness, first rutting occurs and after that protrusion appears.

Protrusion is the result of soil compaction of heavy machinery during high traffic and dry season.

Road slope and elevation are two characteristics that often correlated with increased FRPD. This is physically intuitive because, as the slope and elevation of a road increase, potential energy increases and leads to higher erosive power of the skid log trucks (Loizos and Plati 2008). Pavement deterioration is the decisive deterioration process on inclined roads. While potholes dominate on horizontal road segments, rutting and protrusion are found in high slope segments (Moghadami Rad et al. 2014). Fig. 8 shows the probability of maximum pothole occurrence at zero percent gradients, reducing at 5–8% gradient. The study layout results at lower road gradients (Reid and Dunne 1984). Ruts in upslope and high elevation can be filled with water causing it to drain along the road instead of draining away from the road (Caliskan 2013). Heavily sloped roads (those with slopes greater than 10%) can become rutted very easily, because the driver/operator uses extra capacity of the road when driving with heavy loads or under wet conditions. The probability of protrusion occurrence containing pothole distresses has been recorded at low slope, being significant at 5–8% gradient, disappearing at 0–2% gradient (Fig. 4). Protrusions behave like ruts in slope and elevation.

The results indicate that road surface drainage was effective in preventing the development of deterioration (Fu et al. 2010). Engineered points were less than 20 percent of the pavement deterioration. The average road segment length that was drained by cross-drain culverts or live-stream crossing culverts was appreciably different from the average for the entire database (Fig. 4). As the contact pressure from a tire is mainly supported by the completion layer, the load from the tire can increase the pore water pressure in the road material when drainage is restricted. When water remains on the road surface on low slope segments, potholes appear, while protrusion is seen in mountain roads.

The percentage of canopy that covers the road pavement is an important factor in pothole and protrusion risk and has a significant influence on FRPD as well (Eskioglou 2003). Dense canopies protect the road surface from the water drop. However, when interception decreases in high precipitation, deterioration increases. When canopy tends to be denser, light that reaches the road surface is reduced, and in this case, the road surface remains wet and severe deterio-

M.J. Heidari et al.

ration occurs. The results showed that high canopy, the highest measured was 80%, was sufficient to protect the road surface very well against precipitation and was enough open to let light reach the road surface. Potholes were observed in road classes one and two that have low slope and canopy density. While the concentrated flow of surface water is the cause of erosion, protrusions are formed by pounding water.

Therefore, the total amount of water falling on a certain area in high slope is an indicator for the occurrence of protrusions.

A FRP is susceptible to variations in climatic conditions of the area in which the road is located. Since the segments are neighbors, the historical precipitation data did not vary significantly (Table 6 and 7), while pothole and rutting precipitation varied significantly.

In rainy season, the moisture content in the road becomes higher and consequently the bearing capacity of the FRP is generally reduced causing rutting of the roads. In dry season, the moisture content of the FRP is reduced and this causes road protrusion.

Turns with drainage conditions were the effective variables on road deterioration in both models. The results showed that pavement deterioration increased with decreasing horizontal curve radius in turn. According to the results, due to drainage lakes and uneven load distribution, deterioration was more severe in turns. One reason is that, in view of higher stress on curves, material is dislodged and thrown into the ditches. When the speed is constant, the centrifugal force of the moving trucks increases with the decrease in the radius of the horizontal curve (Kordani and Molan 2014). This can distribute the uneven load on the road surface. If the specific slope was not considered on horizontal curves, the water would be collected on the road surface and the rutting and protrusion would occur severely. Moreover, the pavement layer of the road is damaged by increasing brake on horizontal curves and high longitudinal slopes (Burton et al.

2014). In order to prevent pavement deterioration on road surface, the longitudinal slope should be decreased on horizontal curve to five degrees (Aricak 2015). By increasing the number of turns, pavement deterioration increased but with only two turns, deterioration was less. This is due to the lack of deceleration of the truck driver when he makes two consecutive turns not reducing the vehicle speed due to good visibility.

The pavement deterioration of forest roads varies as a result of length of time since construction, date of maintenance, pavement condition and traffic. The 19 major explanatory variables were considered to investigate the type of deterioration of forest roads with LR

Croat. j. for. eng. 39(2018)2 283

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

and ANN models. The factors that cause deterioration, potholes, ruts, and protrusion, will be the final input parameters for FRPMS. Hassan (2015) reported that logistic regression modeling is feasible for developing deterioration models of subjective distress data of pavement surfaces.

This paper develops two type models using pavement distress data for forest roads. The deterioration modeling was based on FRP condition and response variable (pothole, rutting and protrusion) using two different models (ANN and LR). The results showed that ANN and LR models could be applied for the pothole, rutting, protrusion, and deterioration progression modeling of forest roads. These results are the same as those of Kaur and Pulugurta (2008) with the accuracy of the logistic regression model. Both models predicted two indices on rutting, pothole and protrusion: extent and severity. ANN and LR models were examined by carrying out various trials. The models showed a high area under ROC curve (AUC) between observed and predicted distresses of more than one ratio. This shows an efficacy of the suggested ANN and LR models (Hassan 2015). Although ANN models showed higher efficiency, the results obtained from LR were desirable. LR can describe very well the relationship between pothole, rutting and protrusion and a set of predictor variables due to model responsibility in terms of natural data involving an environmental condition. The models that have been chosen in this study are relevant to all forest environments.

The origin of the model as well as the places and diversities of applications of the models provide an indication of suitability.

Forest road pavement management database consists of many different attributes that are both continuous and categorical in nature. In pavement management, it is often required to determine the type of repair needed for a pavement. This decision is based on the condition of the pavement - whether it is in good condition or fair condition, and also on different attributes such as traffic, weather conditions, etc. It is a complicated process to develop a statistical model based on all these attributes. In this study, a more straightforward approach was used using the actual data. An Artificial Neural Network was generated and then converted to simple rules. The rules were then tested on a test data set and the results showed that the accuracy of the model was approximately 85%.

Furthermore, a logistic regression was used to classify the dataset and the results of the logistic regression model were compared to the ANNs. The accuracy of the logistic regression model was 82%.

284

There are several directions for future work because these generated models are the first models applied to forest roads. Further study is recommended to validate their performance in other forest roads and other conditions. Models that predict the pavement performance in feature years based on the current pavement distress condition can be a very crucial tool for allocating budget among alternative pavement managements and preservation projects for forest management authorities.

Acknowledgment We would like to acknowledge Reza Goudarzi and Mehdi Kheirianpour for the collaboration in field surveys. Appreciation also goes to Chobo O Kaghaz Company for their constructive comments on calibration and editorial advice and support.

6. References Abu Jamous, H., 2013: Parametric Cost Estimation of Road Projects Using Artificial Neural Networks. M.Sc. Thesis, Islamic University – Gaza Deanery of Graduate Studies, 1–100.

Adlinge, S., Gupta, A., 2013: Pavement Deterioration and its Causes. International Journal of Innovative Research and Development 2(4): 9–15.

Anifah, L., Purnama, I.K.E., Hariadi, M., Purnomo, M.H., 2013: Osteoarthritis classification using self organizing map based on gabor kernel and contrast-limited adaptive histogram equalization. The open biomedical engineering journal 7: 18–28.

Aricak, B., 2015: Using remote sensing data to predict road fill areas and areas affected by fill erosion with planned forest road construction: a case study in Kastamonu Regional Forest Directorate (Turkey). Environmental monitoring and assessment 187(7): 417.

Bent, D., Rusu, L., Podean, M., Arba, R., 2012: Web Based pavement maintenance and monitoring system, in: 2nd World Conference on Innovation and Computer Sciences.

Procedia Information Technology & Computer Science, Cluj Napoca, Romania, 353–357 p.

Burton, M., Kim, M., Smit, A., Trevino, M., Wu, H., Murphy, M., Prozzi, J., 2014: Flexible Pavement Narrow Widening Best Practices and Lessons Learned. National Technical Information Service 6748, 1–175.

Caliskan, E., 2013: Environmental impacts of forest road construction on mountainous terrain. Iranian journal of environmental health science & engineering 10(1): 23.

Cambi, M., Certini, G., Neri, F., Marchi, E., 2015: The impact of heavy traffic on forest soils: A review. Forest Ecology and Management 338: 124–138.

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287) Coulter, E.D., Sessions, J., Wing, M.G., 2006: Scheduling forest road maintenance using the analytic hierarchy process and heuristics. Silva Fennica 40(1): 143–160.

Dong, Q., 2011: Enhancement of Pavement Maintenance Decision Making by Evaluating the Effectiveness of Pavement Maintenance Treatments. PhD. Dissertation, University of Tennessee, 1–158.

Eroglu, H., Acar, H., Eker, M., 2003: An Evaluation of The Forest Road Superstructure in Turkey, in: 12th World Forestry Congress. Congress Procceding, Quebec, 4–6.

Eskioglou, P.C.H., 2003: Forest Engineering and Environment Protection. 8th International Conference on Environmental Science and Technology. Lemnos island, Greece 8–10 September, 186–192.

Faghih-Imani, A., Amador Jimenez, L., 2013: Toward Sustainable Pavement Management: Incorporating Environmental Impacts of Pavement Treatments into a PerformanceBased Optimization. Transportation Research Record: Journal of the Transportation Research Board 2366: 13–21.

Fassman, E.A., Blackbourn, S.D., 2011: Road Runoff WaterQuality Mitigation by Permeable Modular Concrete Pavers.

Journal of Irrigation and Drainage Engineering 137(11): 720–729.

Forsyth, A.R., Bubb, K.A., Cox, M.E., 2006: Runoff, sediment loss and water quality from forest roads in a southeast Queensland coastal plain Pinus plantation. Forest Ecology and Management 221 (3–1): 194–206.

Friedman, J., Hastie, T., Tibshirani, R., 2000: Additive logistic regression: a statistical view of boosting (With discussion and a rejoinder by the authors). The Annals of Statistics 28(2): 337–407.

Fu, B., Newham, L.T.H.H., Ramos-Scharron, C.E., RamosScharrón, C.E., 2010: A review of surface erosion and sediment delivery models for unsealed roads. Environmental Modelling & Software 25(1): 1–14.

M.J. Heidari et al.

prove deterioration forecasts for infrastructure asset management. KSCE Journal of Civil Engineering 18(7): 2107–2119.

Hassan, R.A., 2015: Modelling bituminous surfacing distress data using logistic regression. WIT Transactions on The Built Environment 146: 435–446.

Hill, M., 2011. Building design and construction handbook, 6th ed. Donnelley and Sons Company, New York.

Hosmer Jr, D.W., Lemeshow, S., Sturdivant, R.X., 2013. Applied logistic regression. John Wiley & Sons.

Humphrey, L., Arbuckle, R., Moldwin, R., Nordling, J., van de Merwe, J.P., Meunier, J., Crook, T., Abraham, L., 2012: The bladder pain/interstitial cystitis symptom score: development, validation, and identification of a cut score. European urology 61(2): 271–279.

Izenman, A.J., 2008. Modern Multivariate Statistical Techniques, Springer Texts in Statistics. Springer New York, New York, NY. Jaafari, A., Najafi, A., Rezaeian, J., Sattarian, A., Ghajar, I., 2015: Planning road networks in landslide-prone areas: A case study from the northern forests of Iran. Land Use Policy 47: 198–208.

Kaur, D., Pulugurta, H., 2008: Comparative analysis of fuzzy decision tree and logistic regression methods for pavement treatment prediction. WSEAS Transactions on Information Science and Applications 5(6): 979–990.

Kordani, A.A., Molan, A.M., 2014:The effect of combined horizontal curve and longitudinal grade on side friction factors. KSCE Journal of Civil Engineering 19(1): 303–310.

Kumar, K., Parida, M., Katiyar, V.K., 2013: Short Term Traffic Flow Prediction for a Non Urban Highway Using Artificial Neural Network. Procedia – Social and Behavioral Sciences 104, 755–764.

Lee, S.-T., Park, D.-W., Mission, J.L., 2013: Estimation of pavement rehabilitation cost using pavement management data. Structure and Infrastructure Engineering 9(5): 458–464.

Gatto, L.W., 2001: Overwinter changes to vehicle ruts and natural rills and effects on soil erosion potential, in: 10th International Soil Conservation Organization Meeting. National Soil Erosion Laboratory, Purdue University, 378–383.

Lin, O., Hassan, R., Thananjeyan, A., 2014: Pavement condition modelling by using surface condition data–a VicRoads study. ARRB Conference, 26th, Sydney, New South Wales, Australia 8(2): 16 p.

Giroud, J.P., Han, J., 2004: Design Method for Geogrid-Reinforced Unpaved Roads. II. Calibration and Applications.

Journal of Geotechnical and Geoenvironmental Engineering 130(8): 787–797.

Loizos, A., Plati, C., 2008: An alternative approach to pavement roughness evaluation. International Journal of Pavement Engineering 9(1): 69–78.

Gonen, M., 2006: Receiver operating characteristic (ROC) curves. SAS Users Group International (SUGI) 31, 210–231.

Hahne, J.M., Biebmann, F., Jiang, N., Rehbaum, H., Farina, D., Meinecke, F.C., Muller, K.-R., Parra, L.C., 2014: Linear and nonlinear regression techniques for simultaneous and proportional myoelectric control. Neural Systems and Rehabilitation Engineering, IEEE Transactions on 22, 269–279.

Han, D., Kaito, K., Kobayashi, K., 2014: Application of Bayesian estimation method with Markov hazard model to im-

McQueen, J., Timm, D., 2005: Part 2: Pavement Monitoring, Evaluation, and Data Storage: Statistical Analysis of Automated Versus Manual Pavement Condition Surveys. Transportation Research Record: Journal of the Transportation Research Board 1940: 53–62.

Miller, R.H., 2014: Influence of log truck traffic and road hydrology on sediment yield in western Oregon. PhD. Dissertation.

Moghadami Rad, M., Abdi, E., Mohseni Saravi, M., Rouhani, H., Majnounian, B., 2014: Effect of Forest Road Gradient on

Croat. j. for. eng. 39(2018)2 285

M.J. Heidari et al.

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287)

Amount of Runoff and Sediment (Case Study: KohmiyanAzadshahr Forest). Journal of Forest and wood products 66: 389–399.

Moreno-Navarro, F., Sol-Sánchez, M., Rubio-Gámez, M.C., 2015: Exploring the recovery of fatigue damage in bituminous mixtures: the role of healing. Road Materials and Pavement Design 16(1): 75–89.

Movagharnejad, K., Nikzad, M., 2007: Modeling of tomato drying using artificial neural network. Computers and Electronics in Agriculture 59(1–2): 78–85.

Ouma, Y.O., Opudo, J., Nyambenya, S., 2015: Comparison of Fuzzy AHP and Fuzzy TOPSIS for Road Pavement Maintenance Prioritization: Methodological Exposition and Case Study. Advances in Civil Engineering 2015: 1–17.

Owolabi, A., Oladapo Samson, A., 2011: Development of Priority Index Assessment Model for Road Pavements in Nigeria, in: 8th International Conference on Managing Pavement Assets. Managing Pavement Assets, Abeokuta, 12 p.

Pérez, I., Gallego, J., 2010: Rutting prediction of a granular material for base layers of low-traffic roads. Construction and Building Materials 24(3): 340–345.

Peshkin, D.G., 2011: Guidelines for the Preservation of Hightraffic-volume Roadways. Transportation Research Board, 511 p.

Phillips, J., Cripps, E., Lau, J.W., Hodkiewicz, M.R., 2015: Classifying machinery condition using oil samples and binary logistic regression. Mechanical Systems and Signal Processing 60: 316–325.

Potočnik, I., Hribernik, B., Nevečerel, H., Pentek, T., 2015: Maintenance of forest roads – the need for sustainable forest management, in: Forest Engineering Current Situation and Future Challenges. CROJFE, Zalesina, Zagreb, 5–8 p.

Rao, S.G., 2000: Artificial Neural Networks in Hydrology. I: Preliminary Concepts. Journal of Hydrologic Engineering 5(2): 115–123.

Reid, L.M., Dunne, T., 1984: Sediment production from forest road surfaces. Water Resources Research 20(11): 1753– 1761.

Roberts, C.A., Attoh-Okine, N.O., 1998: A Comparative Analysis of Two Artificial Neural Networks Using Pavement Performance Prediction. Computer-Aided Civil and Infrastructure Engineering 13(5): 339–348.

Rodgers, M., Kielty, A., Healy, M.G., 2014: Performance of Granitic, Shale, and Limestone Forest Road Aggregates Subjected to Repeated Loading. Journal of Transportation Engineering 140(4): 4014002.

Russell, C., Eberhart, R.W.D., 1990: Neural Network PC Tools: A Practical Guide. Academic Press, 300 p.

Rusu, L., Taut, D.A.S., Jecan, S., 2015: An Integrated Solution for Pavement Management and Monitoring Systems. Procedia Economics and Finance 27: 14–21.

286

Saha, P., Liu, R., Melson, C., Boyles, S.D., 2014: Network Model for Rural Roadway Tolling with Pavement Deterioration and Repair. Computer-Aided Civil and Infrastructure Engineering 29(5): 315–329.

Salour, F., Erlingsson, S., 2013: Investigation of a pavement structural behaviour during spring thaw using falling weight deflectometer. Road Materials and Pavement Design 14(1): 141–158.

Santos, J., Ferreira, A., 2013: Life-cycle cost analysis system for pavement management at project level. International Journal of Pavement Engineering 14(1): 71–84.

Schaefer, V., White, D., Ceylan, H., Stevens, L., 2008: Design Guide for Improved Quality of Roadway Subgrades and Subbases. InTrans Project Reports.

Schlotjes, M.R., 2013: The development of a diagnostic approach to predicting the probability of road pavement failure. PhD. Dissertation, ResearchSpace, Auckland.

Semeida, A.M., 2015: Derivation of level of service by artificial neural networks at horizontal curves: a case study in Egypt. European Transport Research Review 7(1): 4.

Sen, T., 2013: Pavement management analysis of Hamilton County using HDM-4 and HPMA. Masters Theses and PhD. Dissertation.

Shahnazari, H., Tutunchian, M.A., Mashayekhi, M., Amini, A.A., 2012: Application of Soft Computing for Prediction of Pavement Condition Index. Journal of Transportation Engineering 138(12): 1495–1506.

Sharma, D., McGee, D., Kibria, B.M.G., 2011: Measures of Explained Variation and the Base-Rate Problem for Logistic Regression. American Journal of Biostatistics 2(1): 11–19.

Si, J., Feng, Q., Wen, X., Xi, H., Yu, T., Li, W., Zhao, C., 2015: Modeling soil water content in extreme arid area using an adaptive neuro-fuzzy inference system. Journal of Hydrology 527: 679–687.

Smith, D.M., 1993: Effects of Variable Tire Pressure on Road Surfacings. Volume 2: Analysis of Test Results.

Suman, S.K., Sinha, S., 2012: Pavement Condition Forecasting Through Artificial Neural Network Modelling. International Journal of Emerging Technology and Advanced Engineering 2(11): 474–478.

Sun, L., Cai, X., Yang, J., 2007: Genetic algorithm-based optimum vehicle suspension design using minimum dynamic pavement load as a design criterion. Journal of Sound and Vibration 301(1–2): 18–27.

Sundin, S., Braban-Ledoux, C., 2001: Artificial IntelligenceBased Decision Support Technologies in Pavement Management. Computer-Aided Civil and Infrastructure Engineering 16(2): 143–157.

Tabatabaee, N., Ziyadi, M., Shafahi, Y., 2013: Two-Stage Support Vector Classifier and Recurrent Neural Network Predictor for Pavement Performance Modeling. Journal of Infrastructure Systems 19(3): 266–274.

Croat. j. for. eng. 39(2018)2

Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression ... (271–287) Thube, D.T., 2012: Artificial neural network (ANN) based pavement deterioration models for low volume roads in India. International Journal of Pavement Research and Technology 5(2): 115–120.

Tunay, M., 2006: The assessment of environmentally sensitive forest road construction in Calabrian pine forest areas of Turkey. Journal of Environmental Biology 27(3): 529–535.

Wang, H., 2011: Analysis of tire-pavement interaction and pavement responses using a decoupled modeling approach.

PhD. Dissertation, University of Illinois at Urbana-Champaign.

Wang, H., Al-Qadi, I., 2009: Combined Effect of Moving Wheel Loading and Three-Dimensional Contact Stresses on Perpetual Pavement Responses. Transportation Research Record: Journal of the Transportation Research Board 2095: 53–61.

Xu, G., Bai, L., Sun, Z., 2014: Pavement Deterioration Modeling and Prediction for Kentucky Interstate and Highways.

Proceedings of the 2014 Industrial and Systems Engineering Research Conference, 993 p.

M.J. Heidari et al.

Yang, J., 2004: Road crack condition performance modeling using recurrent Markov chains and artificial neural networks. Graduate Theses and PhD. Dissertations.

Yang, J., Lu, J., Gunaratne, M., Xiang, Q., 2003: Forecasting Overall Pavement Condition with Neural Networks: Application on Florida Highway Network. Transportation Research Record: Journal of the Transportation Research Board 1853: 3–12.

Yee, C.S., Roelofs, T.D., 1980: Planning forest roads to protect salmonid habitat. General Technical Report PNW 1, 9 p.

Zhang, H., Keoleian, G.A., Lepech, M.D., 2013: NetworkLevel Pavement Asset Management System Integrated with Life-Cycle Analysis and Life-Cycle Optimization. Journal of Infrastructure Systems 19(1): 99–107.

Zhang, H., Lepech, M.D., Keoleian, G.A., Qian, S., Li, V.C., 2010: Dynamic Life-Cycle Modeling of Pavement Overlay Systems: Capturing the Impacts of Users, Construction, and Roadway Deterioration. Journal of Infrastructure Systems 16(4): 299–309

Authors’ addresses:

Received: April 09, 2017 Accepted: February 27, 2018

Mohammad Javad Heidari, MSc. e-mail: javad.hedari@gmail.com Assoc. prof. Akbar Najafi, PhD. * e-mail: a.najafi@modares.ac.ir Assist. Prof. Seyedjalil Alavi, PhD. e-mail: j.alavi@modares.ac.ir Tarbiat Modares University Faculty of Natural Resources & Marine Sciences P.O.Box: 14115-111 Tehran IRAN * Corresponding author

Croat. j. for. eng. 39(2018)2 287

Original scientific paper

Current State and Improvement Potential of Forestry Workers Training in Croatia Matija Landekić, Ivan Martinić, Matija Bakarić, Tibor Pentek, Tomislav Poršinsky, Mario Šporčić Abstract This paper discusses the key issues of forestry workers training in Croatia, especially dealing with the providers of vocational training, their profile, training procedures and measures necessary for training improvement. A combined approach of literature review, internet search and questionnaire of training providers was applied in order to collect data on training programs conducted in Croatia. The research was conducted during 2016, and it included 94 legal entities authorized for occupational safety training in the Republic of Croatia, with respect to safe working practice training and vocational training for operating machinery (chainsaw and/or skidder). The analysis used basic descriptive statistics.

Research results showed that 30.85% of the analyzed legal entities provide only training for safe working practice, 15.96% provide both trainings – safe work practice and vocational training for operating machinery, 5.32% of the analyzed entities provide only vocational training for operating machinery, 31.91% do not carry out any form of training in forestry, while 15.96% refused to answer questions. On the other hand, 15.56% of the legal entities, which do not carry out any training or did not answer these questions, have on their official website services posted for vocational training in operating machinery (chainsaw and/or skidder). The key findings of the conducted research have pointed out the great heterogeneity amongst providers of forestry workers training, and certain reductions or limitations in the current training programs, both from the aspect of duration of the theoretical and practical training, and the use of non-transparent criteria and standards in the assessment of training.

As an example of successful solution in forestry workers training, European Chainsaw Standard model (ECS) is shortly presented in the paper. Discussion and conclusion sections provide an overview of legislative and organizational requirements for the application of previously developed European model (ECS) in developing the certification system for training of forestry workers in Croatia.

Keywords: forestry, chainsaw worker, health and safety, vocational training, certification, Croatia

1. Introduction Forestry work is one of the most dangerous occupations with significant human and financial losses (ILO 1981, Bentley et al. 2005, Potočnik et al. 2009, Lindroos and Burström 2010, Adams et al. 2014, Marenče and Krč 2016, Potočnik and Poje 2017). This is confirmed by research of (Driscoll et al. 1995, Bell 2002, Lefort et al. 2003, Cabeças 2007), which classifies forestry profession into the high-risk sector category, based on injury rates which are much higher than in

other sectors. A more detailed analysis of accidents in the forestry sector has shown that the work on wood utilization is far more dangerous than other forestry operations (ILO 1991). Relevant indicators on the level of safety between forestry and other sectors show even greater numerical disproportion when the workforce is not adequately qualified (Klun and Medved 2007) for forestry operations.

The safety level of forestry work has immense benefit from the implementation of professional training for forestry workers (Smith and Thomas 1993). Statistics

Croat. j. for. eng. 39(2018)2 289

M. Landekić et al.

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

confirms that trained workers are less injured (Axelsson 1998, Stadlmann 1997) and rarely fatally hurt. At the very beginning of training, participants realize its importance and thus develop a culture of safety, starting from the constant use of personal protective equipment to more advanced security issues, such as the use of appropriate techniques to cut down the wind damaged trees on sloping ground, etc. (Tsioras 2012). In most European countries regulations oblige employers to provide appropriate training to each person using work tools and machines (Medved 1998). Also, through knowledge acquired by training, work techniques and skills need to be confirmed by a mandatory inspection and certification as evidence of acquired professional knowledge and skills for a safe forestry work (Martinić et al. 2011).

In the context of the current trends in training, it is important to emphasize modularization and introduction of training specialization where chainsaw workers and forest machinery operators represent important and often targeted group. Greater attention is given to entrepreneurship and economic aspects in the design of educational programs (Bernasconi and Schroff 2011). The content of courses is adapted to the social and technical achievements with a stronger emphasis on language skills, improving the entrepreneurial qualifications, increasing competence in the field of machinery and logistics, as well as expansion of methodological skills (Bernasconi and Schroff 2011).

In parallel, the need for certification of training programs, and thereby also testing the skills acquired during training gained in importance. In Germany, the old »Waldarbeiterschule« (forest workers training school) has evolved into forestry training center, and introduced professional licenses i.e. professional certificates for education and training, for example, in Germany, Austria, France (Bernasconi and Schroff 2011).

Today, vocational training is a basic requirement of modern forestry, where traditional forestry chainsaw worker/machinery operator can be translated into a competent associate in the management of forest resources. The complexity and specialization of forestry worker tasks have resulted in a general trend of increasing expansion of the expected skills and, accordingly, the training programs in the forestry sector.

In accordance with the stated, this paper discusses the key issues of forestry workers training in Croatia, especially dealing with the providers of safe working practice training, and vocational training in operating machinery (chainsaw and/or skidder), their services and training procedures. Secondly, the paper presents the concept and implementation process of the European Chainsaw Standard model as an example of

290

good practice where, through forestry training centers, training knowledge and skills must be confirmed by the mandatory inspection and certification. Third, based on the conducted analysis, the paper also elaborates and proposes measures necessary for improvements in the Croatian forestry workers training system.

1.1 Legal framework of forestry workers training in Croatia Requirements and framework conditions of training in European forestry have significantly changed in the last 15 years. The three main areas of change can be seen through: a) change in the socio-economic environment affecting the content of training, b) change in the technological environment with impacts on the use of labor and subjects of specialization and c) change in the systems of training through fundamental reforms, e.g. Copenhagen Process (Bernasconi and Schroff 2011). Copenhagen Process promotes mutual trust, transparency and recognition of competences and qualifications in order to increase mobility and facilitate access to lifelong learning for vocational education and training across Europe. Progressive processes of internationalization and globalization had an additional impact on the above-mentioned changes.

In operational forestry of the Republic of Croatia, the requirements of the legal framework oblige employers to train chainsaw operators/machinery operators (Fig. 1): Þ for safe work practice Þ for operating machinery (chainsaw and/or skidder, forwarder, etc.).

Article 27 and 28 of the Occupational Safety and Health Act (Official Gazette No. 71/14 and 118/14) obliges the employer to train workers to work in a safe manner on the basis of a risk assessment, as follows: 1) before starting work for the first time; 2) implementing changes in the working method; 3) introducing new work equipment or changes; 4) introducing new technology; 5) before assigning a worker to a new job or a new workplace and 6) in response to health damages caused by danger, harmfulness or strain working conditions. In accordance with the Ordinance on Certification of Occupational Safety and Health activities (Official Gazette No. 84/15), the above mentioned training can be conducted by a physical or legal person and by employers for their own use if they are authorized by the Minister of Labor (Ministry of Labor and Pension System) and if they meet the conditions prescribed by the Ordinance on Occupational Safety and Health Training and Professional Examination (Official Croat. j. for. eng. 39(2018)2

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

M. Landekić et al.

ing is not strictly defined by legal regulations. The practical part of training is usually defined in the company of training participants according to a special schedule.

Fig. 1 Entities authorized for forestry workers training in Croatia Gazette No. 112/14) (Fig. 1). The aim of the training is to acquire knowledge and skills, and is carried out under the Plan and Training Programs derived from the content of hazard and risk assessment at workplaces. The Ordinance on Occupational Safety and Health Training and Professional Examination (Official Gazette No. 112/14) stipulates the duration of the theoretical training – at least 7 school hours (sh), the training being carried out by a health and safety expert. On the other hand, the duration of the practical training is not prescribed by the Ordinance, and the training process is carried out a) by the authorized person who directly manages the workers training and b) by health and safety expert who is responsible for the workers training.

The vocational training of forest operators for work with mechanized machinery (e.g. a chainsaw, skidder, etc.) is governed by Article 3 of the Ordinance on Jobs with Special Working Conditions (Official Gazette No. 71/14). The basic requirements for the vocational training are primary school education, age (over 18), and physical and mental ability to perform the work operations. The training program is conducted by an adult vocational school under the authority of the Ministry of Science and Education, and with the prior approval of the Croatian Agency for Vocational Education and Training of the Vocational Training Program (Fig. 1). Organizational form of training for chainsaw and/or skidder operators generally contains regular (R) or consultative-instructive (C-I) teaching, which consists of a theoretical and practical part. The extent and duration of theoretical and practical train-

In accordance with the above legal framework, safety precautions and safety rules for professional forestry work in Croatia are prescribed by the Ordinance on Occupational Safety and Health in Forestry (Official Gazette No. 10/86). On the other hand, according to the Forest Law (Official Gazette 94/14), nonprofessionals (the local population), who are trained in harvesting operations (safe work with a chainsaw), are allowed to buy and process for their own purposes up to 20 m3 of firewood per year (self-processing), with the permission and under the supervision of the person who manages these forests. In state forests, the above mentioned self-processing works reach up to 650,000 m3 per year (Annual Business Report CF Ltd. 2015). Due to the allowed volume of self-processing (20 m3), there are thousands of potential actors who need to be trained in safe work and appropriate use of forestry machines, all according to verified training programs. Deficiency of legal provisions gives an unclear picture of the level and type of qualifications that must be met by non-professionals, as well as of the qualifications of the supervisory staff when the harvesting works are carried out in private forests (Landekić et al. 2017).

Based on the above, the following sections of the paper provide information to forestry experts and practitioners about a) the state of professional training of forestry workers in Croatia and b) trends in the European model of forestry work certification.

2. Materials and methods Collecting of research data was carried out by using a telephone survey with a pre-prepared questionnaire. The advantages of such approach are high accessibility, good quality control, quick data processing, etc. The telephone survey was conducted by a student of the Faculty of Forestry in Zagreb, wherein questioning of the legal entity lasted on average 5 minutes. A specific survey method is used for collecting data either from the general population or from a target population. In this case, the target population, the sampled one, consisted of 94 legal entities authorized for occupational safety training in the Republic of Croatia, with respect to a) safe working practice training and b) vocational training for operating machinery (chainsaw and/or skidder). The list of legal entities was taken from the official register of the Croatian In-

Croat. j. for. eng. 39(2018)2 291

M. Landekić et al.

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

stitute for the Improvement of Occupational Safety, and the research response rate was 84%.

The questionnaire used was developed by the Department of Forest Engineering at the Faculty of Forestry, Zagreb University, and it consisted of three structural parts. The first part of the questionnaire refers to general characteristics of legal entities providing training for forestry workers with three questions.

The second part refers to safe working practice training where, through four questions, participants need to give answers regarding the structure of the training program (duration of theoretical and practical part in school hours, where one school hour is 45 minutes), practical execution and assessment of the training. The third part refers to vocational training for operating machinery (chainsaw and skidder) where, through seven questions, participants also need to give answers regarding the structure, practical execution and assessment of the training. The research was conducted during autumn 2016, where parallel to the questionnaire, on-line website search and analysis of legal entities were conducted to determine their profile and services they provided. The above stated on-line search was also made to compare the accuracy of the key information obtained through the questionnaire.

Methods of analysis and synthesis, comparison and compilation were used in the processing of data and in drafting this paper. Analysis and synthesis method was used in drafting this paper, where various sources were ultimately summarized in a single text.

Method of comparison was used in the practical part of the work to show the frequency of legal entities in categories defined according to theoretical and practical training as well as to display the implementation of the European Chainsaw Standard (ECS) for the member states of the European Forestry and Environmental Skills Council (EFESC). The input data basis, systematization and assessment of the entry and primary processing of the collected data were made in the software package Microsoft Excel. Theoretical part of the work was made using the compilation method.

Additionally, the results of Figs. 3, 4 and 5 are explained in terms of school hours for the theoretical and practical part of the training. In the legend of Fig. 3, 4 and 5, four time-defined categories for the theoretical part of the training are shown with the number of legal entities in brackets (e.g. in Fig. 3, there are 27 legal entities or 61.36% in the category 7 to 16 school hours of theoretical training). Five time-defined categories for the practical part of the training are shown on the x-axis of Fig. 3, 4 and 5, where 27 legal entities are in the category from 7 to 16 sh of theoretical training (Fig. 3), 20 legal entities provide less than 8 sh of prac-

292

tical training, two legal entities provide 9 to 16 sh, one legal entity provides 17 to 24 sh and four legal entities provide more than 33 hours of practical training.

3. Results The research conducted on 94 legal entities, authorized for occupational safety training, included the situation and procedure analysis of forestry workers training in Croatia. Research results in Table 1 show that 30.85% of the investigated legal entities provide only training for safe working practice, 15.96% provide both trainings – safe work practice and vocational training for operating machinery, 5.32% of the investigated entities provide only vocational training for operating machinery, 31.91% do not carry out any form of training in forestry, while 15.96% refused to answer questions. On the other hand, 15.56% of the legal entities, which do not carry out any training or refused to answer this question, have on their official website services posted for vocational training in operating machinery (chainsaw and/or skidder). Also, from 29 legal entities (Table 1), which carry out only safe working practice training, 13.79% have on their official website services posted for vocational training in operating machinery.

Research results show that 49 legal entities carry out a certain type of training in forestry (Table 1 and Fig. 2). From the total number of active legal entities, 81.63% have one or two branch offices (Fig. 2), employ on average 11 to 15 employees and half of employees are specialized in the forestry sector. 3 to 6 branch offices within national borders have 16.33% of legal entities Table 1 Number of legal entities versus type of provided training Category Safe working practice training

Answers from

Information from

telephone survey official website* 29 (30.85%)

4 (13.79%)

15 (15.96%)

–

5 (5.32%)

–

No training in forestry

30 (31.91%)

4 (13.33)

Refused to answer

15 (15.96%)

3 (20%)

In total

94 (100 %)

11 (11.7%)

Safe working practice training and vocational training in operating machinery Vocational training in operating machinery

* related to providing services for vocational training in operating machinery (chainsaw and/or skidder)!

Croat. j. for. eng. 39(2018)2

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

M. Landekić et al.

Fig. 2 Profile of firms providing workers training in forestry (N=49)

Fig. 3 The structure of safe working practice training (N=44)

employing on average 11 to 21 employees (Fig. 2).

Only one company has 30 branch offices and employs 80 workers, of which 75.00% are specialized for training in the forestry sector (Fig. 2).

To the question »Who conducts the training for safe working practice«, 90.90% of legal entities pointed out that the training program was carried out by their own employees, 4.55% stated that the training was carried out by their own employees in collaboration with external partners (colleges, universities, etc.), and only 4.55% pointed out that the training program was conducted only by external collaborators. On the other hand, to the question »Who performs final assessment of participant training«, 86.36% of legal entities pointed out that final assessment was carried out by an employee of the company i.e. examiner, and 6.82% claimed that final assessment was carried out by an employee of the company i.e. company’s examiner in cooperation with external examiner. Also, 6.82% of legal entities indicated that for the practical part of the training, the final grade was given directly by the employer’s trustee on site where the practical training was carried out, and for the theoretical part of the training, the final grade was given by the employee of the company who carried out the training for safe working practice.

3.1 Situation and procedure analysis of forestry workers training in Croatia A more detailed analysis of the duration of theoretical and practical part of training for ensuring safety at work and in operating machinery (chainsaw and/ or skidder) within the time-defined category is shown in Figs. 3, 4 and 5. In addition, for each type of training, the frequency response is given for the question »Who conducts the training program« and »Who performs final assessment of participant training«.

3.1.1 Safe Working Practice Training In the present research of the training legal entities, it was established that, out of 94 duly authorized legal entities, 44 legal entities conducted training of workers for safety at work in the forestry sector (Table 1 and Fig. 3). Frequency response to the question on duration of safe working practice training (Fig. 3) shows that 61.36% of legal entities provide 7 to 16 school hours of theoretical training, while practical training consists of 8 school hours or less. 18.18% of legal entities are classified in the category of 6 school hours or less (Fig. 3) of theoretical training, and 8 school hours or less of practical training. On the other hand, 13.64% of legal entities stated to provide more than 17 hours of theoretical training and more than 33 hours of practical training (Fig. 3).

3.1.2 Vocational training for operating forestry machinery Another aspect of the training is work with forest machinery where, out of 94 duly authorized legal entities, 20 legal entities carry out training of workers to operate with chainsaws (Fig. 4), and 19 of them also carry out, at the same time, the training of workers to operate with skidders (Fig. 5). Frequency response to the question on duration of vocational training in

Croat. j. for. eng. 39(2018)2 293

M. Landekić et al.

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

Fig. 4 The structure of vocational training in operating chainsaws (N=20)

Fig. 5 The structure of vocational training in operating skidder (N=19)

operating chainsaws (Fig. 4) shows that 45% of legal entities provide more than 17 school hours of theoretical training, while for a significant number of legal entities the practical part of the training consists of 33 school hours and more. For 35% of legal entities, the theoretical training consists of 17 to 32 school hours (Fig. 4), whereas practical training hours are heterogeneously grouped within the defined categories. Only 15% of legal entities pointed out that the theoretical part of the training for operating chainsaws consisted of less than 16 school hours (Fig. 4).

At vocational training in operating skidders, 73.68% of legal entities stated to provide more than 17 school hours for theoretical training and more than 33 hours for practical training. Only 15.79% of legal entities stated to provide less than 16 school hours for theoretical training (Fig. 5).

The credibility of the implementation of the vocational training for operating forest machinery and the final assessment of participants training is shown by a frequency response to the three questions (N=20).

Within the first question »Who conducts the vocational training program«, 60% of legal entities stated that the vocational training program was carried out by external partners (colleges, universities, etc.), and 40% stated that the training was conducted by employees of the company. The second question referred to the realization of practical training with the forest machinery in terms of technical requirements i.e. availability of adequate infrastructure such as polygons, instruc-

tors, mechanization, etc. In accordance with the above, 95% of legal entities pointed out that the practical part of the candidate training was conducted in cooperation with the company that met the technical requirements, and only one legal entity claimed that the company that sent employees on training also issued a certificate to the candidates who completed the practical training. As regards the third question, »Who performs final assessment of participant training«, 80% of legal entities pointed out that the final assessment was carried out by a company examiner, while only 20% stated that the final assessment of the training was carried out by an independent examiner.

294

3.2 European chainsaw standards implementation – example of good practice A positive step forward in the training of forestry workers at the level of Europe occurred in 2009 with the establishment of the European Forestry and Environmental Skills Council (EFESC), whose mission is to simplify the mobility of workers in forestry and arboriculture within the EU through the processes of accreditation and promotion of individual national qualifications between the partner countries at European level. The motive behind the development of EFESC was to develop a universal qualification standard for chainsaw users throughout Europe, with the primary aims to: a) reduce accidents and fatalities, b) reduce associated economic and personal costs related to accidents, c) improve and enhance operator skills Croat. j. for. eng. 39(2018)2

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

and efficiency in the workplace, d) enhance operator mobility and employability throughout Europe, e) improve and expand delivery of existing vocational training and assessment standards, and f) encourage and promote life-long learning and continuous professional development (source: http://www.europeanchainsaw.eu/).

The first minimum standard developed by the EFESC, partially implemented in EU member states, is the European Chainsaw Standard (ECS). It consists of five modules / levels (Fig. 6). The developed standard is not legally binding, but is a set of minimum standards that are considered »best practices« with professional and/or semi-professional use of chainsaws with the focus on the safety aspect of the use of operating machinery. In the implementation of the certification of minimum working competence, i.e. authorization and auditing (control) of national training providers, EFESC relies on the national centers, which issue the accreditation, through the process of accreditation, for the certification of training programs to the institutions registered for that activity. Based on the obtained accreditation, such institutions can use, in their own arrangement, the ECS certificate and/or can use the ECS logo on certificates, if they comply with the criteria set by the EFESC general assembly, and if everything is reviewed and approved by the national agency.

Up to November 2016, nine accredited national centers were established in nine different countries in Europe. Those national centers have different legal organizational forms - from economic associations, training centers and network platforms to ministries and other bodies of state administration. Table 2 shows

M. Landekić et al.

Fig. 6 European chainsaw standards levels the process of applying the concept of universal European standards for operating chainsaws within countries that have national centers accredited by the EFESC. Belgium and Germany have issued the highest number of certificates, followed by Spain, Austria and Romania (Table 2). The lowest number of certificates issued was recorded in Italy and the Netherlands. The only country that has established the national agency, but has not started the process of implementation of the ECS, is the United Kingdom.

According to EFESC, the European certification for chainsaw operator (ECS) is not legally binding either for professionals or non-professionals (amateurs),

Table 2 Implementation of European chainsaw standards within EFESC countries (Landekić et al. 2017) Country Austria

Belgium

France

Italy

Netherlands

Germany

Romania

Spain

Starting year of implementation

2015

2012

2015

2015

2014

2014

2015

2015

Number of Assessment Centers

3

2

12

7

1

2

1

1

Number of certificate holders*

152

645

124

35

83

474

170

118

ECS 1 = 152 ECS 1 = 645 ECS 1 = 124 ECS 1 = 30 ECS 1 = 83 ECS 1 = 474 ECS 1 = 170 ECS 1 = 117 Number of certificates issued

Number of certified assessors

ECS 2 = 152 ECS 2 = 209 ECS 2 = 73 ECS 2 = 9

ECS 2 = 56 ECS 2 = 462 ECS 2 = 170 ECS 2 = 68

ECS 3 = 152 ECS 3 = 31 ECS 3 = 9

ECS 3 = 13 ECS 3 = 15 ECS 3 = 462 ECS 3 = 169 ECS 3 = 9

ECS 4 = 115 ECS 4 = 6

ECS 4 = 0

5

14

ECS 4 = 0 44

18

ECS 4 = 0 5

ECS 4 = 433 ECS 4 = 0 28

1

ECS 4 = 0 4

*holder who has e.g. NTPC (National Proficiency Test Council) recognized industry qualification for chainsaw operator can directly attend the third level of European Chainsaw Standard (ECS 3)!

Croat. j. for. eng. 39(2018)2 295

M. Landekić et al.

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

who use chainsaw occasionally. The standard is consistent with the aspect of work quality and forestry work safety, and the need of implementation of the above standard and certification system for forest chainsaw workers is explained by the following assumptions: Þ transparency of skills and competences throughout detailed description of the required qualification for each level Þ enhancement of operator mobility and employability throughout Europe and presence of uniform certificate that vouches for the holder’s skills, independent of language and country Þq uality and traceability of the licensed certificate holder through a registration at the European organization EFESC (source: http://www.europeanchainsaw.eu/).

4. Discussion Technological advancement and specialization of workers, through various forms of education and vocational training in forestry, is highly beneficial for both workers and employers from the technical, safety and professional aspect (Landekić et al. 2017). On the other hand, as Bernasconi and Schroff (2011) stated, there are certain restrictions for training providers that apply on: a) cost demands of training related to the number of hours of field work, polygon infrastructure and specialized tools, b) limited market, i.e. a small number of potential participants, and c) lack of interest of young generation for the so-called »Black Collar« 3D (Dirty, Dangerous and Demanding) jobs – forest chainsaw operators being classified into this category. The above limitations indicate the reduction of the operating portfolio of forestry workers in near future.

Given the numerous challenges facing the forestry sector of the EU, as an area of highly specialized niche, international cooperation and cluster formation becomes increasingly important, with the focus on transfer of knowledge related to training programs. Development and implementation of universal European standards for chainsaw operators is a positive example of cooperation within the EU countries. In developed European countries, in the past decades, specialized institutions have been established, the so-called »forestry training centers« that provide comprehensive training in the technical, safety and vocational aspect.

Training is carried out through the idea of »dual system«, meaning that a combined approach of theoretical learning and practical training can provide the best education possible. With this approach, a high level of

296

knowledge and skills is achieved relating to the proper use of personal protective equipment, operating machinery, auxiliary equipment and procedures in compliance with safety operating procedures. Medved (1998) and Martinić et al. (2011) stated that training and periodical checking of competences of forestry machinery operators have been key factors for work quality and safety in operational forestry in Western Europe in the past decades. Training knowledge and skills are confirmed by the mandatory inspection and certification (e.g. European Chainsaw Certification) as evidence of certain professional knowledge and skills for safe forestry work, which is highly beneficial for both the employee and the employer (Martinić et al. 2011).

Previous research (Martinić et al. 2011, Martinić and Landekić 2012, Landekić et al. 2016, Landekić et al.

2017), dealing with the training of workers in the forestry sector of Croatia, has shown a lesser degree of satisfaction. The above is supported by a recent research, according to which 11.70% of legal entities (N=94) provide training services for operating machinery in forestry at the official web site (Table 1), while in the questionnaire, a negative answer is given to this question. In addition, from a total of 94 interviewed legal entities, 15.96% refused to give any information regarding the provision of forestry workers training in Croatia. The results suggest that a significant part of training for safe forestry work in Croatia is carried out by programs that are not adequately verified nor independently evaluated by the relevant institutions (the Agency and the Ministry). Furthermore, the Ordinance on Occupational Safety and Health Training and Professional Examination (Official Gazette No. 112/14) stipulates the duration of the theoretical part of the training for safe work of at least 7 school hours, while Fig. 3 shows that, from a total of 44 legal entities that applied to perform this type of training, 18.18% did not meet the requirements. As for the practical training, 25 legal entities are classified in the category of 8 school hours or less (Fig. 3), and 72% of them admitted that their practical training for safe work in forestry consisted of only one school hour. The analysis of professional training programs for operating machinery in forestry (chainsaw/skidder), according to the defined categories, shows significant diversity in the duration of theoretical and practical training (Fig. 4 and 5). The frequency of answers for the three key questions (Who conducts the training program; Who conducts the practical part of the training; and Who performs final assessment of participants training) resulted in discordance in both types of training. The following deficiencies were identified: lack of clearly defined criteria and of control over the fulfillment of human and technical Croat. j. for. eng. 39(2018)2

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

preconditions/requirements and resources of training holders, as well as lack of transparency in the final assessment of the candidates and questionable quality of the practical training.

5. Conclusions The most acceptable way of increasing safety in the Croatian forestry, for professional workers and amateurs, is the application and implementation of the achievements of the developed forestry countries, where in the sphere of forestry work safety, the implementation of certification for all aspects of forestry work training is highly emphasized. Considering all the above, the accession of the Republic of Croatia to the European process and programs related to health and safety in forestry is possible by implementation of international best practices through: Þ strong promotion of safety system and awareness of all stakeholders in the supply chain of forestry works: licensed forest contractors, educational and medical institutions, equipment manufacturers, professional societies, professional associations, etc. Þ partnership and cooperation of Croatian forestry sector key organizations in selecting the most relevant organization/institution for acquiring authority of the national accreditation center for the implementation of the European Chainsaw Certificate program, which includes the adoption of a uniform curriculum for the particular type of training Þ amendments and/or preparation of new regulations (acts, ordinances, regulations) to ensure standardization of training for forestry work, including the establishment and functioning of the national accreditation center Þ verification by the national accreditation center of optimally two regional forestry centers for training and evaluation (one for eastern and central part and the other for pre-mountain and Mediterranean part of Croatia) with adequate infrastructure for theoretical and practical training, verified programs, lecturers, trainers, assessors and others.

To conclude, the above measures for the improvement of health and safety of forestry work should be elaborated in detail and implemented through the legislative, structurally organized and educational concept, by using a multidisciplinary research team and a special operational working group in the framework of the national project relevant for the forestry

M. Landekić et al.

sector. This should be a key measure in the field of economy and sustainability of forest production, contained in the National Development Strategy of Forestry and Wood Technology in Croatia.

Acknowledgements Authors are grateful to the CERCA program of the Generalitat de Catalunya for support in the paper development and EFESC national agencies for the provided data related to the implementation of the European Chainsaw Standard.

6. References Adams, G., Armstrong, H., Cosman, M., 2014: Independent forestry safety review – An agenda for change in the forestry sector. http://fica.org.nz/wp-content/uploads/2014/10/ IFSRReportSummary-Web.pdf, 12 p.

Annual Business Report CF Ltd., 2015. ‹http://portal.hrsume. hr/images/stories/godisnja-poslovna-izvjesca/godisnje_izvjesce_za_2015.pdf› [Accessed on December 20, 2016.]

Axelsson, S.Å., 1998: The mechanization of logging operations in Sweden and its effect on occupational safety and health. International Journal of Forest Engineering 9(2): 25–31.

Bell, J.L., 2002: Changes in logging injury rates associated with use of feller-bunchers in West Virginia. Journal of Safety Research 33(4): 463–471.

Bentley, T., Parker, R., Ashby, L., 2005: Understanding felling safety in the New Zealand forest industry. Applied Ergonomics 36(2): 165–175.

Bernasconi, A., Schroff, U., 2011: Professions and Training in Forestry. Results of an Inquiry in Europe and northern America. Federal Office for the Environment, Bern, 84 p.

Cabeças, J.M., 2007: An approach to health and safety in E.U. forestry operations – Hazards and preventive measures. Enterprise and Work Innovation Studies 3: 19–31.

Driscoll, T., Ansari, G., Harrison, J., Frommer, M., Ruck, E., 1995: Traumatic work reated fatalities in forestry and sawmill in workers in Australia. Journal of Safety Research 26(4): 221–233.

European Forestry and Environmental Skills Council (EFESC): European Chainsaw Standard (ECS). Available at: http://www.europeanchainsaw.eu/ International Labour Organisation (ILO), 1981: Occupational Safety and Health problems in the Timber Industry. Geneva, Switzerland.

International Labour Organization (ILO), 1991: Occupational Safety and Health in Forestry. Report II, Forestry and Wood Industries Committee, Second Session. Geneva, Switzerland.

Croat. j. for. eng. 39(2018)2 297

M. Landekić et al.

Current State and Improvement Potential of Forestry Workers Training in Croatia (289–298)

Klun, J., Medved, M., 2007: Fatal accidents in forestry in some European countries. Croatian Journal of Forest Engineering 28(1): 55–62.

Landekić, M., Martinić, I., Šporčić, M., Bakarić, M., 2016: Work technique and safety at work on trees in urban areas.

Proceedings of 6th International Professional and Scientific Conference »Occupational Safety and Health«. September 21–24, 2016, Zadar, Croatia, 362–374.

Landekić, M., Martinić, I., Bakarić, M., Ricart, R.M., Šporčić, M., 2017: Vocational Training of Workers in the Forestry Sector – the Situation in Croatia and Trends in Europe. Šumarski list 140(7–8): 395–407.

Lefort, A.J., de Hoop, C.P., Pine, J.C., 2003: Characteristics of injuries in the logging industry of Louisiana, USA: 1986 to 1998. International Journal of Forest Engineering 14(2): 75– 89.

Lindroos, O., Burström, L., 2010: Accident rates and types among selfemployed private forest owners. Accident Analysis Prevention 42(6): 1729–1735.

Marenče, J., Krč, J., 2016: Possibilities of using small tractors for forestry operations on private property. Croatian Journal of Forest Engineering 37(1): 151–162.

Martinić, I., Landekić, M., Šporčić, M., Lovrić, M., 2011: Forestry at the doorstep of EU – How much are we ready in the area of occupational safety in forestry? Croatian Journal of Forest Engineering 32(1): 431–441.

Martinić, I., Landekić, M., 2012: Evaluation of the possibilitie and conditions for inclusion in the European network of forestry work certification. Elaborate – Defining models for the implementation of certification in the Republic of Croa-

tia including the legislative, organizational and financial framework of implementation, 1–45.

Medved, M., 1998: Nezgode in tveganje pri poklicnem in nepoklicnem delu v gozdu. Gozdarski Vestnik 56(9): 379– 389.

Official Gazette No. 94/14: Law on Forests Official Gazette No. 84/15: Ordinance on authorization for occupational safety and health works Official Gazette No. 71/14: Ordinance on jobs with special working conditions Official Gazette No. 10/86: Ordinance on occupational safety and health in forestry Official Gazette No. 112/14: Ordinance on occupational safety and health training and professional examination Potočnik, I., Pentek, T., Poje, A., 2009: Severity analysis of accidents in forest operations. Croatian Journal of Forest Engineering 30(2): 171–184.

Potočnik, I., Poje, A., 2017: Forestry ergonomics and occupational safety in high ranking scientific journals from 2005– 2016. Croatian Journal of Forest Engineering 38(2): 291–310.

Smith, L.A., Thomas, R.E., 1993: Ergonomics research in the southern United States. Unasylva 44(172): 38–44.

Stadlmann, H., 1997: The accident situation in austrian forestry, with particular reference to farm forests. In the proceedings of the seminar »Safety and Health in Forestry are feasible!«, Konolfingen, October 6–11, 1996, BUWAL, Berne, 127–131.

Tsioras, P.A., 2012: Promotion of safety in forest operations.

International Virtual Conference on Advanced Research in Scientific Areas. December 3–7, 2012, Slovakia, 1395–1399 p.

Authors’ addresses:

Received: October 27, 2017 Accepted: February 12, 2018

298

Assist. prof. Matija Landekić, PhD. * e-mail: mlandekic@sumfak.hr Prof. Ivan Martinić, PhD. e-mail: imartinic@sumfak.hr Matija Bakarić, PhD. e-mail: mbakaric@sumfak.hr Prof. Tibor Pentek, PhD. e-mail: pentek@sumfak.hr Prof. Tomislav Poršinsky, PhD. e-mail: porsinsky@sumfak.hr Prof. Mario Šporčić, PhD. e-mail: sporcic@sumfak.hr Forestry Faculty of Zagreb University Department of Forest Engineering Svetošimunska 25 10 000 Zagreb CROATIA * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests Karlo Beljan, Stjepan Posavec, Jura Čavlović, Krunoslav Teslak, Thomas Knoke Abstract Economic analysis of even-aged fir stand management was illustrated using the example of the forests of the Croatian Dinaric region, as well as their transformation into more stable unevenaged structures. Two scenarios (even-aged, uneven-aged) were simulated against the backdrop of the existing forest stand structure of future forest stand management during a 140-year period using forest growth modeling software MOSES version 3.0 in order to identify economic differences amongst different scenarios both at stand level and at forest level. The research included forest management analysis throughout the transformation period and subsequently the continuation of balanced state forest management. Moreover, the research also provided the opportunity of forest purchase within the price range from 1000 to 12,500 EUR/ha, amid assumed fluctuation of selling prices of timber assortments throughout the simulation period. Discount rates from 1% to 5% were used during the economic analysis. The research findings showed that, according to harvesting costs, Net Present Value and Internal Rate of Return, uneven-aged forest management system, including the transformation period, achieved superior economic results, albeit at discount rates that exceeded 1.24%. The conclusion was reached that, according to all economic criteria, uneven-aged mixed silver fir-beech management system is preferred compared with the pure even-aged silver fir management.

Keywords: silver fir, management transformation, cost control, NPV, IRR

1. Introduction Silver fir (Abies alba Mill.) in the Dinaric region primarily includes Slovenia, Croatia and Bosnia and Herzegovina, as well as to a lesser extent Albania, Montenegro and Serbia (Bončina 2011). The notion of uneven-aged management in Croatia normally refers to silver fir (Matić et al. 1996). This management system is normally adopted in fir-dominated Dinaric mountain forests (Bončina 2011). During the last several decades such forests have been characterised by processes of growing stock of large trees, difficulties in regeneration and gradual silver fir dieback, which were recorded in Croatia (Čavlović et al. 2006, Teslak et al. 2016), Bosnia and Herzegovina (Keren et al. 2014), Slovenia (Bončina et al. 2002, Ficko et al. 2011) and other Central European countries.

management is dominant (Čavlović et al. 2006). Several types of them are present depending on the soil type, altitude, bioclimatic zone and consequently the plant community. In Croatia, pure even-aged forest stands are to be found only in several localities of silver fir forests in the Pannonian and Dinaric regions. The previously mentioned forests within the Pannonian Region are to be found in the western part of Papuk mountain (Šafar and Hajdin 1954, Božić et al. 2011), whereas in the Dinaric region they are located in the lower parts of Velebit and Mala Kapela mountain by the rivers of Gacka and Lika (Vukelić 2000). It is important to highlight that their origin has thus far remained unknown, yet it can be assumed that they were formed through artificial regeneration. Rare cases of successful evenaged silver fir forest management, i.e. in the Italian Alps, have been recorded to date (Bottalico et al. 2014).

Silver fir forests in Croatia account for 12% in the total forest area (Čavlović 2010), in which uneven-aged

During the past 100-odd years, several silver fir forest management systems were adopted in the Dinaric

Croat. j. for. eng. 39(2018)2 299

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

Mountain Region, from even-aged to uneven-aged and plenter (or selection) systems and »freestyle forest« management (Bončina 2011). There are both advantages and disadvantages to even-aged and uneven-aged management systems, yet the general trend in forestry, both in Europe and throughout the world, is to emphasise the advantages of the management system referred to as natural or close-to-nature forest management system (McMahon 1999, O’Hara 2001, Macdonald et al.

2010, Davies and Kerr 2011, Knoke 2012). Uneven-aged forest management system is considered as a close-tonature forest management system due to its features of maintenance of continuous cover stands (Franklin 1989, Mlinšek 1996, Koch and Skovsgaard 1999) primarily for shade tolerant species, yet this is true only in case of observation of small size stands.

There are multiple justifications for the transformation of even-aged coniferous stands into uneven-aged or selection i.e. plenter forests (Hanewinkel 2001). In addition to higher silvicultural and management possibilities (Kenk and Guehne 2001, Malcolm et al. 2001, Schütz 2001, Remeš 2006, Božić et al. 2011), as well as economic potential of uneven-aged forests (Knoke and Plusczyk 2001, Hanewinkel 2002, Price 2002), there is an increasing need for transformation due to a large number of the demands placed by the society upon forests (Salim and Ullsten 1999, Buongiorno 2001, O’Hara 2001). In the context of changing importance of forests and the demands placed upon forests, forest management systems with a rising number of features of adaptability and dynamism are becoming increasingly prominent, whilst the rigid pre-defined framework over a long period is slowly being abandoned.

Forest management system transformation is considered an immense turning point, which is also evident in economic consequences (Davies and Kerr 2015), and hence the only task of forest economics is to analyse forest management transformation processes. Hanewinkel (2002) stated that this type of analysis can be conducted only in silver fir or spruce forests, due to the fact that in such forests both forest management systems can be applied. The same author analyses the European references which, whilst comparing even and uneven-aged management of sprucefir forests, from an economic standpoint clearly gives precedence to uneven-aged management system, whereas the even-aged forest management system is considered more adequate only in rare cases. This is due to a large quantity of valuable logs, a smaller risk of natural disasters, smaller forest management costs (Hanewinkel 2002), as well as regeneration, which is cost free in case of natural uneven-aged forest management (Navarro 2003, Davies and Kerr 2015).

300

It is a fact that, upon transformation of artificially grown forests, regeneration costs play an important role concerning the economic effects (Davies and Kerr 2015). Knoke and Plusczyk (2001) compared evenaged and uneven-aged forest management (including the transformation) of spruce stands. The unevenaged forest transformation process in the previously described case (Knoke and Plusczyk 2001) provides a smaller amount of financial revenue, albeit it is a constant financial revenue, as well as a higher NPV. A similar conclusion was also reached by Pukkala et al.

(2010) on spruce stands in Finland. However, transformation into uneven-aged forests is not always the best scenario from an economic standpoint and the specific situation needs to be considered (volume, increment, regeneration, intensity of previous forest gap dynamics, etc.) and cost-effectiveness needs to be analysed (Price 2012).

In cases of balanced even or uneven-aged stands, the transformation is never recommended from an economic standpoint (Price 2012) and the same applies to cases in which the age of even-aged stands is close to financial maturity (Knoke 2012). Moreover, silvicultural treatments of transformation and discount rates can affect the economic result in such proportions that a specific management system is more adequate or less favourable in relation to another, as was shown by, for instance, Price (2002). According to Hanewinkel (2001), the most economically cost-effective is transformation through cutting in circular gaps and the gradual expansion of these gaps, whilst Price and Price (2006) showed the cutting of the biggest trees in the stand as more advantageous. Macdonald et al.

(2010) concluded that transformation of coniferous forests through clearcutting in circular gaps initiates the creation of side branches that reduce the financial value of timber assortments.

Even-aged silver fir forests in the Dinaric region in Croatia are considered specific in relation to the dominant approach of a combination of single stem and group selection silvicultural systems. An additional feature of such forests is unbalanced age-class structure with prevalent rate of mature (100-year) stands, which results in decision-making dilemmas concerning the future forest management. Starting from the previously mentioned facts and the above-mentioned published research studies, economic analysis can be conducted through application of even-aged forest management and transformation of even-aged to uneven-aged management, which was considered an interesting and significant research task. Hence, two research hypotheses were formulated: Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Karlo Beljan et al.

Table 1 Structural elements of the sampled stand (source: Beljan 2015)

m3/ha

Growing stock

Basal area

pcs./ha

Stand density

m3/ha

Growing stock

m2/ha

Basal area

pcs./ha

Stand density

m3/ha

m2/ha

S

Other hardwoods

Growing stock

m2/ha

Basal area

pcs./ha

Stand density

m3/ha

Common beech Growing stock

m2/ha

Basal area

pcs./ha

DBH, cm

Stand density

Silver fir

10.0–19.9

66

1.21

7.58

9

0.13

0.81

2

0.02

0.15

77

1.36

8.54

20.0–29.9

56

2.88

31.16

2

0.08

0.92

2

0.07

0.62

60

3.03

32.70

30.0–39.9

78

7.78

104.11

1

0.05

0.63

1

0.04

0.37

80

7.87

105.11

40.0–49.9

93

14.70

213.47

1

0.07

1.09

1

0.10

1.44

95

14.87

216.00

50.0–59.9

52

11.92

179.13

–

–

–

–

–

–

52

11.92

179.13

60.0–69.9

11

3.52

52.98

–

–

–

–

–

–

11

3.52

52.98

70.0–79.9

2

0.52

7.94

–

–

–

–

–

–

2

0.52

7.94

S

358

42.53

596.37

13

0.33

3.45

6

0.23

2.58

377

43.09

602.40

Þ a period of transformation from even-aged into uneven-aged stand structure would result in superior economic effects compared with effects of establishment of even-aged balanced ageclass forest Þ management of an established balanced uneven-aged fir-beech forest would be more economically effective in relation to management of establishment of balanced even-aged pure silver fir forest.

For that purpose the objective of this paper was to use a specific example of an even-aged fir forest in order to simultaneously conduct a research and an economic analysis of the two scenarios: even-aged and uneven-aged forest management, to be implemented in the future both at stand and forest level.

2. Material and methods 2.1 Research area and data collection The research was conducted in the Dinaric region covered by beech-fir forests in the Republic of Croatia in the forest of pure silver fir (Abies alba Mill.) on evenaged stands. Coverage area of the pure (>90% of total volume) silver fir forest (management unit Škamnica 44°58’N, 15°08’E) was 567.33 ha. The stands in this specific forest were extremely similar in terms of age, growing stock, basal area and mean annual volume increment. The altitude ranged between 430 and 828 m a.s.l. The soil types were primarily limestone and

dolomite. According to Köppen’s classification, the climate was of Cfwbx type, i.e. it was classified as moderately warm rainy climate. The average annual air temperature was 9.3 °C. From a total of 33 stands, one specific stand was selected in order to conduct a field measurement. First, a group of stands was defined, in which intermediate cut had just been performed and where the rate of silver fir exceeded 90% (18 stands). Out of these 18 stands, the sample stand intended for field measurement was selected by random sampling. Terrain survey in the selected stand, covering an area of 22.21 ha, was conducted in the year 2013 (Table 1). A total of 20 circular plots with 40-metres in diameter were placed within a 100-meter square net oriented towards four cardinal directions. All the trees with DBH exceeding 10 cm were measured. For every tree, its species, the specific position in the three-dimensional space, DBH, and absolute height were determined according to Čavlović and Božić (2008).

2.2 Simulation of stand growth and management scenarios A virtual square shaped stand (173.2x173.2 m) was designed (Beljan et al. 2016) based on both measured and assessed data of the observed stand (Table 1). This virtual stand was integrated into a single-tree growth computer programme for stand growth simulation model MOSES version 3.0 (Pretzsch et al. 2002, Steinmetz 2003, Hasenauer 2006, Mikac et al. 2013) as the

Croat. j. for. eng. 39(2018)2 301

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

initial state for the simulation of management scenarios. Three assumptions were made: Þ the stand structure represented 10% of the forest where cut had just been performed (past management characterised by 10-year cutting cycles and some kind of intermediate cut) Þ the first 10-year period of stand growth simulation was without cut Þ t he use of the same initial state for other nine tenths of the forest (groups of stands) where stand growth simulation was to start sequentially during the following 2–10 years.

Stand growth and future management were simulated for both even-aged and uneven-aged management scenarios and also at both stand and forest level.

2.2.1 Even-aged management scenario Even-aged management scenario and the theoretical even-aged forest were defined by the forest area of 567.33 ha, rotation, theoretical number and area of compartments/sub-compartments, regeneration period and prescribing and scheduling both in-

termediate and regeneration cut (Table 2). It is necessary to distinguish »free« simulation period (without cutting, year 2013–2022) and simulation period of management (with cutting, from the year 2023 to the future).

Stand level growth simulation of the sub-compartment (stand) 1a in which the regeneration would be the first to begin, taking place »on time« (at the age of 100). Stand growth simulation of other stands in the compartment 1 (b up to j) would be similar, albeit with a shift of 1 year and sequentially the prolongation of regeneration would start. Concerning the stands within other compartments (2–11), the management (cutting) would start in 2023 (a stands). In that case, an intermediate cut in 10-year cycle would be simulated until the start of the regeneration in several stands depending on their belonging to several compartments (stands belonging to compartment 11 were scheduled for regeneration last (cut stand age of 220 years)). Stand growth simulations of all 110 stands arranged in the previously described way presented the even-aged forest management scenario.

Table 2 Description of elements whereon even-aged management scenario and theoretical even-aged forest were defined Element of even-aged management/forest

Explanation

Forest area

567.33 ha

Forest area of stands in forest where the rate of silver fir exceeds 90%

Rotation

120 yrs.

In the past, according to Meštrović (2001), 120-year rotation was the most frequently used system of silver fir even-aged management in Croatia

Regeneration period

20 yrs.

20 yr. regeneration period by 3 cuts; theoretically first (preparatory) at the age of 100 yr., second (regeneration) at the age of 110 yr. and after 10 yrs. the final cut. New stand growth cycle begins with regeneration (second) cut in the middle of the regeneration period

Sub-compartment (stand)

110 pcs. (5.15 ha)

In case of 120-yr rotation and 20-yr regeneration period 50% of youngest and oldest age class area overlap, in which case theoretical forest is comprised of 110 stands (Nenadić 1929)

Compartment

11 pcs. (51.5 ha)

110 stands are arranged within 11 compartments (each compartment divided into 10 stands, a up to j); stands of compartment 1 would be regenerated first (felling age of 120 yrs), whilst stands of compartment 11 would be regenerated last (felling age of 220 yrs)

Start of simulation

yr 2013 up to yr. 2022 a stands of all compartments in 2013, … , j stands of all compartments in 2022

a stands of all compartments in 2023 (1a regeneration, 2–11a intermediate cut), … , j stands of all Start of management yr 2023 up to yr. 2032 compartments in 2032 (cut) Intermediate/thinning 10 yr. cycle cut

Intermediate cut in stands during period between start of management and first regeneration; thinning in new cycle growth stands (first at the age of 30 yr, last at the age of 90 yr); cut amount estimates according to Matić (1989) as quotient of stand volume and stand age in decades (i.e.: V=675 m3/ha, age=170 y; cut volume=675 m3/ha /17=40 m3/ha)

Regeneration felling

Start first in stand 1a (preparatory in 2023, regeneration in 2033, final in 2043) and last starts after 110 years in stand 11j (preparatory in 2133, regeneration in 2143, final in 2153); estimation of felling amount: one third of current stand volume for preparatory, one half of current stand volume (accumulated last 10-y increment included) for regeneration cut, total remaining stand volume (accumulated last 10-y increment included) for final cut

302

20 yr. period

Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Karlo Beljan et al.

2.2.2 Uneven-aged management scenario Theoretical uneven-aged forest and management scenario was defined by the same forest area of 567.33 ha, selection cut cycle of 10 years, theoretical number of 10 stands (compartments 1–10) each covering an area of 56.73 ha and achievement of balanced selection stand structure with aimed silver fir/beach proportion of 80% vs. 20%.

The management scenario was based on stand level simulation of compartment 1, where simulation started in 2013, whilst the transformation (1th cut) would begin in 2023. Stand growth simulation of other stands (compartments 2 up to 10) was similar, but with a shift of 1 year (the assumption was made that those stands would have similar structure at the commencement of the simulation in 2014 lasting up to 2022 (state just after the cut) and the transformation would start in 2024 lasting up to 2032 (state just before the cut) similarly as compartment 1).

The estimate of the cut for each cut cycle was based on theoretical rate of current annual volume increment standing at 2.5% (10-year cut intensity of 25%) according to Klepac (1961). The commencement of stand transformation to uneven-aged structure was simulated by establishment of 5 initial regeneration gaps by cutting of all trees in circle diameter of 20–25 m on the »virtual« stand research object covering an area of 3 ha (1.7 gaps per ha). The other segment of the performed cut was implemented on the remaining stand area as thinning. Concerning the following cut cycles, the enlargement of the initial gaps by cut of circlemarginal trees, establishment of new regeneration gaps and thinning were simulated.

costs which, for even-aged silver fir forests stand at 22.09 EUR/ha (Beljan 2015). According to available calculations, average direct costs of wood harvesting (1997–2013), in relation to the selling price achieved, stand at 1:0.654 (±0.08) (Beljan 2015). In other words, wood harvesting costs per 1 Euro of generated revenue stand at 0.65 Euro ± 8 Euro cents. This ratio was also projected in the future. The costs also include the share of 5% of the selling price of stumpage, which, according to the National Forest Law (OG, 140/2005) is collected from legal entities involved in forestry.

Purchase of (investment in) both the forest and land within the price range between 1000 and 12,500 EUR/ha was analysed for the purpose of this research.

2.3.2 Assessment of economic returns and profitability Economic analysis encompasses the period commencing from 2013 (actual situation in the field) up to 140 years in the future for an even-aged scenario and 120 years for uneven-aged scenario. Subsequently, the cash flow is continuous and in balanced structure theory it is assumed to be for an unlimited period of time.

The comparison of economic effectiveness amongst specific scenarios focused on net present value (NPV) (Klemperer 2003) for unlimited time period according to the Faustmann (1968) concept and internal rate of return (Damodaran 2002). Interest rates ranging from 1% to 5% were used concerning economic analysis features, with 2% taken as a reference value.

According to Knoke and Plusczyk (2001), net present value Jeven-age for an unlimited time period of the even-aged scenario was calculated by the following mathematical expression:

2.3 Economic analysis of silvicultural treatments  T 2.3.1 Cash flow 1 1

   J R v t = ∑ × + (1,0 Cash flow of each specific scenario was estimated even −age  ( )   t  ( 1,0r )T − 1 t =0 1,0 r ( ) based on i) the results of simulation of forest manage ment using the forest growth modelling software   T  MOSES, ii) timber assortment tables, iii) simulated T−t 1 1  × 1  (1)  v ( t )  +    Jeven −age R 1,0 r R v t = ∑the Croatian × × selling price, iv) cost in compliance with ( ) ( ) s     t  1,0r )T − 1 t =0  ( 1,0r )t  (1,0r ) exForest Law (OG, 140/2005), v) estimated through   (  penses arising from wood harvesting. Simulation of the future selling prices of beech and fir assortments here: was taken from earlier study of Beljan et al. (2017). A R[v(t)] profit during the current rotation simulation of timber prices on the market was made Rs[v(t)] profit during the upcoming rotation based on the collected data on achieved selling prices t time of the current rotation for the surveyed forest using Monte Carlo methodolT time of the upcoming rotation ogy (Beljan et al. 2017). r discount rate.

On the other hand, costs (both indirect and direct) are revaluation expressed by consumption of producThe first part of the mathematical expression (betion elements. Indirect costs include administrative fore the square bracket) referred to NPV (of the current

∑

∑

Croat. j. for. eng. 39(2018)2 303

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

rotation). From the commencement of the simulation to the final felling, all the profits (R[v(t)]) were discounted to the start through the discount rate (r). In this way the analysis of only a segment of the rotation was included (e.g., from 100 to 220 years of age, depending on the felling age in a specific compartment according to Table 2. The second part of the mathematical expression referred to NPV of the entire upcoming rotation shown over an unlimited time period. Irrespective of the fact that the time of the regeneration felling determined the commencement of production of a new stand, all the revenues and expenditures were capitalised/discounted according to the time of seeding felling (Nenadić 1922) and hence the factor 1.0r u − 1 (Navarro 2003) was used.

Net present value Juneven-age of uneven-aged scenario over an unlimited time period was described through the expression according to Knoke and Plusczyk (2001): J uneven −age=

1

(1,0r )

t

1

∑ (1,0r ) t

t

Fig. 1 Comparison of dynamics in growing stock at stand level.   During the first decade, 1 1  the dynamics of fluctuations in growing    R v t × R  v ( t )  +  × × ( )equal stock at both levels s  was cc t    1,0r ) − 1  ( 1,0r )   (

   1 1     × R  v ( t )  + × Rs  v ( t )  ×

cc t    − 1,0 r 1 1,0 r ( ) ( )   

(2)

Where: R[v(t)] profit during transformation Rs[v(t)] profit after transformation t time of transformation cc time after transformation r discount rate.

The first part of the mathematical expression (before the square bracket) discounted the profit throughout the duration of the transformation. Following the establishment of the uneven-aged structure, a continuation of forest management was expected upon equal net circulating capital generated every (cc) year, according to Navarro (2003) shown over an unlimited period of time.

3. Results 3.1 Growth data A comparison of dynamics in growing stock at the level of stand and forest showed the fundamental difference between scenarios (Fig. 1). In Fig. 1, the scenario for an even-aged management system was pre-

304

sented through a sub-compartment (stand) 1a, whilst the uneven-aged scenario was presented through a compartment 1. The end of the simulation at the stand level (sub-compartment 1a) was defined by the time of the final felling of the even-aged scenario and, from the temporal aspect, it matched the previously established balanced selection stand structure) (compartment 1). It is important to emphasize the fact that fluctuations in growing stock of the uneven-aged scenario within the theoretical values occurred 30 years prior to the even-aged scenario.

The development of growing stock at forest level is primarily conditioned by the features of the specific scenario. A rise in average growing stock at forest level up to a remarkable 750 m3/ha (Fig. 1) was shown in case of the scenario for the even-aged management system, due to the extension of the felling age up to 220 years old. The uneven-aged scenario continuously reduced the growing stock to the minimum amount after 65 years of simulation, when a gradual increase in growing stock and the establishment of balanced selection stand structure commenced.

3.1.1 Even-aged management scenario At the beginning of the analysed time period, the growing stock both at stand and forest level was above the optimal level (Table 1), for details see (Čavlović 2013) and (Beljan 2015). Simulation at forest level in even-aged management scenario included individual Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299â&#x20AC;&#x201C;312) Karlo Beljan et al.

Fig. 2 Dynamics of growing stock in even-aged scenario divided into compartments (a compartment is comprised of ten stands/ sub-compartments) at forest level. The growing stock for individual compartment is the average of its 10 sub-compartments simulations of each sub-compartment (110 pieces) distributed in 11 compartments shown in Fig. 2. The growing stock in a specific year for an individual compartment was the average of all sub-compartments of the corresponding compartment (Fig. 2).

A balanced even-aged forest structure was expected to be established after 140 years. It would include all age classes, whilst the forest would comprise 110 even-aged stands, where under the tree tops of the old stand there would be a new one (for instance, under 120-year old stand there would be a young â&#x20AC;&#x201C; 10-year old stand). Growing stock per hectare at forest level corresponded to half the stock of the 3rd compartment, which had the highest growing stock amongst all the compartments (Fig. 2). This was also an indicator of a balanced even-aged forest structure.

cycles, the gaps intended for regeneration were concentrically expanded through cutting of the circlemarginal trees, whilst selection cutting was performed throughout the remainder of the stand. This procedure was repeated until the entire stand was regenerated and the uneven-aged structure established after 110 years at stand level, i.e. after 120 years at forest level (Fig. 3). A gradual decrease in the proportion of silver fir in the growing stock enabled its substantial regeneration. It is important to highlight that this also resulted in beech regeneration which, at the beginning of the simulation, covered only 3.45 m3/ha (Table 1), whilst after 70 years it achieved the desired growing stock.

Simulation at forest level was the average of simulations of all the stands distributed into 10 compartments (Fig. 3). The dynamics of growing stock at that level followed the trend of individual stand and decreased from the commencement of the simulation period. That happened during the first 60 years of simulation, after which the increase in growing stock started and it lasted until 120th year of the simulation, when the uneven-aged structure at forest level was established (Fig. 3). The same picture shows that the balanced uneven-aged forest was gradually achieved; first, in the compartment No. 1 and in all the subsequent compartments upon one-year interval. Analogous to the indicator of normality for even-aged management scenario, growing stock at forest level corresponded to half the growing stock between two cutting cycles (uneven-aged forest management).

3.1.2 Uneven-aged management scenario This scenario includes a transformation of the management system from even-aged into uneven-aged.

Tree felling into 5 gaps of the radius from 25 m to 30 m was simulated over the surface represented by the virtual object of the research (3 ha). Not all the trees within the gap intended for forest regeneration were cut, but due to the seeding of the soil with seeds several trees were left. The cutting in first cutting cycles was spatially concentrated for the opening of the gaps intended for regeneration. Through subsequent cutting

Fig. 3 Development of growing stock of uneven-aged scenario at both stand and forest level

Croat. j. for. eng. 39(2018)2 305

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

3.2 Economic analysis 3.2.1 Cash flow Cash flow can be observed both at stand and at forest level, albeit the forest level covers a significantly larger area. At forest level, revenue and cost were present every year and their sum was comprised of revenue and cost of all the stands in annual yield area (Fig. 4). Revenue and cost are closely connected. In the event of higher revenues, the expenditures will also be higher, yet disproportionately, as shown in Fig. 4.

Cash flow of both scenarios was not even roughly identical throughout the period.

The scenario for even-aged management of different felling-age (Table 2) prescribed for a specific compartment resulted in different dimensions of trees and hence in different revenue. During the first decade, the cash flow was roughly identical due to the fact that every year 1/10 of the forest area was cut with equal intensity (Fig. 4). During the subsequent 20 years, both revenues and costs dropped drastically, due to the fact that the felling age regularly exceeded 120-year rotation. Throughout that period (10–30 years of simulation) annual surface of final cut stood at 8.55 ha (the area covered by one sub-compartment). In other words, most of the yield was achieved by the intermediate yield – with an inferior financial result. After that, every year the final felling was achieved over an equal area covering 8.55 ha, yet every subsequent year it was performed in the sub-compartment that was 1

Fig. 4 Cash flow of forest management in both scenarios at forest level. Investment costs of both forest and land purchase have not been shown

306

year older and hence its financial value was also higher. The older the stand, the higher its financial value, yet up to a point which occurred in this specific situation after 50 years of simulation. Fig. 4 also shows the previously mentioned recurrent drop in revenue and cost. The decrease in cash flow was the result of lower revenue from stands, which were considerably old and had an increased share of dead trees and wood of inferior value. As the end of the simulation approached, cash flow showed a deceasing number of fluctuations and eventually, expectedly, it entirely harmonised.

Cash flow in case of uneven-aged management scenario assumed entirely different attributes (Fig. 4).

The sum of revenues and costs of all the stands in a specific year is considered the cash flow at forest level.

At the commencement of the projection period, the revenue from beech could not be expected due to the fact that its share was too insignificant. Nevertheless, the first revenue was generated in that aspect after 40 years of simulation of uneven-aged forest management. On the other hand, cash flow from silver fir at the commencement of the projection period showed the highest amounts due to a more intensive cutting of accumulated growing stock (Fig. 4). Summary cash flow (silver fir and common beech) recorded the lowest values between the 60th and the 80th year of the simulation, as a direct consequence of inevitable reduction in growing stock, as shown in Fig. 3. It is also important to highlight that both revenues and costs at the commencement of the simulation were almost double compared with those at the end – when the balanced uneven-aged forest structure has been established. Another objective was met in this scenario, i.e. the forest can be provided with such features that it can generate equal profit throughout the year without any time limits.

3.2.2 Economic consequences Net present value over an unlimited time period was shown as a specific crossword, from the values for different costs of purchase of (investment in) a forest can be read (Table 3). Cases regarded as unprofitable from an economic standpoint have been shaded in grey. Following the comparison amongst different NPV scenarios, shown in Table 3, it was evident that even-aged scenario was more adequate throughout an entire range of investment costs, albeit only at discount rate of 1%. Break-even point for uneven-aged transformation scenario was given when an interest rate of 1.24% was applied, irrespective of the proportion of investment costs. In other words, uneven-aged scenario was more adequate when discount rate that exceeded 1.24% was applied.

Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Karlo Beljan et al.

Table 3 Net present value [EUR/ha] upon application of different purchase prices and discount rates Interest rate, % Investment

1

EUR/ha

2

3

4

5

EA*

UA**

EA*

UA**

EA*

UA**

EA*

UA**

EA*

UA**

1000

11,286

9368

6077

6153

4071

4600

2982

3599

2300

2903

2500

9786

7868

4577

4653

2571

3100

1482

2099

800

1403

5000

7286

5368

2077

2153

71

600

–1018

–401

–1700

–1097

7500

4785

2868

–423

–347

–2429

–1900

–3519

–2901

–4201

–3597

10,000

2285

368

–2923

–2847

–4929

–4400

–6019

–5401

–6701

–6097

12,500

–215

–2132

–5423

–5347

–7429

–6900

–8519

–7901

–9201

–8597

*even-aged management scenario, **uneven-aged management scenario

Comparison at forest level (Table 3) provided an approximate overview of the economic features of forest management, as opposed to the analysis at stand level (Fig. 5). At stand level and at reference discount rate of 2%, the dependence of NPV on investment costs and the management system were shown. Absolutely expectedly, higher values of NPV were achieved by stands purchased at lower prices, which would be

Fig. 5 Comparison amongst NPV at stand level at reference discount rate of 2% and investment cost of a) 1000, b) 2500, c) 5000, d) 7500, e) 10,000, f) 12,000 EUR/ha. Due to the features of a specific management scenario the forest has been divided into 10 or 11 compartments. A specific compartment comprises 10 stands and hence values for each stand have been provided in the graph

cut in the forthcoming future. NPV of even-aged scenario was higher in relation to the uneven-aged scenario in all the compartments management with felling age that ranges from 120 to 170 years old (Table 2, Fig. 5). Irrespective of the fact that at discount rate of 2% and investment of 7500 EUR/ha NPV at forest level were negative for both scenarios (Table 3), a proportion of stands still achieved a positive NPV (Fig. 5).

An overview of changes in mutual ratio of NPV at other discount rates would be identical to the overview presented in Fig. 5, yet the amount of NPV at a higher rate would be lower and vice versa. Management in compartments, whose turn for cutting is earlier, will be more profitable compared with the compartments in which cutting occurs later and which can have a negative NPV (Fig. 5).

Internal rate of return (IRR) fluctuated substantially, depending on the levels of the investment cost (Table 4). Initially, upon the lowest investment cost, IRR may appear to be unrealistically high, yet in case of the opportunity of investment in such forests at that price, IRR would be identical to the values shown in Table 4. On the other hand, the highest investment cost resulted in a considerably low IRR. From this table, it is important to highlight that IRRs in uneven-aged scenario are higher (in most cases) compared with those in even-aged scenario, whilst the interval of fluctuations of IRR (min-max) was shorter upon higher investment costs.

The applied discount rate has the greatest impact on economic comparison between the two scenarios.

It is extremely important to make a comparison between the scenarios at identical discount rate. Table 3 shows the impact of different rates on NPV, which will

Croat. j. for. eng. 39(2018)2 307

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

Table 4 Comparison of internal rate of return (IRR) both at forest and stand level for a period of 140 years Internal rate of return, % Investment cost EUR/ha

Even-aged

Uneven-aged

Forest level

Stand Min.

Stand Max.

Forest level

Stand Min.

Stand Max.

1000

25.4

24.9

28.7

27.3

12.2

>150

2500

7.0

5.8

10.6

8.8

6.0

26.0

5000

3.1

2.4

4.6

3.6

2.8

4.8

7500

1.9

1.5

2.4

1.8

1.5

2.2

10,000

1.3

1.0

1.5

1.1

0.9

1.2

12,500

1.0

0.8

1.1

0.7

0.7

0.8

regularly be lower at a higher discount rate and it is likely that NPV can also be negative in case high discount rate is applied (Table 3, Fig. 5). The highest discount rate that can be applied in order for NPV to equal zero has been shown for different investment amounts in Table 4 (column Forest level). Each discount rate inferior in relation to that provided therein will achieve a positive NPV. Consequently, it is evident that low discount rates positively affected evenaged scenario. Comparison between the two scenarios showed that the difference in NPV was not constant, i.e. it was less evident at higher rates (Table 3).

The results presented thus far show management until the establishment of balanced even-aged i.e. uneven-aged forest and subsequently management through theoretical balanced forest. Nevertheless, upon comparison of management only in balanced forests, the comparison was slightly different. Evenaged balanced forest achieved constant average annual yield of 10.72 m3/ha, whilst uneven-aged that of 11.72 m3/ha (the difference of 8.5%). Upon ignoring the fluctuations in price, it stood at 106.39 EUR/ha for even-aged, or 127.98 EUR/ha for uneven-aged scenario (the difference of 16.86%).

4. Discussion 4.1 Forest planning The main directions of future forest management need to be analysed against the backdrop of the current state of affairs in even-aged silver fir forests in the Dinaric Region. Two scenarios (both even-aged and uneven-aged scenario) which can be implemented in practice have been analysed. Even-aged scenario im-

308

plies the continuation of the currently implemented forest management, whereas uneven-aged is considered as a turning point in forest management, which requires a transformation during the first phase.

The scenario for the even-aged management system is a kind of imitation of the management system applied thus far in such forests. During the period in which pure silver fir forests in the Croatian Dinaric region were managed exclusively through even-aged management system, the rotation period of 120 (100) years with a regeneration period of 20 (10) years and regeneration felling applied twice (Meštrović 2001) was resorted to. Due to ecological characteristics, the scenario with the rotation period of 120 years, regeneration period of 20 years and regeneration felling applied 3 times was analysed.

On the other hand, the scenario for uneven-aged forest management implied a transformation from even-aged to uneven-aged forest. This process requires a change in the stand structure from homogeneous even-aged into a more complex uneven-aged (O’Hara 2001, Pommerening 2006). Several factors played a crucial role in a successful transformation.

Most authors (Hanewinkel and Pretzsch 2000, Malcolm et al. 2001, Schütz 2001, Mason and Kerr 2004, Čavlović and Božić 2007, Francetić 2010, Božić et al. 2011, Knoke 2012) agree that regeneration (both natural or artificial) plays the most important role. Even-aged stand can establish the uneven-aged structure by a natural process through natural disasters (Koop 1989, Peterken 1996), yet upon implementation of silvicultural measures the process is considerably shortened (Schütz 2001). The ideal timing for the commencement of the transformation is up to one half of the prescribed rotation period of even-aged forest management (Schütz 2001, Knoke 2012) and the process should commence after the full seed yield (Malcolm et al.

2001). Hence, in this research the places in the virtual forest which were naturally regenerated were used as centres of regeneration nuclei (gaps). In addition, this was supported by the fact that clearcutting in small gaps and their gradual expansion through cutting of the circle-marginal trees were the most adequate for pure coniferous stands (Hanewinkel and Pretzsch 2000, Hanewinkel 2001) both from the silvicultural (Malcolm et al. 2001) and economic standpoint (Hanewinkel 2001). In the previously mentioned forest, the height of the mean basal-area tree was 27.2 m (Beljan 2015).

Consequently, according to Malcolm et al. (2001), the diameter of the circle-shaped regeneration patch ranged between 25 m and 30 m, depending on the characteristics of the terrain and the arrangement of the trees in the area covered by the virtual stand.

Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Karlo Beljan et al.

4.2 Economic analysis and cost control in forest management Economic comparison between two main forest management systems can be made only with silver fir or spruce forests, since they enable the application of both systems (Hanewinkel 2002) and transformation from even-aged into uneven-aged management system and vice versa. The economic feature of the transformation can be analysed both at stand level (Buongiorno 2001, Hanewinkel 2001, Knoke et al. 2001, Knoke and Plusczyk 2001) and at forest level (Buongiorno and Gilles (1987), Price and Price (2006)). However, the conclusion on the economic result of forest management needs to be made exclusively at forest level and, if possible, at normal balanced forest level. It is important to point out that the comparison at stand level depends on the selection of stands for the purpose of comparison. The selection of different stands would result in a different comparison. In this specific research, there was a total of 1000 different combinations of stands for comparison and hence caution needs to be exercised during the interpretation of results at this level and consequently comparison at forest level is required. Although a wide range of potential buying prices of forest was examined, IRR results can seem unrealistic. Analysed purchase prices ranging from 1 EUR/ha to 12,500 EUR/ha are hypothetical. So the investor’s ability to negotiate prices is crucial for economic results. If an investor can make an agreement and invest (buy) a forest for 1 EUR/ha, IRR will be 150% because standing timber is around 600 m3/ha.

Other analysed buying prices (like 5000 or 7000 EUR/ha) will result in IRR as in similar research. In other words, all possibilities have been examined, the worst and the best ones.

Transformation from the even-aged into unevenaged management system is considered as a kind of turning point in forest management, which also affects the economic aspect. It is financially unprofitable to commence the process of transformation in the event when even-aged stand is »close« to financial maturity (Knoke 2012), which was also the case in this research.

However, Knoke (2012) reached this conclusion based on the analysis conducted at stand level. The characteristics of the object of research determine which management system is more cost-effective. In case of the Croatian Dinaric region, clearcutting in small gaps is the only logical option since the piled up and homogeneous standing timber make difficulties in natural regeneration. Nevertheless, in that case it is possible to reduce the value of timber assortments due to the growth of side branches on the remaining trees located on the edge of the circular patch (Macdonald et

al. 2010) and increase harvesting costs. Both the economic and ecological justification of the transformation of forest management system from pure stands into mixed is relative (Knoke et al. 2005). However, an advantage to mixed stands is certainly the reduction of risk spread between two or more forest tree species (Bauhus et al. 2017). In case of the existence of a balanced even-aged or uneven-aged forest, transformation is never recommended from the economic standpoint (Price 2012). Silvicultural characteristics, primarily the age class ratio in case of even-aged management and tree age structure of all the stands in uneven-aged management, are the factors that define the cost-effectiveness of transformation. In other words, each forest, due to its specific features, requires a special analysis (Price 2012). Natural regeneration is the most important economic factor in this process (Davies and Kerr 2011), since it represents the creation of capital that is a gift of nature (Navarro 2003). Irrespective of the fact that, from silvicultural standpoint, both management systems can be applied, it is clear that, from the economic aspect, a balanced unevenaged forest has a higher financial impact.

Both cash flows (Fig. 4) give the impression that simulations take the forest into a direction in which the profit will be reduced. It is important to highlight that a unique method has been presented, in which a normal balanced forest can be established generating optimal profits that will be continuous in theory. Otherwise, huge profits from the forest can be short-term and without the perspectives of financial gains in the long-term future.

As is the case with other types of decisions, the same applies to economic decisions, which are normally made against the backdrop of risk and uncertainty and this is particularly evident in long-term projects, as is the case in forestry. Most economic studies on forest management ignore the uncertainty of the future cash flows (Knoke et al. 2005), whereas in this research risk was considered as an integral part of economic analysis. In addition to the potential risk from natural disasters, it is also necessary to consider the uncertainty of changes in selling prices of timber assortments (Beljan et al. 2017) and timber harvesting costs. Timber assortment production is characterised by long production cycles and it is influenced by natural disaster risks. Forestry investment risk differs from country to country and it is directly affected by the risk carried by a specific country. A country whose economy implies lower risk will also be the one to apply a lower discount rate (Snowdon and Harou 2013). The selected discount rate that is applied in forest management analysis plays a decisive role in achieving good

Croat. j. for. eng. 39(2018)2 309

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312)

economic results (Klemperer et al. 1994, Price 1997, Brukas et al. 2001, Kanas 2008). Premium of the investment risk in forestry production is extremely low as opposed to other branches of the economy, where both the yield and risk are higher.

Bončina, A., 2011: History, current status and future prospects of uneven-aged forest management in the Dinaric region: an overview. Forestry 84(5): 467–478.

5. Conclusions

Bottalico, F., Travaglini, D., Fiorentini, S., Nocentini, S., 2014: Stand dynamics and natural regeneration in silver fir (Abies alba Mill.) plantations after traditional rotation age. IforestBiogeosciences and Forestry 7(5): 313–323.

In the silver fir forests located in the Croatian Dinaric region, which are currently even-aged, both management systems can be applied. Nevertheless, due to their current characteristics, the piled up standing timber into even-aged structure, they are not sustainable over the long term and hence they need to be directed towards the uneven-aged structure through a gradual transformation. The economic effectiveness of the transformation is determined primarily by the current characteristics of a specific forest and it can in some cases be cost-effective over a long period of time.

Consequently, it can be concluded that, from the longrun economic perspective, even-aged silver fir forests need to be transformed into balanced uneven-aged forests, which will be continuously providing higher monetary and non-monetary (public benefit function) values. During the transformation process, discount rates of up to 1.24% favourably affect even-aged management, whereas the increasing rates have a positive impact on uneven-aged management. Upon comparative analysis at the level of the established balanced forest, uneven-aged management will achieve a higher Net Present Value irrespective of the discount rate, as a consequence of a higher and financially more valuable forest increment/yield.

6. References Bauhus, J., Forrester, D.I., Gardiner, B., Jactel, H., Vallejo, R., Pretzsch, H., 2017: Ecological stability of mixed-species forests. In: Mixed-Species Forests: Ecology and Management (Pretzsch, H., Forrester, D.I., Bauhus, J. eds.), Springer Berlin Heidelberg, Berlin, Heidelberg, 337–382 p.

Beljan, K., 2015: Economic analysis of even-aged silver fir (Abies alba Mill.) forest management. PhD thesis. University of Zagreb, Croatia, 166 p.

Beljan, K., Posavec, S., Orsag, S., Teslak, K., 2017: Simulation model for prediction of timber assortment price trends in Croatia – a case study of Forest Office Brinje. Drvna industrija 68(2): 145–152.

Beljan, K., Vedriš, M., Mikac, S., Teslak, K., 2016: Generating virtual even-aged silver fir stand structure based on the measured sample plots. South-east European forestry 7(2): 119– 127.

310

Bončina, A., Diaci, J., Cenčić, L., 2002: Comparison of the two main types of selection forests in Slovenia: distribution, site conditions, stand structure, regeneration and management.

Forestry 75(4): 365–373.

Božić, M., Čavlović, J., Goršić, E., Teslak, K., 2011: Dynamics of establishing the selection structure in beech-fir stands on Papuk. Croatian Journal of Forest Engineering 21(1): 287–298.

Brukas, V., Thorsen, B.J., Helles, F., Tarp, P., 2001: Discount rate and harvest policy: implications for Baltic forestry. Forest Policy and Economics 2(2): 143–156.

Buongiorno, J., 2001: Quantifying the implications of transformation from even to uneven-aged forest stands. Forest Ecology and Management 151(1–3): 121–132.

Buongiorno, J., Gilles, J.K., 1987: Forest management and economics – A primer in quantitative methods. Macmillan Publishing, New York, 285 p. Čavlović, J., 2010: First national forest inventory in Republic of Croatia. Ministry of regional development and forestry & Faculty of Forestry, Zagreb, 300 p. Čavlović, J., 2013: Basics of forest management planning. ­University of Zagreb – Faculty of Forestry, Zagreb, 322 p. Čavlović, J., Božić, M., 2007: The establishment and preservation of a balanced structure of beech-fir stands. Glasnik za šumske pokuse 42: 75–86. Čavlović, J., Božić, M., 2008: National forest inventory in Croatia – methods of field assesment. University of Zagreb – Faculty of Forestry, Zagreb, 146 p. Čavlović, J., Božić, M., Bončina, A., 2006: Stand structure of an uneven-aged fir-beech forest with an irregular diameter structure: modeling the development of the Belevine forest, Croatia. European Journal of Forest Research 125(4): 325–333.

Damodaran, A., 2002: Investment valuation – Tools and techniques for determining the value of any asset. John Wiley & Sons, Inc. New York, 992 p.

Davies, O., Kerr, G., 2015: Comparing the costs and revenues of transformation to continuous cover forestry for sitka spruce in Great Britain. Forests 6(7): 2424–2449.

Davies, O., Kerr, K., 2011: The costs and revenues of transformation to continuous cover forestry. The Research Agency of the Forestry Commission, 63 p.

Faustmann, M., 1968: Berechnung des Wertes, welchen Waldboden, sowie noch haubare Holzbestände für die Waldwirtschaft besitzen. Allgemeine Forst und Jagdzeitung 15: 441–455.

Ficko, A., Poljanec, A., Bončina, A., 2011: Do changes in spatial distribution, structure and abundance of silver fir (Abies alba

Croat. j. for. eng. 39(2018)2

Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Karlo Beljan et al.

Mill.) indicate its decline? Forest Ecology and Management 261(4): 844–854.

Francetić, I., 2010. Structural characteristics of pure fir forests on the locality of Forest offices Brinje and Otočac. Specialist master thesis. University of Zagreb – Faculty of Forestry, 83 p.

Franklin, J.F., 1989: Toward a new forestry, American Forests 37(11–12): 37–44.

Hanewinkel, M., 2001: Economic aspects of the transformation from even-aged pure stands of Norway spruce to unevenaged mixed stands of Norway spruce and beech. Forest Ecology and Management 151(1–3): 181–193.

Hanewinkel, M., 2002: Comparative economic investigations of even-aged and uneven-aged silviculture systems: a critical analysis of different methods. Forestry 75(4): 473–481.

Hanewinkel, M., Pretzsch, H., 2000: Modelling the conversion from even-aged to uneven-aged stands of norway spruce (Picea abies L. Karst.) with a distance-dependent growth simulator. Forest Ecology and Management 134(1–3): 55–70.

Koch, N.E., Skovsgaard, J.P., 1999: Sustainable management of planted forests: some comparisons between Central Europe and the United States. New Forests 17(1–3): 11–22.

Koop, H., 1989: Forest dynamics. Springer Verlag, Berlin, 229 p.

Macdonald, E., Gardiner, B., Mason, W., 2010: The effects of transformation of even-aged stands to continuous cover forestry on conifer log quality and wood properties in the UK. Forestry 83(1): 1–16.

Malcolm, D.C., Mason, W.L., Clarke, G.C., 2001: The transformation of conifer forests in Britain – regeneration, gap size and silvicultural systems. Forest Ecology and Management 151(1–3): 7–23.

Mason, W., Kerr, G., 2004: Transforming even-aged conifer stands to continuous cover management. Forestry Commission Information Note 40, Edinburgh.

Matić, S., 1989: The intensity of thinning and its impact on the stability, productivity and regeneration stands of oak. Glasnik za šumske pokuse 25: 261–278.

Hasenauer, H., 2006: The growth model MOSES 3.0. In: Sustainable Forest Managemet: Growth Models for Europe (Hasenauer, H. ed.). BOKU, Vienna, 64–71.

Matić, S., Oršanić, M., Anić, I., 1996: Some characteristics and problems of Silver fir (Abies alba Mill.) selection forests in Croatia. Šumarski list 120(3–4): 91–99.

Kanas, A., 2008: On real interest rate dynamics and regime switching. Journal of Banking & Finance 32(10): 2089–2098.

McMahon, J.P., 1999: International expectations for sustainable forestry: a review from US forest industry. New Forests 17:329–338.

Kenk, G., Guehne, S., 2001: Management of transformation in central Europe. Forest Ecology and Management 151(1–3): 107–119.

Keren, S., Motta, R., Govedar, Z., Lucic, R., Medarevic, M., Diaci, J., 2014: Comparative structural dynamics of the Janj mixed old-growth mountain forest in Bosnia and Herzegovina: are conifers in a long-term decline? Forests 5(6): 1243– 1266.

Meštrović, Š., 2001: Managing forests of silver fir. In: Silver Fir in Croatia (Klepac, D. ed.), Academia scientiarum forestariarum, Zagreb, 548–560.

Mikac, S., Klopf, M., Anić, I., Hasenauer, H., 2013: Using the tree growth model MOSES to assess the dynamics of Dinaric old-growth mixed beech-fir forest ecosystems. Plant Biosystems 147(3): 664–671.

Klemperer, D.W., 2003: Forest resource economics and finance, McGraw-Hill, 551 p.

Mlinšek, D., 1996: From clear-cutting to close-to-nature silviculture system, IUFRO News 25(4): 6–8.

Klemperer, D.W., Cathcart, J.F., Haring, T., Alig, R.J., 1994: Risk and the discount rate in forestry. Canadian Journal of Forest Research-Revue 24(2): 390–397.

Navarro, G.A., 2003: A Re-examing the theories supporting the so-called Faustmann Formula. In: Recent Accomplishments in Applied Forest Economics Research (Helles, F., Strange, N. and Wichmann, L. eds.), Kluwer Academic Publishers, Dordrecht/Boston/London, 19–38.

Klepac, D., 1961: New management system of uneven aged forests. Agricultural-Forestry Chamber of NR Croatia, Zagreb, 46 p.

Knoke, T., 2012: The economics of continuous cover forestry.

In: Continuous Cover Forestry (Pukkala, T. and Gadow, K. eds.), Springer, 166–193.

Knoke, T., Moog, M., Plusczyk, N., 2001: On the effect of volatile stumpage prices on the economic attractiveness of a silvicultural transformation strategy. Forest Policy and Economics 2(3–4): 229–240.

Knoke, T., Plusczyk, N., 2001: On economic consequences of transformation of a spruce (Picea abies (L.) Karst.) dominated stand from regular into irregular age structure. Forest Ecology and Management 151(1–3): 163–179.

Knoke, T., Stimm, B., Ammer, C., Moog, M., 2005: Mixed forests reconsidered: A forest economics contribution on an ecological concept. Forest Ecology and Management 213(1–3): 102–116.

Nenadić, Đ., 1922: Calculating forest value and forest statics.

Croatian Forestry Society, Zagreb, 423 p.

Nenadić, Đ., 1929: Forest management. Tisak zaklade tiskare Narodnih novina u Zagrebu, Zagreb, 302 p.

O’Hara, K.L., 2001: The silviculture of transformation – a commentary. Forest Ecology and Management 151(1–3): 81–86.

OG, 140/2005: Forest law. Official Gazette 140/2005.

Peterken, G.F., 1996: Part I: Definitions, significance and survival, Part II: Temperate and boreal natural forests. In: Natural Woodland: Ecology and Conservation in Northern Temperate Regions (Peterken, G.F. ed.), Cambridge University press, Cambridge, 227 p.

Pommerening, A., 2006: Transformation to continuous cover forestry in a changing environment. Forest Ecology and Management 224(3): 227–228.

Croat. j. for. eng. 39(2018)2 311

Karlo Beljan et al. Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests (299–312) Pretzsch, H., Biber, P., Dursky, J., Gadow, K., Hasenauer, H., Kändler, G., Genk, G., Kublin, E., Nagel, J., Pukkala, T., Skovsgaard, J.P., Sodtke, R., Sterba, H., 2002: Recommendations for standardized documentation and further development of forest growth. Forstwissenschaftliches Centralblatt vereinigt mit Tharandter forstliches Jahrbuch 121(3): 138–151.

Price, C., 1997: A critical note on a long-running debate in forest economics. Forestry 70(4): 389–397.

Price, C., 2002: The economics of transformation from evenaged to uneven-aged forestry. In: Recent Accomplishments in Applied Forest Research (Helles, F., Strange, N. and Wichmann, L. eds.), Kluwer academic Publishers, Dordrecht/Boston/London, 3–17.

Price, C., 2012: Normal forest structures and the costs of ageclass transformation. In: Scandinavian Forest Economics (Toppinen, A., Karppinen, H. and Kleemola, K. eds.). Scandinavian Society of Forest Economics, Helsinki, 237–238.

Price, M., Price, C., 2006: Creaming the best, or creatively transforming? Might felling the biggest trees first be a winwin strategy? Forest Ecology and Management 224(3): 297– 303.

Pukkala, T., Lahde, E., Laiho, O., 2010: Optimizing the structure and management of uneven-sized stands of Finland.

Forestry 83(2): 129–142.

Remeš, J., 2006: Transformation of even-aged spruce stands at the School Forest Enterprise Kostelec nad Černymi lesy: Structure and final cutting of mature stand. Journal of Forest Science 52(4): 158–171.

Salim, E., Ullsten, O., 1999: Our forests, our future: Report of the World Commission on Forests and Sustainable Development, 38 p.

Schütz, J.P., 2001: Opportunities and strategies of transforming regular forests to irregular forests. Forest Ecology and Management 151(1–3): 87–94.

Snowdon, P., Harou, P., 2013: Guide to economic appraisal of forestry investments and programmes in Europe. Forestry Commission, 31 p.

Steinmetz, P., 2003: MOSES 3.0. Forest growth modelling software – User manual, 19 p. Šafar, J., Hajdin, Ž., 1954: The problem of silver fir expanding areal at the hilly area between the Sava and Drava river. Šumarski list 78(9–10):486–495.

Teslak, K., Vedriš, M., Gašparović, M., Žunić, M., Jura, Č., 2016: Stand regeneration characteristics of beech and fir forests in Gorski Kotar region. South-east European forestry 7(2): 99–108.

Vukelić, M., 2000: Osvrt na obnovu prirodnih šuma. Šumarski list 124(3–4): 227–230.

Authors’ addresses: Karlo Beljan, PhD. * e-mail: kbeljan@sumfak.hr Assoc. prof. Stjepan Posavec, PhD. e-mail: sposavec@sumfak.hr Prof. Jura Čavlović, PhD. e-mail: jcavlovic@sumfak.hr Assist. prof. Krunoslav Teslak, PhD. e-mail: kteslak@sumfak.hr University of Zagreb Faculty of Forestry Department of Forest Inventory and Management Svetošimunska 25 10000 Zagreb CROATIA

Received: November 9, 2017 Accepted: March 7, 2018

312

Prof. Thomas Knoke, PhD. e-mail: knoke@tum.de Technische Universität München (TUM) Institute of Forest Management School of Life Sciences Weihenstephan Department of Ecology and Ecosystem Management Hans-Carl-von-Carlowitz-Platz 2 85354 Freising GERMANY * Corresponding author Croat. j. for. eng. 39(2018)2

Original scientific paper

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass for Biorefinery Products Alan Antonović, Damir Barčić, Jaroslav Kljak, Josip Ištvanić, Tomislav Podvorec, Juraj Stanešić Abstract Open-air fires or forest fires are becoming a key factor in reducing the forest surface areas and they are one of the major factors of devastation and degradation of forests and forest land and their ecosystems in the Mediterranean, mainly in coastal karst. They cause extreme material and economic damage, and they negatively affect biological and landscape diversity. After the forest fire, significant quantities of fired trees are left behind, representing a significant amount of lignocellulosic biomass available for conversion into a variety of biobased products. The question arises as to what degree they are chemically degraded, or whether they still have the properties required for further application in mechanical or chemical processing.

The main aim of this paper was to study the group chemical composition as a biomass chemical property of the Aleppo pine (Pinus halepensis Mill.) sapwood before and after the impact of low ground fire and high fire of the treetops at tree height of 0, 2 and 4 m. Therefore, the impact of forest fires on the Allepo pine sapwood group chemical composition was studied in terms of quality for further application in production of biorefinery products. In addition, research results on group chemical composition of the same unfired and fired Aleppo pine wood bark from previous study were used for comparison with sapwood from this study.

The obtained results show that the distribution of the main chemical components of Aleppo pine unfired wood bark and sapwood is similar to the results of previous studies for different wood species. That means that the bark contains a significantly higher content of ash, accessory materials (extractives) and lignins, and a significantly lower content of polysaccharides cellulose and polyoses (hemicellulose) than sapwood. The bark results from previous studies show a significant difference in reduced ash, cellulose and lignin content, and in the increased accessory materials and wood polyoses (hemicellulose) content between the unfired and fired wood. Furthermore, the content of individual chemical components of fired bark at different forest fires heights of 0, 2 and 4 m for each sample did not differ significantly. Contrary to fired bark, no significant differences have been observed in the chemical composition of sapwood between unfired and fired wood, not even resulting from different forest fires heights. It can be concluded that the forest fire did not have any effect on Aleppo pine sapwood, where the fired wood bark took over all the damage caused by high temperature during the forest fire. In addition, the fired sapwood still retains the chemical properties required for further application in biorefinery biobased products.

Keywords: Aleppo pine (Pinus halepensis Mill.), sapwood group chemical composition, forest fires, fired wood

1. Introduction The current model of production and consumption, which largely relies on fossil-based resources and has a decisive impact on the environment and the availability of natural resources, is approaching its peak.

Significant steps are being taken around the world to

move from today’s fossil based economy to a more sustainable economy based on biomass. A key factor in the realization of a successful biobased economy or bioeconomy is the production of a range of biobased products and bioenergy to substitute their fossil-derived equivalents by processing a wide variety of biological feedstock. The EU has declared the biobased products

Croat. j. for. eng. 39(2018)2 313

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

sector to be a priority area with high potential for future growth, reindustrialization, and addressing societal challenges. Furthermore, the circular economy is defined as an economy that is restorative and regenerative by design, and which aims to keep products, components and materials at their highest utility and value at all times, distinguishing between technical and biological-cycles. It focuses on the efficient use of finite resources and ensures that these resources are reused as long as possible (IEA Bioenergy 2017, AEBIOM 2017).

A bioeconomy, as well as circular economy, uses as many renewable raw materials as possible for products such as chemicals, materials and energy. As a part of bioeconomy and circular economy, biorefining, as the sustainable processing of biomass into a range of marketable biobased products and bionenergy/biofuels, is an innovative and efficient approach to use the available biomass resources for the synergistic co-production of chemicals, materials and energy. A biorefinery is a multidisciplinary and complex concept addressing, at the same time, the production of value-added bioproducts (chemical building blocks, materials), and bioenergy (biofuels, power, and heat) from biomass, within a sustainability assessment carried out along the entire value chain and life cycle. The development and implementation of biorefinery processes is of the upmost importance and constitutes the keystone for establishing a bioeconomy based on bioresources, and it is an absolute necessity and the key to meet this vision of a biobased economy. It can use various combinations of feedstock and conversion technologies to produce a variety of products and includes key technologies to effectively use (components from) biomass, involves converting biomass into valuable components and marketable products. In addition, the European Commission has set a long-term goal to develop a competitive, resource efficient and low carbon economy by 2050, and biorefinery products are expected to play an important role (IEA Bioenergy 2017).

Contrary to petro-resources, whose nature and composition variations are relatively limited, under the terms bioresource or biomass, compounds are gathered of very different nature. Biomass is defined as biodegradable products, wastes, and residues of biological origin from agriculture, forestry, and aquaculture, and comes from a wide range of raw materials that include wood, agricultural crops, by-products of wood processing, manure, and the organic fraction of waste products (Directive 2008/28/EC). Biomass, as a form of renewable raw material, has the advantages of being easily stored, transported, and used as a flexible load where and when the raw material is needed.

This makes biomass unique among other options of renewable raw materials (AEBIOM 2017).

314

Wood as a lignocellulosic biomass is a multicomponent, hygroscopic, anisotropic, fibrous, porous, biodegradable and renewable raw material. It is generally clear that wood has unique structural and chemical characteristics that show a wide spectrum of end-use possibilities. For these reasons, it can be assumed that the basic knowledge of the structure and chemical composition of wood is of essential importance, considering the choice optimization of certain wood species for different applications. Every wood species is unique in its chemical composition and varies from species to species, as well as its chemical, physical, and mechanical properties. Wood is a complex heterogeneous mixture of key structural organic components such as cellulose, hemicellulose, and lignin along with accessory organic and inorganic composites. From the chemical point of view, wood consists of 40–45% cellulose, 25–35% hemicelluloses, 15–30% lignin and up to 10% other compounds. All the main wood components are highweight-molecular polymers and form an interwoven network in the wood cell wall; consequently it can be said that the wood is a natural polymer. The chemical composition of wood tissues (sapwood and heartwood), as well as bark, is equally complex, and varies between and within species. Comparing the chemical composition of tissues and bark, it can be concluded that the bark contains a higher content of ash, accessory materials (extractives) and lignin, and a lower content of polysaccharides cellulose and polyoses (hemicellulose). The qualitative and quantitative characterization of such components in the biomass is essential for its application perspectives. Hence, an overall characterization of biomass is indispensable to expand the bioeconomy sector worldwide (Antonović 2010, Antonović, 2017).

The end-use processing pathways of wood, as a lignocellulosic biomass for biorefinery products, depend on its physicochemical properties. These properties are composed of the following: Þ biochemical composition (a) wood chemistry – cellulose, hemicellulose and lignin; (b) non-wood chemistry – saccharides, lipids and proteins Þ moisture content (a) intrinsic moisture; (b) extrinsic moisture Þ mineral matter content (a) major elements; (b) trace elements; (c) nutrients; (d) salts Þ elemental composition of organic matter (C, H, N, S, O) Þ physical properties (a) density; and (b) grindability.

Two fundamental aspects related to such biomass are: Þ to extend and improve the basic knowledge on composition and properties Croat. j. for. eng. 39(2018)2

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Þ apply this knowledge for the most advanced and sustainable utilization of biomass.

According to that, the systematic identification, quantification and characterization of chemical composition of such biomass are the initial and most important steps during the research and application of biorefinery products. The primary aspect in utilizing biomass for such products is to understand its basic composition and properties (Strezov and Evans 2015).

Recently in Croatia, and worldwide, concern has arisen with respect to the increasing number of forest fires. The causes of fires, as a result of certain human activities, are manifold and numerous. Open-air fires or forest fires are becoming a key factor in reducing the forest surface areas and they are one of the major factors of devastation and degradation of forests and forest land and their ecosystems in the Mediterranean, mainly in coastal karst. In addition to causing extreme material and economic damage, in a large number of cases they negatively affect biological and landscape diversity (Prgin 2005). The impact of open space fire, in terms of ecology, was explored by many authors.

These are extremely important problems of natural renewal and vegetation succession, occurrence of erosion processes after the fire and prevention of forest soil degradation. Another aspect, often neglected, refers to the technical properties of the fired wood.

Therefore, an important species for the Croatian Mediterranean karst was studied in this paper, the Aleppo pine (Pinus halepensis Mill.), the most important species for the afforestation of the Dinaric karst (Saracino et al. 1997, Pentek 1998, Espelta et al. 2002, Rodrigo et al. 2004, Tapias et al. 2004, Pausas et al. 2008, Španjol et al. 2011, Pentek et al. 2011, Pentek et al. 2014).

Forest fires occur when all the elements of a fire triangle come together in a susceptible area: an ignition source is brought into contact with a combustible material such as vegetation that is subjected to sufficient heat and has an adequate supply of oxygen from the ambient air. High moisture content usually prevents ignition and slows propagation, because higher temperatures are required to evaporate any water within the material and heat the material to its fire point. Dense forests usually provide more shade, resulting in lower ambient temperatures and higher humidity, and are therefore less susceptible to forest fires. Less dense material such as grasses and leaves are easier to ignite because they contain less water than denser material such as branches and trunks. Plants continuously lose water by evapotranspiration, but water loss is usually balanced by water absorbed from the soil, humidity, or rain. When this balance is not maintained, plants dry out and become more flammable; this being often a consequence of droughts (Prgin 2005).

A. Antonović et al.

According to the occurrence method, fires are grouped into: Þ natural (uncontrollable, wild, sifted) Þ artificial (controlled, planned) (Dimitrov 1987, Španjol 1996).

According to the fuel material type, fires are classified as follows: Þ underground fire or soil fire (roots and peat); Þ ground or low fire Þ fire in the tops or high fire Þ fire of lonely trees and shrubs (Španjol 1996).

Underground fire or soil fire (roots and peat) is affected by humus and peat layers beneath forestry mats. It advances very slowly but constantly. Peat fires may last (smolder) for several months and constantly threaten to emerge from the surface and become a dangerous groundwater fire. Damage is great because of the damage to the roots of trees that are then dried.

Such fire is very difficult to detect and it is difficult to extinguish. Ground fire or low fire occurs when the upper layer of forest mats, booming shrubs and young stands, are ignited. This is the most common type of fire that occurs in all types of forests. Strong ground fire in forests, where the trees have a thin bark, damages the tree roots, and causes cambium dying and trees drying, as well as dying of the whole stands. The temperature of 54 °C is sufficient for destroying the cambium. Fire in the tops or high fire is the kind of fire that surely destroys the entire forest ecosystem, and it is most frequent in coniferous forests. With the socalled flying fires, fire spreads from one top to another. With the help of the wind, this fire can destroy large forest complexes. The fire of lonely trees and shrubs is mainly caused by the lightning strike and in that case the tree burns down. It is most often connected to large forest areas. However, the cause may also be the burning of the fire close to the trees (Španjol 1996).

When trees are exposed to elevated temperatures caused by forest fires, changes can occur in their chemical composition, affecting their properties. The extent of changes depends on the temperature level and length of time under exposure conditions. Combustion of cellulose and lignin is preceded by thermal degradation, where gaseous and liquid products are formed as well as a solid residue of charcoal. Some gases and liquids, when mixed with air, burn with a flame, whereas the charcoal burns in the air without flame. Wood burns »indirectly«, in the sense that wood does not actually burn, but combustion takes place as a reaction between oxygen and the gases released from the wood (high molecular weight components). Under the influence of heat, wood produces easily substances that react eagerly with oxygen, making wood susceptible to

Croat. j. for. eng. 39(2018)2 315

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

fire. Ignition and combustion of wood is mainly based on thermal decomposition of cellulose and reactions of thermal degradation products with each other and with gases in the air, mainly oxygen. When temperature increases, cellulose starts to degrade. The decomposition products either remain inside the material or are released as gases. Gaseous substances react with each other and oxygen, releasing a large amount of heat that further induces degradation and combustion reactions (Thomas and McAlpine 2010).

According to the above mentioned, after the forest fires, significant quantities of fired Aleppo pine trees (Pinus halepensis Mill.) are left behind as a lignocellulosic biomass. The question arises as to what degree they are chemically degraded or whether they still have all the properties required for further application in mechanical or chemical processing. The main aim of this study was to chemically characterize fired Aleppo pine in terms of biomass quality and contribute to a better understanding of the possibility of using fired wood species, as a very large raw material base in the Republic of Croatia, in the further production of biorefinery products. Therefore, the impact of forest fires on chemical composition of the Aleppo pine sapwood group was studied as a quality property. The study was carried out by determining the group chemical composition (accessory materials, mineral substances, cellulose, hemicellulose and lignin) of Aleppo pine fired wood sapwood and unfired wood sapwood for comparison, by sampling the rings at the tree height of 0 m, 2 m 4 m (also referred as forest fire heights). Based on the results of the chemical analysis of the fired Allepo pine wood sapwood, the impact of forest fires on changes of chemical composition was determined by comparison with the unfired wood sapwood, as well as its possibility for further application in production of biorefinery products. In addition, research results on group chemical composition of unfired and fired Aleppo pine wood bark from previous study (Antonović et al. 2017) were used for comparison with sapwood from this study.

2. Materials and methods For studying the quality of fired wood as biomass for the production of biorefinery products, Aleppo pine wood (Pinus halepensis Mill.) was chosen, as the most common wood species in the coastal karst, especially in Dalmatia, the islands and the Dalmatian Zagora of the Republic of Croatia.

Aleppo pine wood covers more and more areas thanks not only to new afforestation but also to its biological properties of natural expansion and regeneration on fired surfaces. Monoculture of Aleppo pine fa-

316

vors faster spread of fires than mixed forests of Mediterranean hardwoods. The use of Aleppo pine wood in Mediterranean countries has a wide application. It is especially appreciated as quality wood in pulp production, and in some countries, wood industry is based on the Aleppo pine raw materials. Furthermore, it is used for heating, small technical wood in agriculture, mining wood, in construction for internal joinery, sawmill boards and others. Aleppo pine forests favor the development of economic activities, employment of the population, raising the quality of life of people and creating their local culture identity. With the opportunities offered by Aleppo pine forests, as well as other forest areas, new possibilities will emerge for the development of forestry and wood processing, providing employment of the local population, with a clear concept of forestry development on karst areas that have been passive until now (Meštrović 1977, Matić 1986).

2.1 Wood sampling location and forest fires characteristics Sampling of unfired and fired Aleppo pine wood was carried out by the company »Croatian Forests Ltd.«, in Split Forest Administration – Forest Department Šibenik, management unit Jelinjak. Wood sampling location and forest fire characteristics were as follows: Þ fire type – ground fire (law) and treetops fire (high) Þ fired surface size – 19.25 ha Þ forest breeding Þ high (Allepo pine forest culture) – 3.15 ha (16.36%) Þ low (garrigue) – 14.9 ha (77.40%) Þ agriculture land (perennial crops) – 1.20 ha (6.23% Þ forest description – old Allepo pine forest culture (Pinus halepensis Mill.) 53 years old, the bushy layer and plants ground layer are very rare, ground is very rocky, 10º inclination, 50m altitude Þ sample mark Þ »a« unfired tree Þ »b« fired tree (Fig. 1, 2 and 3).

2.2 Sampling and preparation of wood for chemical analysis For the purpose of studying the influence of different types of forest fires (ground fire – low and treetops fire – high) on the chemical composition of fired (b), and unfired (a) wood of Aleppo pine for comparison, ring samples of 10–30 cm thickness were taken at three different tree heights (fires heights) as follows: Croat. j. for. eng. 39(2018)2

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Fig. 1 Forwarder pulling out the fired Aleppo pine biomass from the fired forest

Fig. 2 Removal of fired Aleppo pine biomass from fire site Þ first ring sample at a height of 0 m (the first ring from the cutting point, next to the stump) Þ second ring sample at a height of 2 m Þ third ring sample at a height of 4 m.

A. Antonović et al.

Fig. 3 Fired Aleppo pine biomass on fire site Rings were taken immediately after cutting trees according to the standard TAPPI T257 cm 02 – Sampling and preparation of wood for analysis, and taken to the laboratory.

After 14 days of drying in the laboratory, wood macroscopic parts were mechanically separated on each ring. This means that bark (B), sapwood (S) and heartwood (H) were separated. The sapwood (S) was additionally divided into two parts and labeled as S1 (the part of sapwood close to the bark) and S2 (the second part of sapwood close to the heartwood) (see Fig. 4). The aim of that separation was to determine the influence of fire heat and temperature penetration into the wood depth on wood chemical composition as a quality property. The first part of sapwood close to the bark (S1) was taken for this research. The results of chemical composition of bark (B) group of the same Allepo pine wood was also taken from the previous

Fig. 4 Unfired wood (a) and fired wood (b) of Allepo pine with wood macroscopic parts Croat. j. for. eng. 39(2018)2 317

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Fig. 5 Schematic view of sample chemical analysis study (Antonović et al. 2017) for comparison with sapwood (S1) results obtained in this research.

Furthermore, after the separation of macroscopic parts of each ring, sapwood (S1) samples were prepared according to previous studies (Antonović et al.

2007, Antonović et al. 2008, Antonović et al. 2010) and in accordance with the laboratory analytical procedure Preparation of Samples for Compositional Analysis (Hames et al. 2008). Air-dried sapwood samples were milled to different particle sizes using a knife-mill Fritsch – Pulverisette 19. After milling, samples were sieved through standardized sieves. The milled particles, which passed the sieve screen of 0.71 mm and remained on the sieve of 0.325 mm, were used for further chemical analysis, due to their ideal particle size for all isolation methods of group chemical composition, as recommended in previous studies. For screening, laboratory electromagnetic sieves shaker Cisa RP.08 (shaking time t=15±1 min) was used.

After sapwood (S1) grinding and sieving, three smaller samples were taken of each criterion, on which all the chemical analysis were performed, and the results are presented as the mean values of these three samples.

2.3 Wood group chemical components isolation Isolation methods for determining the content of the chemical composition of sapwood (S1) sample

318

group, namely ash, accessory materials (extractives), cellulose, hemicellulose (polyoses) and lignin, were conducted in compliance with previous studies (Antonović et al. 2007, Antonović et al. 2008, Sluiter et al. 2005a, Sluiter et al. 2005b, Sluiter et al. 2008). Sample compositional analysis consisted of a series of isolation methods of the main components, which can be schematically presented as shown in Fig. 5. A small portion of the prepared sample was first used to determine the ash content, and the other major part for prior sample extraction (treatment with a solvent mixture of methanol, CH3OH and benzene, C6H6 in the volume ratio 1:1) to remove the accessory materials from the sample which could interfere during further chemical analysis.

Thus, additional residual solid content was determined as a content of accessory materials or extractives. Furthermore, sulfonic acid lignin or Klason’s lignin (by treatment with 72% sulfuric acid, H2SO4) and polysaccharides cellulose (by treatment with a solvent mixture of nitric acid, HNO3 and ethanol, C2H5OH in a volume ratio of 1:4) were isolated from the extracted sample.

The content of hemicellulose (polyose) was determined by calculation according to the share of other mentioned components in the samples. The hemicellulose content was calculated according to following expression: WP = 100 – (% A + % AM + % C + % L) in %. All used chemicals were of high purity (p.a.) and were obtained from commercial sources.

Croat. j. for. eng. 39(2018)2

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

3. Results with discussion Forest fires front is the portion sustaining continuous flaming combustion, where unfired material meets active flames, or the smoldering transition between unfired and fired material. As the front approaches, the fire heats both the surrounding air and woody material through convection and thermal radiation. First, wood is dried as water is vaporized at a temperature of 100 °C. Next, the pyrolysis of wood at 230 °C releases flammable gases. Finally, wood can smolder at 380 °C or, when heated sufficiently, ignite at 590 °C. Even before the flames of a forest fire arrive at a particular location, heat transfer from the forest fires front warms the air from 800 to 900 °C, which pre-heats and dries flammable materials, causing materials to ignite faster and allowing the fire to spread faster (Bakšić et al. 2015).

Forest fires emit a complex mixture of particles and gases into the atmosphere. The diversity in composition of combustion products results from a wide range of wood species, wood chemistry, and fire behavior. The chemical characteristics of wood affect the rate of combustion and influence the overall forest fire behavior. At temperatures above 100 °C, chemical bonds begin to break. The rate at which the bonds are broken increases as the temperature increases.

Between 100 °C and 200 °C, noncombustible products, such as carbon dioxide, traces of organic compounds and water vapor, are produced. Above 200 °C, the cellulose breaks down, producing tars and flammable volatiles that can diffuse into the surrounding environment. If volatile compounds are mixed with air and heated to the ignition temperature, combustion reactions occur. The energy from these exothermic reactions radiates to the solid material, thereby enhancing the combustion reactions. If the burning mixture accumulates enough energy to emit radiation in the visible spectrum, the phenomenon is known as flaming combustion. Above 450 °C, all volatile material is gone. The residue that remains is an activated char that can be oxidized to carbon dioxide, carbon monoxide and water vapor. Oxidation of the char is referred to as afterglow (Alexander 1982, Basin 2011).

The thermal degradation of wood caused by forest fires can be represented as the sum of the thermal degradation reactions of the individual wood chemical components (group chemical composition), namely ash, accessory materials, cellulose, lignin and polyoses (hemicellulose). (Antonović et al. 2016, Krička et al. 2016). The influence of individual chemical components on thermal degradation reactions of sapwood depends on the species and its moisture content, and forest fires exposure period as a function of

A. Antonović et al.

temperature for the components and for wood itself.

The chemical components of wood are thermally degraded at different speeds, and the degradation takes place in the following order: hemicellulose → cellulose → lignin. The degradation of holocellulose, which consists of the cellulose plus hemicelluloses, follows most closely the degradation of wood. Lignin generally degrades at a slower rate than holocellulose, although the degradation period begins somewhat earlier than for the holocellulose. Cellulose and wood appear to degrade at similar rates, although wood begins to degrade at slightly lower temperatures than cellulose but at higher temperatures than holocellulose. This lower degradation temperature of wood is primarily due to hemicelluloses in wood and holocellulose. The wood degradation resembles more closely the degradation pattern of cellulose and holocellulose than the degradation pattern of lignin. This is reasonable because cellulose and holocellulose account for approximately 50% and 75% of wood, respectively (Antonović et al. 2017, Jurišić et al. 2017, Krička et al. 2017).

Table 1 Bark (B) from previous study and average values of sapwood (S1) group chemical composition of unfired and fired Aleppo pine wood Wood chemical component Sample

0

a

2

4

0

b

2

4

A

AM

C

L

WP

%

%

%

%

%

B

3.63

8.61

24.12

49.45

14.18

S1

0.40

1.19

49.41

27.24

21.76

B

3.42

9.58

24.15

49.55

13.30

S1

0.42

1.60

44.51

25.99

27.48

B

5.89

8.37

24.79

49.94

11.01

S1

0.42

1.04

53.85

29.85

14.83

B

2.72

12.47

18.44

46.71

19.67

S1

0.47

2.68

47.11

33.35

16.38

B

2.53

13.17

20.50

47.18

16.61

S1

0.45

1.58

49.40

31.84

16.73

B

3.10

13.17

19.20

47.13

17.40

S1

0.51

1.11

44.81

32.37

21.20

A – ash; AM – accessory materials; C – cellulose; L – lignin; WP – polyoses (hemicellulose); a – unfired wood; b – fired wood; 0 – height 0 m; 2 – height 2 m; 4 – height 4 m; B – bark; S1 – sapwood

Croat. j. for. eng. 39(2018)2 319

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Table 1 and Fig. 6 show the unfired and fired wood bark (B) from previous study (Antonović et al. 2017) for comparison and chemical components contents of the first part of sapwood close to the bark (S1) of the Aleppo pine depending on the sampling heights or forest fires heights. The bark and sapwood were from the same Aleppo pine wood. Comparing the results of the Aleppo pine unfired bark and sapwood with the results of previous researches of different wood species (Antonović et al. 2007), it can be concluded that the distribution of chemical components is similar. When comparing the unfired wood bark with sapwood chemical composition, the distribution is also similar, which means that the bark contains a significantly higher content of ash, accessory materials (extractives) and lignins, and a significantly lower content of polysaccharides cellulose and polyoses (hemicellulose) than sapwood. The same was noticed for fired wood bark in comparison with fired sapwood.

The inorganic content of wood species, usually referred to as its ash content, is an approximate measure of its mineral salts and other inorganic matter content.

The ash content of fired wood bark decreased in comparison to unfired wood bark, which is probably due to rapid water evaporation, where mineral salts were dissolved at high temperatures caused by fires. Furthermore, during the burning in the presence of oxygen and the appearance of flame, a part of inorganic substances were evaporated and thus the loss of mass of inorganic substances was assumed.

Previous research (Antonović et al. 2017) showed that the chemical components of the bark wood cell wall (cellulose, hemicellulose, and lignin) was thermally degraded under the influence of high temperatures caused by forest fires, i.e. the high polymeric compounds were degraded into compounds of low molecular weight. It is assumed that many new low molecular weight compounds, which occurred during the fire and elevated temperatures, due to their chemical character similar to the different accessory materials groups mentioned above, were extracted during Soxhlet extraction together with the original accessory materials and thus joined their total content. The same is explained for re-condensation (re-polymerization) in the above mentioned previous studies. Based on the sapwood results, there is no significant difference in the accessory materials content between unfired and fired wood, not even resulting from different forest fires heights.

Cellulose is principally responsible for the production of flammable volatiles. The decrease in cellulose content occurs through dehydration, hydrolysis, oxidation, decarboxylation and transglycosylation. The

320

primary reaction of the high-temperature pathway is depolymerization. This takes place when the cellulose structure has absorbed enough energy to activate the cleavage of the glycosidic linkage to produce glucose, which is then dehydrated to levoglucosan (1, 6-anhydro-ß-D-glucopyranose) and oligosaccharides. During thermal degradation, water and acids are produced from both hemicellulose and cellulose. The degradation reaction of cellulose is an exothermic reaction, beginning at 240–350 °C, where it is decomposed to anhydrocellulose and levoglucosan. The bark results from the previous study (Antonović et al. 2017) show a significant difference in reduced cellulose content between the unfired and fired wood. Based on the sapwood results, there are no significant differences in the cellulose content between unfired and fired wood, not even resulting from different forest fires heights.

Although belonging to the same group of polysaccharides, wood polyoses (hemicellulose) differ from cellulose by the composition of different sugar units, by a much shorter molecular chain and by molecule chain branching. Hemicelluloses are less stable thermally than cellulose and evolve more noncombustible gases and less tar. As the obtained results show, increased hemicellulose content in the fired wood bark compared with unfired wood bark, and considering that hemicellulose was mathematically calculated, it should be assumed that the increased hemicellulose content is caused by thermal decomposition of other chemical components of the wood bark group chemical composition (cellulose and lignin) into low molecular weight compounds that are joined to the sum of the hemicelulose total content. Based on the sapwood results, there are no significant differences in the hemicellulose content between unfired and fired wood, not even resulting from different forest fires heights.

Thermal degradation of lignin yields phenols from cleavage of ether and carbon–carbon linkages and produces more residual char than does thermal degradation of cellulose. Dehydration reactions around 200 °C are primarily responsible for thermal degradation of lignin and one part of decreased content. The other part of lignin decreased content occurs between 150 °C and 300 °C and cleavage of a- and ß–aryl-alkyl-ether linkages, around 300 °C, aliphatic side chains start splitting off from the aromatic ring, and finally, the carbon-carbon linkage between lignin structural units is cleaved at 370–400 °C. The degradation reaction of lignin is also an exothermic reaction, with peaks occurring between 225 °C and 450 °C. Similarly as in cellulose content, the bark results show a significant difference in reduced lignin content between the unfired and fired wood.

Based on the sapwood results, there are no significant differences in the lignin content between unfired and Croat. j. for. eng. 39(2018)2

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313â&#x20AC;&#x201C;324)

A. AntonoviÄ&#x2021; et al.

Fig. 6 Bark (B) and sapwood (S1) group chemical components content of unfired and fired Aleppo pine wood in dependence on sampling height (forest fire height) Croat. j. for. eng. 39(2018)2 321

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

fired wood, not even resulting from different forest fires heights.

Further research on the same wood samples should analyze the impact of forest fires on the second part of sapwood and heartwood at different tree heights (fire heights) to see if the fire has affected their chemical composition, and to make comparison with the results of the present study.

4. Conclusions The main aim of this paper was to study the group chemical composition of the Aleppo pine (Pinus halepensis Mill.) first part of sapwood close to the bark before and after the impact of low ground fire and high fire of the treetops at tree height of 0, 2 and 4 m.

Therefore, the impact of forest fires on the Allepo pine sapwood group chemical composition was studied in terms of quality for further application in production of biorefinery products. In addition, research results on group chemical composition of unfired and fired Aleppo pine wood bark from previous studies were used for comparison with sapwood from this study.

The obtained results show that the distribution of the main chemical components of Allepo pine unfired wood bark and sapwood is similar to the results of previous studies for different wood species. The distribution is also similar when bark is compared to chemical composition of sapwood of different wood species. This means that the bark contains a significantly higher content of ash, accessory materials (extractives) and lignins, and a significantly lower content of polysaccharides cellulose and polyoses (hemicellulose) than sapwood.

The bark results from previous studies show a significant difference in reduced ash, cellulose and lignin content, and in the increased accessory materials and wood polyoses (hemicellulose) content between the unfired and fired wood. Furthermore, the content of individual chemical components of fired bark at different forest fires heights of 0, 2 and 4m for each sample does not differ significantly, except for ash, which can be explained by phytocenological criteria of different sampling locations. Based on the results, there are no significant differences for sapwood in the chemical composition between unfired and fired wood, not even resulting from different forest fires heights.

It can be concluded that the forest fire did not have any effect on Aleppo pine sapwood. The fired wood bark takes over all the damage caused by the high temperature during the forest fire. It can be assumed that fired sapwood still has all the chemical properties for

322

further application in mechanical or chemical processing. Therefore, in practical terms, it is necessary to ascertain the possible utilization and use of wood biomass that has been exposed to forest fires. This can reduce the extremely high costs of recovery after the forest fires.

The study presented in this paper contributes to a better understanding of the possibility of using fired wood species, as a very large raw material base in the Republic of Croatia, in the further production of biorefinery products.

5. References Alexander, M.E., 1982: Calcualting and interpreting forest fire intesities. Can. J. Bot. 60(4): 349–357.

Antonović, A., Jambreković, V., Pervan, S., Ištvanić, J., Moro, M., Zule, J., 2007: Influence of sampling location on sapwood group chemical composition of beech wood (Fagus sylvatica L.). Drvna industrija 58(3):119–125 Antonović, A., Jambreković, V., Pervan, S., Ištvanić, J., Greger, K., Bublić, A., 2008: A supplement to the research of native lignin of beech sapwood (Fagus sylvatica L.). Wood research (Bratislava). Slovak Forest Product Research Institute, Bratislava, Slovak Republic 53(1): 55–68.

Antonović, A., Jambreković, V., Franjić, J., Španić, N., Pervan, S., Ištvanić, J., Bublić, A., 2010: Influence of sampling location on content and chemical composition of the beech native lignin (Fagus sylvatica L.). Periodicum Biologorum 112(3): 327–332.

Antonović, A., 2010: Kemija drva. Interna skripta, Šumarski fakultet.

Antonović, A., Krička, T., Matin, A., Jurišić, V., Bilandžija, N., Voća, N., Stanešić, J., 2016: Biochar quantification and its properties in relation to the raw material. 51st Croatian and 11th International Symposium on Agriculture, Proceedings, 445–449.

Antonović, A., Krička, T., Matin, A., Voća, N., Jurišić, V., Bilandžija, N., Grubor, M., Stanešić. J., 2017: Lignocellulosic Composition of Some Important Oilseeds and Grains Biomass in the Republic of Croatia. Proceedings SA 2017/Vila, Sonja; Antunović, Zvonko (ur.). Osijek: Poljoprivredni fakultet Sveučilišta Josipa Jurja Strossmayera u Osijeku, 623–626.

Antonović, A., Barčić, D., Ištvanić, J., Medved, S., Podvorec, T., Stanešić, J., 2017: The Forest Fires Impact on Bark Chemical Composition of the Aleppo Pine (Pinus halepensis Mill.).

3rd International Scientific Conference Wood Technology & Product Design, 11–14. September, Ohrid, Republic of Macedonia, 130–142.

Bakšić, N., Vučetić, M., Španjol, Ž., 2015: Potencijalna opasnost od požara otvorenog prostora u Republici Hrvatskoj.

Vatrogastvo i upravljanje požarima, 5(2): 30–41.

Croat. j. for. eng. 39(2018)2

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

A. Antonović et al.

Basin, T., 2011: In: Fire in Mediterranean Ecosystems: ecology, evolution and management, Cambrigde University Press, 515 p.

Pausas, J.G., Llovet, J., Rodrigo, A., Vallejo, R., 2008: Are wildfires a disaster in the Mediterranean basin? – A review.

Int. J. Wildland Fire 17(6): 713–723.

Directive 2008/28/EC of the European Parliament and of the Council of 11 March 2008 amending Directive 2005/32/EC establishing a framework for the setting of ecodesign requirements for energy-using products, as well as Council Directive 92/42/EEC and Directives 96/57/EC and 2000/55/ EC, as regards the implementing powers conferred on the Commission.

Pentek, T., 1998: Forest fire prevention roads as a special category of forest roads and factors that influence their distribution in space. Glasnik za šumske pokuse, 35: 93–141.

European Biomass Association – AEBIOM, 2017: Statistical Report 2017 – European Bioenergy Outlook. Full report, 12–31.

Pentek, T., Nevečerel, H., Ecimović, T., Lepoglavac, K., Papa, I., Tomašić, Ž., 2014: Strategic planing of forest road network in Croatia – analysis of present situation as basis for future activities. Nova mehanizacija šumarstva 35(1): 63–78.

Espelta, J.M., Rodrigo, A., Habrouk, A., Meghelli, N., Ordoñez, J.L., Retana, J., 2002: Land use changes, natural regeneration patterns and restoration practices after large wildfire in NE Spain: challenges for fire ecology landscape restoration. In: Trabaud, L., Prodon, R., (eds) Fire and biological processes. Backhuys Publishers, Leiden.

Hames, B., Ruiz, R., Scarlata, C., Sluiter, A., Sluiter, J., Templeton, D., 2008: Preparation of samples for compositional analysis. Laboratory analytical procedure. National renewable energy laboratory, 1–9.

International Energy Agency – IEA, 2017: IEA Bioenergy – Annual Report 2017, 87–100.

Jurišić, V., Voća, N., Bilandžija, N., Krička, T., Antonović, A., Grubor, M., Matin, A., Kontek, M., 2017: Pirolitička svojstva važnijih energetskih kultura u RH. Proceedings SA 2017/ Vila, Sonja; Antunović, Zvonko (ur.). Osijek: Poljoprivredni fakultet Sveučilišta Josipa Jurja Strossmayera u Osijeku, 651–655.

Krička, T., Jurišić, V., Matin, A., Bilandžija, N., Antonović, A., 2016: Mogućnosti pretvorbe i iskorištenja ostataka poljoprivredne biomase nakon procesa pirolize. 51st Croatian and 11th International Symposium on Agriculture. Proceedings, 485–489.

Krička, T., Voća, N., Jurišić, V., Matin, A., Bilandžija, N., Antonović, A., Slipčević, D., 2017: Pretvorba poljoprivrednih ostataka i energetskih kultura u energiju i proizvode dodane vrijednosti – bioulje i biougljen. Proceedings SA 2017/Vila, Sonja; Antunović, Zvonko (ur.). Osijek: Poljoprivredni fakultet Sveučilišta Josipa Jurja Strossmayera u Osijeku, 659–663.

Matić, S., 1986: Šumske kulture alepskog bora i njihova uloga u šumarstvu Mediterana. Glasnik za šumske pokuse pos. izd. 2: 125–145.

Meštrović, Š., 1977: Značaj šumskih kultura u primorskom području krša. Šumarski list 101(8–9): 382–385.

Moreira, F., Duarte, I., Catry, F., Acácio, V., 2007: Cork extraction as a key factor determining post-fire cork oak survival in a mountain region of southern Portugal. Forest Ecology and Management 253(1–3): 145–162.

Pentek, T., Pičman, D., Nevečerel, H., Lepoglavac, K., Papa, I., Potočnik, I., 2011: Primary forest opening of different relief areas in the Republic of Croatia. Croatian Journal of Forest Engineering 32 (1): 401–416.

Prgin, D., 2005: Alepski bor (Pinus halepensis Mill.) prvorazredna vrsta za podizanje šuma na mediteranskom kršu. Šumarski list 129 (1–2): 71–80.

Rodrigo, A., Retana, J., Pico, X., 2004: Direct regeneration is not the only response of Mediterranean forests to large fires.

Ecology 85(3): 716–729.

Saracino, A., Pacella, R., Leone, V., Borghetti, M., 1997: Seed dispersal and changing seed charachteristics in a Pinus halepensis Mill. Forest after fire. Plant Ecol. 130(1): 13–19.

Sluiter, A., Hames, B., Ruiz, R., Scarlata, C., Sluiter, J., Templeton, D., 2005a: Determination of ash in biomass. Laboratory analytical procedure. National renewable energy laboratory, 1–5 p.

Sluiter, A., Ruiz, R., Scarlata, C., Sluiter, J., Templeton, D., 2005b: Determination of extractives in biomass. Laboratory analytical procedure. National renewable energy laboratory, 1–12 p.

Sluiter, A., Hames, B., Ruiz, R., Scarlata, C., Sluiter, J., Templeton, D., Crocker, D., 2008: Determination of structural carbohydrates and lignin in biomass. Laboratory analytical procedure. National renewable energy laboratory, 1–16 p.

Strezov, V., Evans, T.J., 2015: Biomass Processing Technologies. CRC Press. Španjol, Ž., 1996: Biološko–ekološke i vegetacijske posljedice požara u borovim sastojinama i njihova obnova. Disertacija, Šumarski fakultet Sveučilišta u Zagrebu, Zagreb. Španjol, Ž., Rosavec, R., Barčić, D., Galić, I., 2011: Zapaljivost i gorivost sastojina alepskog bora (Pinus halepensis Mill.

Croatian Journal of Forest Engineering 32(1): 121–129.

Tapias, R., Climent, J., Pardos, J.A., Gil, J., 2004: Life histories in Mediterranean pines. Plant Ecol. 171(1–2): 53–68.

Thomas, P.A., McAlpine, R., 2010: Patterns of fire over time.

In: Fire in the Forest, Cambrigde University Press, 225 p.

TAPPI T257 cm-02 – Sampling and preparing wood for analysis. TAPPI. TAPPI T 264 cm-97 – Preparation of Wood for Chemical Analysis. TAPPI.

Croat. j. for. eng. 39(2018)2 323

A. Antonović et al.

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Authors’ addresses:

Received: May 15, 2018 Accepted: May 27, 2018

324

Assoc. prof. Alan Antonović, PhD. * e-mail: alan.antonovic@zg.htnet.hr Assoc. prof. Damir Barčić, PhD. damir.barcic@zg.htnet.hr Assoc. prof. Jaroslav Kljak, PhD. e-mail: jkljak@sumfak.hr Assist. prof. Josip Ištvanić, PhD. e-mail: jistvanic@sumfak.hr Tomislav Podvorec e-mail: t.podvorec.55@gmail.com Juraj Stanešić, uni.bacc.techn.lign. e-mail: juraj.stanesic@gmail.com University of Zagreb Faculty of Forestry Svetošimunska cesta 25 10002 Zagreb CROATIA * Corresponding author Croat. j. for. eng. 39(2018)2

A. Antonović et al.

CONTENTS

The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass ... (313–324)

Original scientific papers Okey Francis Obi, Rien Visser Including Exogenous Factors in the Evaluation of Harvesting Crew Technical Efficiency using a Multi-Step Data Envelopment Analysis Procedure

153

Eduardo Tolosana, Raffaele Spinelli, Giovanni Aminti, Rubén Laina, Ignacio López-Vicens Productivity, Efficiency and Environmental Effects of Whole-Tree Harvesting in Spanish Coppice Stands Using a Drive-to-Tree Disc Saw Feller-Buncher

163

Piotr S. Mederski, Mariusz Bembenek, Zbigniew Karaszewski, Zenon Pilarek, Agnieszka Łacka Investigation of Log Length Accuracy and Harvester Efficiency in Processing of Oak Trees

173

Zbigniew Karaszewski, Agnieszka Łacka, Piotr S. Mederski, Mariusz Bembenek Impact of Season and Harvester Engine RPM on Pine Wood Damage from Feed Roller Spikes

183

Omar Mologni, Peter Dyson, Dzhamal Amishev, Andrea Rosario Proto, Giuseppe Zimbalatti, Raffaele Cavalli, Stefano Grigolato Tensile Force Monitoring on Large Winch-Assist Forwarders Operating in British Columbia

193

Václav Štícha, Jaroslav Holuša, Roman Sloup, Jan Macků, Jiří Trombik A Mobile Hydraulic Winch for Use in Small-Scale Forestry

205

Nopparat Kaakkurivaara, Tomi Kaakkurivaara Productivity and Cost Analysis of Three Timber Extraction Methods on Steep Terrain in Thailand

213

Matevž Mihelič, Raffaele Spinelli, Anton Poje Production of Wood Chips from Logging Residue under Space-Constrained Conditions

223

Milan Marusiak, Jindřich Neruda Dynamic Soil Pressures Caused by Travelling Forest Machines

233

Ahmad Solgi, Ramin Naghdi, Eric R. Labelle, Petros A. Tsioras, Ali Salehi Comparison of Sampling Methods Used to Evaluate Forest Soil Bulk Density

247

Ehsan Abdi Root Tensile Force and Resistance of Several Tree and Shrub Species of Hyrcanian Forest, Iran

255

Mohammad Javad Heidari, Akbar Najafi, Seyedjalil Alavi Pavement Deterioration Modeling for Forest Roads Based on Logistic Regression and Artificial Neural Networks

324

271

Matija Landekić, Ivan Martinić, Matija Bakarić, Tibor Pentek, Tomislav Poršinsky, Mario Šporčić Current State and Improvement Potential of Forestry Workers Training in Croatia

289

Karlo Beljan, Stjepan Posavec, Jura Čavlović, Krunoslav Teslak, Thomas Knoke Economic Consequences of Different Management Approaches to Even-Aged Silver Fir Forests

299

Alan Antonović, Damir Barčić, Jaroslav Kljak, Josip Ištvanić, Tomislav Podvorec, Juraj Stanešić The Quality of Fired Aleppo Pine Wood (Pinus Halepensis Mill.) Biomass for Biorefinery Products

313

Croat. j. for. eng. 39(2018)2