CROJFE - Volume 31, Issue 2

31

Issue 2

2010

2

Original scientific paper – Izvorni znanstveni rad

Development of a Multi-Attribute Spatial Decision Support System in Selecting Timber Harvesting Systems Martin Kühmaier, Karl Stampfer Abstract – Nacrtak Strategic and tactical decisions in timber harvesting planning have long-term consequences on the further development of forests. Decisions about harvesting activities are often based on intuition and the consequences of these actions cannot be determined exactly. A GIS based evaluation model was designed to support the timber harvesting decision making process. It compares harvesting systems and selects the best suitable system in consideration of stakeholder interests and environmental conditions. The developed model is made up of four stages.

First, the area of interest is defined. Then, a technological evaluation of harvesting systems capability determines their compatibility with location factors. Only acceptable systems are included into the third stage, the utility analysis. Using evaluation criteria, it transforms them into comparable values and ranks these values. The last stage of the model provides a metric that estimates consequences of different treatment scenarios. The main processes have been automated in ESRI® ArcGIS by using ModelBuilder™ extension. The model has been demonstrated in a 1100 ha sized forest enterprise in steep terrain in the South of Lower Austria. One scenario determined the possible benefits of implementing »cable forwarders« as new harvesting technology. Five of seven criteria could be improved; including a reduction of stand damage by 2 percent points and an increase of contribution margin from 40 to 46 /m³. Improving forest road network generated a positive effect on productivity and fuel consumption, but the overall economic benefit was too low to recommend the construction of the road. The model suggests that a combination of increasing forest road density and technology improvement could lead to tripling productivity, increasing contribution margin from 40 to 56 /m³ and lowering the damage rate by 53% and injury rate by 93%. This example shows that this SDSS can help the user to determine the best suitable alternatives.

Keywords: timber harvesting, forest road network, decision support, evaluation model, utility analysis

1. Introduction – Uvod Timber harvesting is often one of the main objectives of forest management. It increases the contribution margin to the forest enterprise, but can also have positive effects on long-term ecological and social values. Efficient harvesting operations are based on a well established forest road network, best suitable equipment and machines and experienced forest workers (Stampfer 2010). Decisions in selecting these items are mostly based on experience and intuition and often do not consider a long-term and sustainable strategy of resource management (Lüthy 1998). Such a decision making process can cause difCroat. j. for. eng. 31(2010)2

ficulty when reacting to change, e. g. in harvesting volume or technology. Admittedly, the increase in production costs and the development of new technologies presupposes a continuous review of the systems used.

To estimate the effects of changes in management a decision support tool is helpful, especially for tactical and strategic goals. Changing forest road network or using different harvesting systems can have large consequences on costs, ecological and social impacts and machine and work force capacity. Until now, economic efficiency has been the most important criterion for selecting harvesting systems (Lüthy 1998,

75

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

Meyer et al. 2001, Lubello 2008). The non-consideration of ecological and social criteria may impose negative side-effects and risks that revoke the economic advantage. Therefore, a well-grounded analysis of harvesting systems should take stand and terrain data as well as ecological, economic and social impacts into account (Mendoza 1989, Næsset 1997, Sheppard et al. 2005, Wolfslehner et al. 2008, Kangas et al. 2008). Since this decision problem consists of several criteria and bears many trade-offs, a satisfactory solution can hardly be found without using technical and mathematical tools. For that reason a multi-criteria, computer-aided DSS is a good approach (Vacik and Lexer 2001, Lexer et al. 2005, Kangas and Kangas 2005). Harvesting operations are carried out at a spatial level and are best considered using GIS technology. Equipment and work force have to be transported to the operation area and the harvested timber will be transported from the stand to the saw mill. The accessibility of the forest area depends on existing infrastructure and the roughness of the terrain.

In recent years some studies have been published to estimate best suitable harvesting systems on the basis of forest districts or compartments. Lüthy (1998) focused on the development of a SDSS concerning harvesting system evaluations in steep terrain. The case study included a technological evaluation and rough cost estimation. Yoshioka and Sakai (2005) analyzed the amount and availability of forest biomass as an energy resource in mountainous regions. This study was based on a GIS analysis including three machinery types (skidder, tower yarder, and sledge yarder) and three biomass resources (logging residues, thinned trees, and broadleaved forests). The resources with the lowest procurement costs were selected. Lubello (2008) implemented a GIS-based SDSS for extracting operations. The model outputs show feasible working areas of each system (skidder, forwarder, cable forwarder, tower yarder, sledge yarder), and the technical and optimized distribution of systems with costs. A similar approach was made by Adams et al. (2003) for 500 hectares of mountainous terrain in south–west Virginia. They analyzed harvest system allocations for wheeled skidder, tracked skidder, cable yarder and helicopter. In Austria a technological evaluation of harvesting systems based on stand and terrain data has been carried out by Mallinger (2002). Nevertheless, none of these studies took ecological or social criteria into account, which is essential for a comprehensive analysis of the impacts of harvesting operations.

The aim of this study is to develop a SDSS for identifying best suitable harvesting systems and to estimate ecological, economical and social conse-

76

quences of timber harvesting operations. The main focus of the model is on improving forest road networks and/or implementing new harvesting technologies.

2. Materials and methods – Materijal i metode 2.1 SDSS architecture – Arhitektura SDSS-a 2.1.1 Development approach – Pristup razvoju A master model was developed that involved iterative communication and negotiating among the users; the decision analyst and the software engineer that together define the decision scope and decision-making process (Lexer et al. 2005). This master model combines different but related aspects of the DSS-development. For example, the process model represents the flow of data and information throughout the modelled planning and decision-making process, and describes the exchange of information among various DSS components. The formal model includes the algorithms, rules, and mathematical equations needed to formally describe the modelled system. Finally, the software engineer has to create the implementation model, which comprises software architecture and technical solutions to implement the master model (Lexer et al. 2005). Based on iterative discussions and a negotiation process, the major processes of the master model were developed (Fig. 1).

2.1.2 Dividing the decision problem – Podjela problema odlu~ivanja The basic decision problem can be divided into two parts: Þ (a) which harvesting systems are suitable at particular locations within a defined project area, Þ (b) which harvesting system is the best suitable when including considerations of economic, ecological and social effects.

The master model was designed so that both questions could be analyzed successively but within the same general analytical process. The decision process includes a set of medium to long-term objectives for the management of timber harvesting. The evaluation process within the master model is made up of four stages. First, the investigation area has to be defined. In the next step a technological evaluation of harvesting systems is implemented, where the capability of harvesting systems is determined by comparing their specification data with location factors (Löffler 1984). Concordant systems are inCroat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

M. Kühmaier and K. Stampfer

Fig. 1 Schematic representation of the main processes of the master model Slika 1.Shematski prikaz glavnih procesa glavnoga modela cluded into the third stage, the utility analysis. The analysis calculates the best suitable system by considering evaluation criteria, transforming them into comparable values, and aggregating and ranking these values. The last stage of the model analyses the consequences of the harvesting program for different scenarios.

2.1.3 Analysis mask – Podru~je analize For estimating potential timber harvesting areas, desirable zones (e. g. forests) have been intersected with non-desirable (protected or prohibited areas).

This combination of data layers generated the analysis mask for technological evaluation and utility analysis.

2.1.4 Creating technological layers – Odre|ivanje tehnolo{kih pokazatelja 10 different harvesting systems have been taken into account for the technological evaluation. They differ in four grades of mechanization and three Croat. j. for. eng. 31(2010)2

working methods (Stampfer 2002). For the technological evaluation, four criteria have been chosen.

They act as specification data for the applicability of the selected harvesting system under given site conditions. The slope, expressed in %, is a limiting factor for wheeled (30%) and tracked (60%) machines. The given limits are average values; they can vary depending on relief and soil bearing capacity. The extraction distance is a limiting factor for cable-supported machines, e. g. tower yarders (400 m) and skidders (100 m). The limiting DBH for harvester and processor depends on the type of harvesting head. A strongly varying morphology is a restricting factor for ground-based systems as a result of reduced trafficability. According to the extraction operation, harvesting systems can be divided into cut-tolength (CTL), tree-length (TL) and whole-tree (WT) methods (Table 1).

By combining stand and terrain conditions with the equipment specifications, a »technology layer« has been calculated for every harvesting system. To

77

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

Table 1 Harvesting systems and equipment subject to stand and terrain data Tablica 1. Sustavi pridobivanja drva i strojevi koji se odnose na sastojinske i terenske podatke

1 2 3 4 5 6

Harvesting System – Sustav pridobivanja drva Chain saw & Wood-pick, CTL Motorna pila i capin; sortimentna metoda Chain saw & Forwarder, CTL Motorna pila i forvarder; sortimentna metoda Chain saw & Cable Forwarder, CTL Motorna pila i forvarder s vitlom; sortimentna metoda Chain saw & Skidder, TL Motorna pila i skider; deblovna metoda Wheeled Harvester & Forwarder, CTL Kota~ni harvester i forvarder, sortimentna metoda Tracked Harvester & Tower Yarder, CTL Gusjeni~ni harvester i stupna {umska `i~ara; sortimentna metoda

Tracked Harvester & Cable Forwarder, CTL Gusjeni~ni harvester i forvarder s vitlom; sortimentna metoda Chain saw & Tower Yarder, CTL 8 Motorna pila i stupna {umska `i~ara; sortimentna metoda Chain saw & Tower Yarder & Processor, WT 9 Motorna pila, stupna {umska `i~ara i procesor; stablovna metoda Chain saw & Helicopter & Processor, TL 10 Motorna pila, helikopter i procesor; deblovna metoda 7

Technological Specification – Tehnolo{ke zna~ajke Slope 30–60%, Terrain accessible Nagib terena 30–60 %, pristupa~an teren Slope < 30%, Terrain accessible Nagib terena < 30 %, pristupa~an teren Slope < 60%, Terrain accessible Nagib terena < 60 %, pristupa~an teren Slope < 30%, Terrain accessible Nagib terena < 30 %, pristupa~an teren Slope < 30%, Terrain accessible, DBH max. 40 cm Nagib terena < 30 %, pristupa~an teren, prsni promjer maks. 40 cm Slope < 60%, Extraction distance < 800 m, Terrain accessible, DBH max. 40 cm Nagib terena < 60 %, srednja udaljenost privla~enja < 800 m, pristupa~an teren, prsni promjer maks. 40 cm Slope < 60%, Terrain accessible, DBH max. 40 cm Nagib terena < 60 %, pristupa~an teren, prsni promjer maks. 40 cm Slope < 100%, Extraction distance < 400 m Nagib terena < 100 %, pristupa~an teren, srednja udaljenost privla~enja < 400 m Slope < 100%, Extraction distance <400 m Nagib terena < 100 %, srednja udaljenost privla~enja < 400 m –

move machines of ground-based systems to the utilization area the harvesting sites have to be accessible. This means that machines are able to drive to the harvesting site. If not, and generally usable zones were surrounded by non-usable ones, these areas have been shifted to the next possible technology layer. Furthermore climate data could be considered to determine periods without the possibility of carrying out harvesting operations as a result of high snow cover, and to estimate advantageous periods for trafficability caused by frozen underground. The technology layers act as input data for the next step – the utility analysis.

each of these criteria. The task is now to aggregate these utility functions to describe the overall utility of the alternatives. This aggregation is done by weighting of the criteria in the utility function with respect to their importance. The relations between the weights of different criteria describe the tradeoffs between the criteria (Kangas et al. 2008). The best suitable alternative is the one with the highest overall utility. The most applied multi-attribute utility function is the linear additive utility function.

2.1.5 A multiple criteria utility model to evaluate alternatives – Vi{ekriterijski model korisnosti za procjenu zamjenskih rje{enja

where: Ui describes the overall utility of alternative i (or priority of alternative i) cij is the performance of alternative i with respect to criterion j and aj is the importance weight of criterion j.

To evaluate the overall utility of decision alternatives for cases where there is more than one possible solution, an approach borrowed from multiple-attribute utility theory (MAUT) was adopted. This method requires the mathematical characterization of the preferences of the decision maker over a set of attributes (Goicoechea et al. 1982). In a case of MAUT, it is assumed that there are a certain number of criteria (m) and a unidimensional utility function for

78

m

Ui = ∑ aj c ij j =1

(1)

In this equation, it is assumed that the criteria values cij are already in utility scale or are scaled with a value function. Typically it is required that: m

∑a j =1

j

=1

(2)

Croat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

otherwise the utility could always be increased by increasing the weights. The tradeoffs between criterion k and k’ can be calculated from the ratio of the weights ak/ak’. In general, the marginal rate of substitutions between criteria k and k’ can be calculated as a ratio of partial derivatives of the utility function as U' k a k l= (3) = U' k ' a k ' This means that the decision maker is willing to give up units of criterion k’ in order to increase the value of criterion k by one (Kangas et al. 2008).

Evaluation criteria should be independent from each other, i.e. one goal does not influence the performance of another goal. Indicators are variables, which indicate the status of criteria. For the evaluation model, ecological criteria (impacts on soil, global warming potential, stand damage), economic criteria (contribution margin, relocation time) and social criteria (employment, working safety) have been chosen. Criteria (bold) and indicators are presented in Table 2.

The calculation of the absolute values of the criteria either depends on machines and/or system (e.g. impacts on soil, stand damage, working safety); or on

M. Kühmaier and K. Stampfer

both system and stand data (all other criteria). The latter are based on productivity models, which also include a mode for tree volume, slope and extraction distance (Stampfer 2002, Kühmaier 2010). The criteria values have been scaled by preference functions.

There are several methods for estimating preference functions. In this study the natural scale values have been scaled with score range procedure for all data within the project area (Kangas et al. 2008). vi = (ci – min(c)) / (max(c) – min(c))

(4)

The best alternative is assumed to have a value of one, and the worst the value zero. In this case, if min (c) > 0, the alternatives do not follow a ratio scale, but an interval scale. Interval scale can be interpreted as local scale, the length of the interval depends on specific planning situation (Kainulainen et al. 2007). As an example, the linear preference function for the criterion »contribution margin« has a value of zero below –20 /m³ and a value of one above 100 /m³ (Fig. 2).

In direct weighting methods, which have been used for this study, the estimation is based on direct questions concerning the importance of criteria in the decision situation at hand. SMART and AHP are popular direct methods (von Winterfeldt and Ed-

Table 2 Harvesting systems and equipment subject to stand and terrain data Tablica 2. Sustavi pridobivanja drva i strojevi koji se odnose na sastojinske i terenske podatke Criterion – Uvjeti Indicator – Pokazatelji Impacts on Soil – Utjecaj na tlo Bearing Pressure in kPa – Nosivost tla, kPa

Input Data – Ulazni podaci Machine Weight – Te`ina stroja Tyre Dimension – Dimenzija guma

Global Warming Potential – Potencijal globalnoga zagrijavanja Fuel Consumption in kg CO2-Equivalent – Potro{nja goriva u kg ekvivalent CO2

Fuel Consumption – Potro{nja goriva System Productivity – Proizvodnost sustava

Stand Damage – O{te}enost sastojine Damage on Remaining Stand in % – O{te}enost preostaloga dijela sastojine, %

Damage per Harvesting System O{te}enost po sustavu pridobivanja drva

Contribution Margin – Kontribucijska mar`a Contribution Margin in /m³ – Kontribucijska mar`a, /m³ Relocation Time – Vrijeme premje{tanja Aggregation of Harvesting Areas in % – Zbroj po sje~inama, % Employment – Zaposlenost Demand in Work Force in h/m³ – Potreba za radnom snagom, h/m3 Working Safety – Sigurnost pri radu Accidents in n/m³ – Broj nezgoda, n/m3

Croat. j. for. eng. 31(2010)2

Source/Calculation Izvor/Izra~un Rowland 1972 Maclaurin 2000 Suvinen 2006 Nordfjell et al. 2003 Klvac et al. 2009 Berg & Lindholm 2005 Stampfer 2002 Limbeck-Lilienau 2004 Wratschko 2006

Revenues – Prihodi Hourly System Costs – Tro{kovi sustava po satu System Productivity – Proizvodnost sustava Technological Layers Tehnolo{ki pokazatelji Demand in Work Force Potreba za radnom snagom System Productivity – Proizvodnost sustava

See chapter 2.1.4

Injury rates – Broj ozljeda

Manwaring et al. 1998 Jänich 2009 Eiwegger 2009

Sterba 1983 Stampfer 2009

Stampfer 2009 Stampfer 2010

79

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

Fig. 2 Value function for »contribution margin« Slika 2. Vrijednosna funkcija za kontribucijsku mar`u wards 1986, Saaty 1977). The overall utility for each alternative (harvesting system) and for each subarea has been calculated with Equation (1). The best suitable harvesting system is the one with the highest overall utility calculated for each subarea (e.g. stand, raster cell). The spatial allocation of the best suitable harvesting systems can be viewed directly on the screen.

2.1.6 Analysis and comparison of treatment scenarios – Analiza i usporedba scenarija postupaka The overall effects of the evaluation process have been calculated by spatial aggregation of the evaluation criteria. For the aggregation, only the values of the best suitable harvesting systems (estimated by utility analysis) within the project area have been included for a certain planning period. The evaluation model within the SDSS enables the user to calculate the benefits to climate protection by reduction of greenhouse gas emissions, the contribution to the enterprise profit, the contribution to full employment by increasing labour utilization, injury quotas, equipment and labour relocation time, and the demand of equipment and workforce. These data could also be used as an index for the evaluation of the quality of several harvesting treatments.

2.2 Model implementation – Primjena modela GIS software was used to implement the model.

The main processes have been automated in ESRI® ArcGIS by using the ModelBuilder™ extension. Sup-

80

porting calculations have been carried out in Microsoft® Excel and Microsoft® Access. The necessary analogue data has been digitised and together with the digital information they have been harmonized in GIS by using the same projection and connecting them by primary index. Analyses in GIS are based on raster calculations. The configuration of all the calculations is composed of modules that have been generated with the ModelBuilder™ extension based on Python scripts. ModelBuilder™ is an application in ArcGIS that allows creation, editing and management of models. Models give the possibility to automate the workflow and to execute calculations multiple times. The idea behind using this calculation is to make calculations easier, to chain together workflows by using the output of one tool as the input to another tool, but also to have some possibilities to check the intermediary results. The created models have been implemented in ESRI® ArcToolbox. The GUI is similar to the standard software ESRI® ArcGIS, but with additional features and a help function. The models can be executed in the Toolbox using its dialog or the Command Line window.

2.3 Project area – Podru~je istra`ivanja The SDSS was demonstrated for a region of approximately 1100 ha in the South of Lower Austria (15°39’ longitude East, 47°52’ latitude North). This region is called Tiefental with a main elevation of 800 m. According to Kilian et al. (1994) beech forests (Fagus sylvatica L.) with fir (Abies alba Mill.), sycamore (Acer pseudoplatanus L.) and ash, spruce-fir-beech forests (Picea abies L.) and spruce-fir forests with oak (Quercus robur L.) are the natural vegetation composition. On shallow and exposed dolomite soils, pine forests (Pinus sylvestris L.) are expected. The climate is characterized by cold winters with average January temperatures of –2.6° C, and hot summers with average July temperatures of 15.5° C. The annual average temperature is 6.5° C. The average annual precipitation is about 1300 mm. Depending on the sea level, the duration of the snow cover ranges from 50 to 140 days. 93% of the area is located on calcareous sites and 7% on recent landfills. 10.5% of the forest area is provided on flat terrain (< 30%), and 23% is located in steep terrain (> 60%). Current forests in the project area are dominated by Norway spruce (59.5%) and Scots pine (24.7%) and European larch (7.1%) and broadleaved trees (8.3%) with less importance as crop species. The utilization method is based on small-area operations and single tree forest management systems. Natural regeneration is preferred. The annual cut is about 5350 m³ of timber, transported on forest roads with a density of approximately 34.8 running meters per ha.

Croat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

M. Kühmaier and K. Stampfer

Table 3 Criteria weighting Tablica 3. Te`inski faktori obilje`ja Criterion Obilje`je

Impacts on Soil Utjecaj na tlo

Global Warming Potential Potencijal globalnoga zagrijavanja

Stand Damage O{te}enost sastojine

Weight Te`ina

5%

10%

15%

3. Results – Rezultati The evaluation process analyzed the best suitable harvesting systems for four scenarios within the Tiefental region: Þ (a) before implementing cable forwarder technology and before improving forest road network, Þ (b) after implementing cable forwarder technology, Þ (c) after improving forest road network, Þ (d) combining b and c.

Scenario a is used as a zero option and will be compared with all other scenarios. For the Tiefental SDSS the weighting of the criteria has been done in the following way: Contribution margin is the most important criteria, followed by working safety and stand damage. Global warming potential, employment, impacts on soil and relocation time have minor importance with a weighting factor of 5 to 10% (Table 3). These preferences have been developed together with the forest managers of Tiefental region.

Contribution Margin Relocation Time Vrijeme Kontribucijska mar`a premje{tanja 35%

Employment Zaposlenost

Working Safety Sigurnost pri radu

10%

20%

5%

3.1 Scenario b: Implementing cable forwarder technology – Scenarij b: Primjena forvardera s vitlom The project area is characterised by steep terrain so that wheel-based systems can hardly be implemented. Only in some small flat parts in the Northern region, »harvester-forwarder« technology can be used. The potential of extracting timber with a forwarder is about 6% of the project area, but the possibility of extracting timber with a skidder to the forest road increases the potential harvesting area for the system »chainsaw-skidder« up to 79%. Hand delivery could be implemented on moderately sloped areas that cover 56% of Tiefental region. Tracked harvester in combination with cable forwarder might be used on 665 ha. As a result of excellent road density within the project area, tower yarders could be used in nearly all areas of Tiefental (Table 4). Advantageous periods for trafficability caused by frozen underground and no or low snow cover comprises approximately three weeks from the end of November till mid-December. From mid-December till the end

Table 4 Potential harvesting areas based on technological evaluation for scenario b Tablica 4. Mogu}e sje~ine zasnovane na tehnolo{koj procjeni scenarija b Technological layer – Tehnolo{ki pokazatelj

Potential Harvesting Area Mogu}e podru~je pridobivanja drva ha

%

Chain saw & Helicopter & Processor – Motorna pila, helikopter i procesor

1098

100

Chain saw & Tower Yarder (& Processor) – Motorna pila, stupna {umska `i~ara i procesor

1091

99

Chain saw & Cable Forwarder – Motorna pila i forvarder s vitlom

677

62

Tracked Harvester & Tower Yarder/Cable Forwarder – Gusjeni~ni harvester i stupna {umska `i~ara/forvarder s vitlom

665

61

Chain saw & Wood-pick – Motorna pila i capin

616

56

Chain saw & Skidder – Motorna pila i skider

866

79

There of skidding from forest road – Od toga po tlu kretnim sustavima

805

73

Chain saw & Forwarder – Motorna pila i forvarder

61

6

Wheeled Harvester & Forwarder – Kota~ni harvester i forvarder

49

4

Croat. j. for. eng. 31(2010)2

81

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

Table 5 Harvesting volume before and after implementing cable forwarder technology Tablica 5. Sje~ivi obujam prije i poslije uvo|enja forvardera s vitlom System – Sustav

Scenario a – Scenarij a

Wheeled Harvester & Forwarder – Kota~ni harvester i forvarder Chain saw & Wood-Pick – Motorna pila i capin Tracked Harvester & Cable Forwarder – Gusjeni~ni harvester i forvarder s vitlom Chain saw & Cable Forwarder – Motorna pila i forvarder s vitlom Tracked Harvester & Tower Yarder – Gusjeni~ni harvester i stupna {umska `i~ara Chain saw & Skidder – Motorna pila i skider Chain saw & Tower Yarder & Processor – Motorna pila, stupna {umska `i~ara i procesor Chain saw & Helicopter & Processor – Motorna pila, helikopter i procesor

of March harvesting is normally not possible because of too high snow cover.

843 ha of forest covered area and an average volume of 5354 m³/year are intended for harvesting operations within the next ten years in the Wittgenstein region. Cable forwarders are forwarders equipped with a winch, which increases traction control during extraction operations. The model suggests that with this technology the range of application for forwarders has been boosted up to more than 60%. Given the steep terrain in the project area, extracting with tower yarder was the most favourable extraction operation. 90% of timber was to be harvested with this technology. After introducing cable forwarder technology, the composition of the best suitable harvesting systems, as suggested by the model output, has changed dramatically within regions with slope range of 30 to 60%. 56% of the potential harvesting volume could then be harvested by »tracked harvester & cable forwarder« (Table 5).

After this technological innovation in areas with slope < 60%, all other harvesting systems will be almost fully replaced by cable forwarders (Fig. 3).

Scenario b – Scenarij b

183 m³

3%

3 m³

0%

38 m³

1%

0 m³

0%

0 m³

0%

2983 m³

56%

0 m³

0%

14 m³

0%

2782 m³

52%

3 m³

0%

241 m³

5%

241 m³

5%

2051 m³

38%

2051 m³

38%

59 m³

1%

59 m³

1%

5354 m³

100%

5354 m³

100%

The applicability of cable forwarders in areas with slopes < 60% might be explained by fewer impacts on the remaining stand because of a more careful extraction process, higher system productivity and no setup times, fewer greenhouse gas emissions, higher contribution margin because of lower harvesting costs, fewer equipment rotation times, and higher working safety as a result of fully mechanized harvesting systems. On the other hand, the impacts on the soil increase because of higher bearing pressure of heavy cable forwarders and less people could be employed for harvesting the same timber volume. In the Tiefental region the productivity has only slightly increased after implementing cable forwarder technology, with contribution margin increased by 6 /m³.

By reasons of the increased application of fully mechanised systems the injury rate could be decreased by 36%. On the other hand employment effects are also decreasing by 35%. After taking into account all criteria, the implementation of cable forwarder technology looks favourable (Table 6).

In this example technologically highly developed systems are more preferred than partially mecha-

Table 6 Impacts before and after implementing cable forwarder technology Tablica 6. Utjecaji prije i poslije uvo|enja forvardera s vitlom Indicator – Pokazatelji Productivity – Proizvodnost Bearing Pressure – Nosivost tla Fuel Consumption – Potro{nja goriva Stand Damage – O{te}enost sastojine Contribution Margin – Kontribucijska mar`a Demand in Work Force – Potreba za radnom snagom Injury Rate – U~estalost ozljeda

82

Scenario a – Scenarij a 7 m³/h 50 kPa 4.91 kg CO2/m³ 29% 40 /m³ 0.51 h/m³ 49.48/million m³

Scenario b – Scenarij b 8 m³/h 200 kPa 4.83 kg CO2/m³ 27% 46 /m³ 0.33 h/m³ 31.49/million m³

Variation – Varijabilnost +14% +300% –2% –7% +15% –35% –36%

Croat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

M. Kühmaier and K. Stampfer

Fig. 3 Example of the Model map showing spatial distribution of best suitable harvesting systems after implementing cable forwarder technology Slika 3. Primjerna modelna karta s prostornom raspodjelom najpogodnijih sustava pridobivanja drva nakon uvo|enja u primjenu forvardera s vitlom

nised systems. If the trade-offs are not too high, operational work with chain saw will be prevented.

Therefore the systems »Chain saw & Skidder« and »Chain saw & Forwarder« are not likely in the evaluation model. »Chain saw & Tower Yarder« in cutCroat. j. for. eng. 31(2010)2

to-length method will be proposed only in steep terrain and high extraction distances. Extraction operations with helicopter will only be suggested if there is no other system applicable. The model also gives the possibility to estimate the equipment and work-

83

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

Table 7 Equipment and workforce demand Tablica 7. Zahtjevi za stojevima i radnom snagom Category – Kategorija Employment (Work force) – Zaposlenost (radna snaga) Capacity demand: Chain saw Uporaba: motorna pila Capacity demand: Tracked Harvester Uporaba: gusjeni~ni harvester Capacity demand: Cable Forwarder Uporaba: forvarder s vitlom Capacity demand: Tower Yarder Uporaba: stupna {umska `i~ara Capacity demand: Tower Yarder & Processor Uporaba: stupna {umska `i~ara i procesor Capacity demand: Helicopter & Processor Uporaba: helikopter i procesor

h/year 2273 h 1155 h 148 h 136 h 43 h 334 h 2h

force demand, which accounts for 2273 h of manpower and 136 h of operating time for cable forwarders (Table 7).

3.2 Scenario c and d: Improving forest road network – Scenariji c i d: pobolj{anje mre`e {umskih cesta The implementation of a new harvesting technology might be a good opportunity to improve the conditions of harvesting operations, but in non-accessible regions this would have no effect. Investing in infrastructure, higher road density has positive effects on the productivity of harvesting operations because of reduced extraction distance. The following example shows the effects of improving forest road network for an 11 ha area in Tiefental region, which can only be harvested by tower yarders or helicopter. There have been no harvesting operations in recent years, so there is 2595 m³ available for harvesting within the next 20 years. The new forest road gives the possibility to improve currently used systems and to open the area for new harvesting tech-

nologies. Before improving the road network, the technological layers for the scenarios a, b and c are identical. Only tower yarders and helicopters can be used for extracting timber.

Before building the forest road, chain saw/tower yarder and chain saw/tower yarder/processor have been selected as best suitable harvesting systems by the model (Scenario a/b/c – Table 8). For these three scenarios the proposed systems are identical, but the impacts on the evaluation criteria may differ. After building the new road, the whole harvesting area is accessible for cable forwarder but also for tracked harvester technology and the utility analysis suggested using these two machine types for the whole area (Scenario d – Table 8).

After implementing the new forest road (Scenario c), the timber harvesting productivity could be increased by 50% in comparison to zero option (Scenario a). This increase is the result of the shorter average extraction distance, which could be reduced from 400 to 150 m. Further effects of higher productivity are lower fuel consumption, higher contribution margin, but also a lower employment rate.

There is no difference between scenario a and b because in this area cable forwarder technology can only be used after road construction. The effects after implementing both new forest road and cable forwarder technology (Scenario d) were much more impressive. Productivity could be tripled. Fuel consumption, damage to the remaining stand and the injury rate were also much lower. These could be explained in using more efficient equipment and in the safety standards and better control mechanism of highly mechanised systems. On the other hand, there are also some negative effects, e.g. much higher bearing pressure after driving in the stand and lower demand in work force as a result of improved technology (Table 9).

After harvesting the whole timber volume with Scenario d, CO2 emissions could be reduced by 5059 kg compared to scenario a/b. Contribution margin also increased by approximately EUR 40 000, but the employment rate decreased by 1330 hours. Although

Table 8 Harvesting volume before and after improving forest road network Tablica 8. Sje~ivi obujam prije i poslije pobolj{anja mre`e {umskih cesta System – Sustav Tracked Harvester & Cable Forwarder – Gusjeni~ni harvester i forvarder s vitlom Chain saw & Tower Yarder – Motorna pila i stupna {umska `i~ara Chain saw & Tower Yarder & Processor Motorna pila, stupna {umska `i~ara i procesor

84

Scenario a/b/c – Scenarij a/b/c 0 m³ 0% 834 m³ 32% 1761 m³

68%

2595 m³

100%

Scenario d – Scenarij d 2595 m³ 100% 0 m³ 0% 0% 2595 m³

100%

Croat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

M. Kühmaier and K. Stampfer

Table 9 Impacts before and after improving forest road network Tablica 9. Utjecaj prije i poslije pobolj{anja mre`e {umskih cesta Indicator – Pokazatelji Productivity – Proizvodnost Bearing Pressure – Nosivost tla Fuel Consumption – Potro{nja goriva Stand Damage – O{te}enost sastojine Contribution Margin – Kontribucijska mar`a Demand in Work Force – Potreba za radnom snagom Injury Rate – U~estalost ozljeda

Scenario a Scenario a 6 m³/h 15 kPa 5.93 kg CO2/m³ 36% 40 /m³ 0.62 h/m³ 84.05/mil. m³

Scenario c Scenario c 9 m³/h 15 kPa 5.10 kg CO2/m³ 36% 48 /m³ 0.51 h/m³ 84.05/mil. m³

Variation Varijabilnost +50% ±0% –14% ±0% +20% –18% ±0%

Scenario d Scenario d 19 m³/h 332 kPa 3.98 kg CO2/m³ 17% 56 /m³ 0.11 h/m³ 6.03/mil. m³

Variation Varijabilnost +217% +2113 % –33% –53% +40% –82% –93%

Table 10 Cost analysis of forest road network improvement Tablica 10. Analiza tro{ka pobolj{anja mre`e {umskih cesta

Contribution margin/year before road construction – Kontribucijska mar`a prije izgradnje {umske ceste Yearly payment – Godi{nji tro{ak Contribution margin/year after road construction – Kontribucijska mar`a godinu dana nakon izgradnje {umske ceste Difference against Scenario a/b – Razlika izme|u slu~aja a i b Recommendation – Preporuka

five of seven criteria could be improved and the evaluation process suggests the implementation of cable forwarder technology, the overall evaluation of building a forest road does not need to be positive at all.

Therefore also the effects of the construction phase should be implemented into the evaluation.

These results have further been validated by an economic evaluation. As part of the model, the contribution margin has been calculated for the area of interest by using average revenues, hourly system costs and productivity models (Kühmaier 2010). The construction of the forest road in the Tiefental region involves costs of approximately EUR 28 per running meter. To access the area, a new forest road with a total length of 700 m is required, resulting in an overall planned forest road cost of EUR 19 600. Considering a payment schedule period of 20 years and a yearly nominal interest rate of 3.5% an ordinary annuity of EUR 1379 has been estimated (Table 10).

After considering the yearly contribution margin before forest road construction and deducting the annual payments, a contribution margin between EUR 4849 and EUR 5887 has been calculated. The research shows that harvesting operations with the suggested systems are always positive within the inCroat. j. for. eng. 31(2010)2

Scenario a Scenarij a

Scenario c Scenarij c

Scenario d Scenarij d

5190

6228

7226

–

1379

1379

–

4849

5887

– –

–341 negative

697 positive

vestigation area. From an economic view, scenario d is the most favourable one, followed by a/b and c.

These results show that an improving forest road network could enhance timber harvesting operations, but the payments for the road construction might not be settled within the investigation period and this will have a negative recommendation for upgrading infrastructure and efficiency in forest operations. Therefore, it is important to go one step further and also to include possible technological improvements, which can only be implemented after upgrading infrastructure. In our example, positive effects of scenario d can be seen against all other scenarios as a result of including infrastructure and technology improvement. Scenario d achieves a yearly contribution margin of EUR 5887, which is more than EUR 1000 higher than building the forest road without technology improvement, and EUR 700 higher than the zero option.

4. Summary – Zaklju~ci The aim of this study was to develop a SDSS for identifying the best suitable harvesting systems and to estimate ecological, economical and social effects

85

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

on timber harvesting operations after improving forest road networks and/or implementing new harvesting technologies. The model has been implemented in GIS and demonstrated in a 1100 ha forest enterprise in steep terrain in the Southern part of Lower Austria. In evaluating specific scenarios, the implementation of cable forwarder technology had positive effects on productivity, CO2 emissions, stand damage, contribution margin, and injury rate.

On the other hand, the introduction of this technology had negative effects on the bearing capacity and employment rate. Generally speaking, the introduction of cable forwarders would have more positive than negative effects.

For another project area of 11 ha, the improvement of the forest road network also generated positive effects on productivity, fuel consumption and contribution margin, but the economic effect was too low to recommend the construction of the road.

Only a combination of increasing forest road density and implementing cable forwarder technology lead to a positive decision with tripling productivity, increasing contribution margin from 40 to 56 /m³ and lowering the damage rate by 53% and the injury rate by 93%. Admittedly, impacts on soil and the employment rate deteriorated. This example also shows the need of involving both infrastructure and technology improvement for planning harvesting operations.

As with any multiple-attribute preference model, this approach generates a cardinally scaled order of all decision alternatives with respect to their expected utility. However, the decision maker must be aware that the resulting solution may just be a best-compromise solution based on subjective rationality (Mollaghasemi and Pet-Edwards 1997). Sensitivity analysis is one of the powerful tools of decision support systems. However, the implementation of this decision model in GIS has to be modified by an experienced user. By using a model base management system (MBMS) and storing model components in an object-oriented data base, the flexibility could be improved. In this case, the capabilities for spatial analysis could be substantially improved by the integration of ModelBuilder™ into ESRI® ArcGIS.

5. References – Literatura Adams, J. D., Visser, R. J. M, Prisley, S. P., 2003: Modeling Steep Terrain Harvesting Risks using GIS. Austro2003: High Tech Forest Operations for Mountainous Terrain, October 5–9, Schlaegl – Austria, 10 p.

Berg, S., Lindholm, E., 2005: Energy use and environmental impacts of forest operations in Sweden. Journal of Cleaner Production, 13 (1): 33–42.

86

Eiwegger, A., 2009: Arbeitssicherheit bei der Seilrückung – Auswertung von Arbeitsunfällen (Working safety for cable yarding operations – Analysis of accidents). Diplomarbeit am Institut für Forsttechnik, Universität für Bodenkultur Wien, 57 p.

Goicoechea, A., Hansen, D. R., Duckstein, L., 1982: Multiobjective Decision Analysis with Engineering and Business Applications. Wiley, New York, 519 p.

Jänich, K., 2009: Unfallbericht 2008 (Injury report 2008).

Niedersächsisches Forstliches Bildungszentrum (Hrsg.), 28 p.

Kainulainen, T., Leskinen, P., Korhonen, P., Haara, A., Hujala, T., 2009. A statistical approach to assess interval scale preferences in discerete choice problems. Journal of the Operational Research Society, 60: 252–258.

Kangas, A., Kangas, J., 2005: Multiple criteria decision support in forest management – the approach, methods applied, and experiences gained. Forest Ecology and Management, 207(1–2): 133–143.

Kangas, A., Kangas, J., Kurttila, M., 2008: Decision support for forest management. Berlin: Springer, XII, 222 p.

Kilian, W., Müller F., Starlinger, F., 1994: Die forstlichen Wuchsgebiete Österreichs. Eine Naturraumgliederung nach waldökologischen Gesichtspunkten (Forestal growth zones of Austria. Natural regions according to forest ecological criteria), Bericht Nr. 82. Forstliche Bundesversuchsanstalt, Wien, p. 60.

Klvac, R., Skoupy, A., 2009: Characteristic fuel consumption and exhaust emissions in fully mechanized logging operations. Journal of Forest Research – Article in Press, 1–7.

Kühmaier, M., 2010: Multikriterielle Entscheidungsunterstützung in der Holzernte (Multiobjective decision support in timber harvesting). Dissertation am Institut für Forsttechnik. Universität für Bodenkultur Wien, 200 p.

Laukkanen, S., Palander, T., Kangas, J. Kangas, A., 2005: Evaluation of the multicriteria approval method for timber-harvesting group decision support. Silva Fennica, 39(2): 249–264.

Lexer, M.J., Vacik, H., Palmetzhofer, D., Oitzinger, G., 2005: A decision support tool to improve forestry extension services for small private landowners in southern Austria.

Computers and Electronics in Agriculture, 49(1): 81-102.

Limbeck-Lilienau, B., 2004: Residual stand damage caused by mechanised harvesting systems. In: Proceedings of the Austro2003 meeting: High Tech Forest Operations for Mountainous Terrain. CD ROM. Limbeck-Lilienau, Steinmüller and Stampfer (editors). October 5–9, 2003, Schlaegl – Austria. 11 p.

Löffler, H. D., 1984: Terrain classification for forestry: final report to the Joint FAO/ECE/ILO Committee on the project 14.3.1.2.4 / Hans D. Löffler. – Munich: Univ. of Munich.

Lubello, D., 2008: Sviluppo di un modello per la pianificazione integrate dei sistemi di utilizzazione (A rule-based Croat. j. for. eng. 31(2010)2

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

M. Kühmaier and K. Stampfer

SDSS for integrated forest harvesting planning). Tesi di dottorato – Universitá degli Studi di Padova, Dipartimento Territorio e Sistemi Agro-Forestali, 213 p.

Saaty, T. L., 1977: A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology, 15(3): 234–281.

Lüthy, D., 1998: Entwicklung eines »Spatial decision support« Systems (SDSS) für die Holzernteplanung in steilen Geländeverhältnissen (Development of a SDSS for timber harvesting in steep terrain). Zürich: Vdf, Hochschul-Verl. an der ETH. 260 p. ISBN 3-7281-2654-3.

Sheppard, S. R. J., Meitner, M., 2005: Using multi-criteria analysis and visualisation for sustainable forest management planning with stakeholder groups. Forest Ecology and Management, 207(1–2): 171–187.

Maclaurin, E. B., 2000: The soft soil performance of wheeled and tracked vehicles. J Defence Sci, 5(4).

Mallinger, A., 2002: Technologieinventur. Berechnung von Potentialen in der Holzernte durch Rasterdatenanalyse und visuelles Programmieren mit »MapModels« (Technological inventory. Estimation of potential harvesting areas by visual programming and using »MapModels«), Projektarbeit UNIGIS 2000 (MAS), Institut für Geographie und Angewandte Informatik an der Universität Salzburg, 73 p.

Stampfer, K., 2002: Optimierung von Holzerntesystemen im Gebirge (Optimisation of timber harvesting systems in mountainous regions), Habilitationsschrift eingereicht an der Universität für Bodenkultur Wien, Juni, 96 p.

Stampfer, K., 2009: System costs, fuel consumption, and demand in work force of harvesting equipment in Austria, Manuscript.

Stampfer, K., 2010: Forstliches Ingenieurwesen (Forest engineering), Vorlesungsunterlagen Studienjahr 2010/11, Universität für Bodenkultur Wien.

Manwaring, J. C., Conway, G. A, Garrett, L. C., 1998: Epidemiology and Prevention of Helicopter External Load Accidents. Journal of Safety Research, 29(2): 107–121.

Sterba, H., 1983: Sortentafeln für Fichte (Assortment tables for Norway Spruce). Österreichischer Agrarverlag, Wien, 151 p.

Mendoza, G. A., Sprouse, W., 1989: Forest planning and decision making under fuzzy environments: an overview and illustrations. Forest Science, 35(2): 481–502.

Suvinen, A., 2006: A GIS-based simulation model for terrain tractability. Journal of Terramechanics, 43 (4): 427–449.

Meyer, T., Stückelberger, J., Hollenstein, K., Attenberger, M., Hänggli, T., 2001: effor2: Holzerntekonzept für die Untersuchungsbetriebe im Kanton Schwyz (Timber harvesting concept for several areas in canton Schwyz). Eidgenössiche Technische Hochschule Zürich. Interne Berichte Nr. 11. 98 p.

Mollaghasemi, M., Pet-Edwards, J., 1997: Technical Briefing: Making Multiple-Objective Decisions. IEEE Computer Society Press, Los Alamitos.

Næsset, E., 1997: A spatial decision support system for long-term forest management plan by means of linear programming and a geographical information system. Scand.

J. For. Res., 12(1) 77–88.

Nordfjell, T., Athanassiadis, D., Talbot, B., 2003: Fuel Consumption in Forwarders. International Journal of Forest Engineering, 14(2) Rowland, D., 1972: Tracked vehicle ground pressure and ist effect on soft ground performance. Proceedings of the fourth international conference of the ISTVS, Stockholm.

Vacik, H., Lexer, M. J., 2001: Application of a spatial decision support system in managing the protection forests of Vienna for sustained yield of water resources. For. Ecol.

Manage, 143(1–3): 65–76.

Von Winterfeldt, D., Edwards, W., 1986: Decision analysis and behavioral research. Cambridge University Press, Cambridge.

Wolfslehner, B., Vacik, H., 2008: Evaluating sustainable forest management strategies with the Analytic Network Process in a Pressure-State-Response framework. Journal of Environmental Management, 88 (1): 1–10.

Wratschko, B., 2006: Einsatzmöglichkeiten von Seilforwardern (Potential use of cable forwarders). Diplomarbeit am Institut für Forsttechnik, Universität für Bodenkultur Wien, 66 p.

Yoshioka, T., Sakai, H., 2005: Amount and availability of forest biomass as an energy resource in a mountainous region in Japan: a GIS-based analysis. Croatian Journal of Forest Engineering 26 (2): 59–68.

Sa`etak

Razvoj vi{eatributnoga prostornoga sustava za pomo} pri odlu~ivanju kod odabira sustava pridobivanja drva Cilj je ove studije bio razvoj prostornoga sustava koji bi pomogao pri odlu~ivanju (SDSS) u odabiru najpogodnijega na~ina pridobivanja drva uz procjenu ekolo{kih, ekonomskih i dru{tvenih utjecaja na radove pri pridobivanju drva nakon pobolj{anja {umske prometne infrastrukture i/ili nakon primjene novih postupaka sje~e, izradbe i

Croat. j. for. eng. 31(2010)2

87

M. Kühmaier and K. Stampfer

Development of a Multi-Attribute Spatial Decision Support System ... (75–88)

transporta drva. Spomenuti je model razvijan u ~etiri ciklusa. Nakon definiranja podru~ja primijenjena je tehnolo{ka prosudba sustava pridobivanja drva s obzirom na lokalne ~imbenike. Prikladni su sustavi zatim uspore|eni primjenom vi{eatributne teorije korisnosti (MAUT) zbog odabiranja najpogodnijega sustava pridobivanja drva. Sedam nezavisnih uvjeta i pokazatelja kori{teno je pri odabiru: utjecaj na tlo (nosivost tla, kPa), potencijal globalnoga zagrijavanja (potro{nja goriva u kg ekvivalent uglji~noga dioksida, kg CO2), o{te}enost sastojine (%), kontribucijska mar`a ( /m3), vrijeme premje{tanja (zbroj tehnolo{kih pokazatelja, %), zaposlenost (potreba za radnom snagom, h/m3) i sigurnost pri radu (broj nezgoda/m3). Za prevo|enje apsolutnih u usporedive vrijednosti kori{tene su lokalne sklonosne funkcije. Te`inski faktori za obilje`ja uspostavljeni su uz pomo} upravitelja lokalnih {umarskih poduze}a. Analizom scenarija procijenjene su posljedice razli~itih zamjenskih postupaka pridobivanja drva.

Model je ugra|en u geografski informacijski sustav (GIS) i predo~en na primjeru {umarskoga poduze}a koje gospodari s 1100 hektara {uma na nagnutim terenima u Donjoj Austriji. Pri ocjeni odre|enih scenarija kori{tenje forvardera s vitlom imalo je pozitivan utjecaj na proizvodnost (+14 %), emisiju uglji~noga dioksida (–2 %), o{te}ivanje sastojine (–7 %), kontribucijske mar`e (+15 %) i na ozljede radnika (–36 %). Op}enito se mo`e re}i da primjena forvardera s vitlom ima vi{e pozitivnih nego negativnih posljedica. Primijenjeni model sugerira izvo`enje 56 % drva spomenutom tehnologijom i smanjenje primjene stupnih {umskih `i~ara s 90 na 38 %.

Za jednu drugu povr{inu veli~ine 11 hektara pobolj{anje {umske prometne infrastrukture dalo je pozitivan utjecaj na proizvodnost (+50 %), potro{nju goriva (–14 %) i kontribucijsku mar`u (+20 %), ali je ekonomski utjecaj bio prenizak za davanje preporuke za izgradnju {umske ceste. Samo je kombinacija pove}anja gusto}e {umskih cesta i primjena forvardera s vitlom dovela do pozitivnoga pomaka utrostru~ivanjem proizvodnosti, pove}anjem grani~noga prihoda s 40 na 56 %, smanjenjem razine o{te}enosti sastojine za 53 % i smanjenjem razine ozljeda za 93 %. Istodobno su se pogor{ali utjecaj na podlogu (od 15 do 332 kPa) i zaposlenost (–82 %).

Izno{enje drva stupnim {umskim `i~arama u potpunosti je zamijenjeno izvo`enjem drva forvarderom s vitlom. Taj primjer pokazuje potrebu uklju~ivanja kako infrastrukture, tako i tehnologije za operativno planiranje pridobivanja drva.

Kao i svi vi{eatributni modeli korisnosti, ovaj pristup rezultira kardinalnim nizom svih ina~ica zamjenskih rje{enja s obzirom na njihovu o~ekivanu korist. Kakogod, donositelj odluka mora biti svjestan da krajnje rje{enje mo`e biti samo kompromisno rje{enje zasnovano na subjektivnoj racionalnosti. Analiza je osjetljivosti sna`niji alat sustava za potporu pri odlu~ivanju. Ugra|ivanje ovoga modela odlu~ivanja u GIS mora nadgledati iskusni korisnik. Kori{tenjem modela sustava upravljanja i pohranom sastavnica modela u objektno orijentiranu bazu podataka mo`e se pove}ati fleksibilnost modela. U tom slu~aju mogu}nosti prostorne analize uvelike se mogu pobolj{ati primjenom alata ModelBuilder™ u ra~unalnom programu ESRI® ArcGIS. Provodi se daljnje usavr{avanje opisanih postupaka, a model }e se testirati na drugim istra`ivanim podru~jima, posebice na ravnim terenima, o ~em zasad ne postoji dovoljno spoznaja. U budu}nosti bi se mogla razmotriti i uporaba podataka LIDAR-a visoke razlu~ivosti. Za pobolj{anje rezultata koristit }e se jo{ ve}i broj kriterija zadovoljavaju}e kakvo}e, a za postizanje bolje prilago|enosti korisniku usavr{it }e se korisni~ko su~elje i kontrola me|udjelovanja korisnika i modela. Provjerit }e se prakti~na primjenjivost ovoga modela pri planiranju {umskih cesta i dobavi energetskoga drva.

Klju~ne rije~i: pridobivanje drva, mre`a {umskih cesta, sustav za pomo} pri odlu~ivanju, procjena korisnosti

Authors’ address – Adresa autorâ:

Received (Primljeno): October 16, 2010 Accepted (Prihva}eno): November 17, 2010

88

Martin Kühmaier, MSc. e-mail: martin.kuehmaier@boku.ac.at Assoc. Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Applied Life Sciences Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Straße 82/3 A-1190 Wien AUSTRIA Croat. j. for. eng. 31(2010)2

Original scientific paper – Izvorni znanstveni rad

Learning Curves of Harvester Operators Frank Thomas Purfürst Abstract – Nacrtak Single grip harvesters may reach very high productivity levels but are very expensive forestry machines. Therefore, they have to work very efficiently. This paper will analyse the learning curves of harvester operators and the influence of human factors described in a qualitative and quantitative manner. Based on long term production logging files (StanForD), the performance of 32 operators was collected over a period of three years (3,351 stands, 0.65 mil. m³). 16 of the operators were beginners and their learning curve was analysed and drawn as a sigmoid function. Most operators begin their career between 50% and 60% of the mean performance and double their performance (200%) by the end of the learning phase. The differences and variation in the learning curve between individual operators are large and generalisations are to be made with care. A learning curve phase productivity loss of 24% was calculated for the average 8 month duration, which equates to approximately 45000 Euro.

Keywords: harvester operator, learning curves, harvester productivity, harvester operations, human influence.

1. Introduction – Uvod The modern forest industry uses thousands of single grip harvesters. They may reach very high productivity levels but are very expensive forestry machines. Therefore, they have to work very efficient. Numerous different factors have an influence on this productivity, many of which have already been determined and described in literature (Purfürst 2009). One factor is often disregarded when considering mechanical work performance; the human factor. As Kirk et al. (1997) mentioned: »A skilled operator is essential if the investment in the machinery is to be maximized by the contractor«.

Purfürst (2009) analysed 53 harvester productivity models in regards to the operator effect. Most of these models do not use the operator as an influencing factor, only mentioning the experience of the operator. However, the intra-individual (Purfürst 2010) and temporally observed variability in the operator performance level can be very high and will increase with the experience of the operator. The replacement of an experienced operator by a new harvester operator alone can be estimated by a production loss of about 49,650 (Gellerstedt et al. 2005). The training costs are still not taken into account, however, and can account for up to 15,000 (Gellerstedt et al.

Croat. j. for. eng. 31(2010)2

2005). Therefore it is necessary to briefly present a rough outline of the learning curve concept.

Driving a single-grip harvester is a complex and an exacting type of work. Performance varies over time. Within a day there are variations of the whole mental system based on tiredness and daily rhythms.

Throughout the seasons there are also variations based on light, weight of roundwood and driving conditions. But there still are variations without an external influence factor. When a harvester operator starts his career he usually has a very low performance. Over time he makes less mistakes, learns to ignore unimportant information and his coordination skills increase. Through repeating work cycles very often he will develop an »automatic« mode of working where functions are coordinated by the cerebellum instead of by the cerebrum. With work experience he gets well trained and his and the system – performance increases over time. This effect is based on the aptitude of the operator.

The relation between productivity and experience is called a learning curve. A learning curve describes the level of performance through learning over time. It can be calculated by the quotient of learning results and the time needed. To measure the experience of a harvester operator directly still remains a problem. Currently, only productivity, time

89

F. T. Purfürst

and some influencing factors can be measured. The assumption is that the more time the operator spends the more familiar he/she will become with the machinery, hence his/her skill level increases. However, this does not presume that the productivity of the whole harvesting system increases proportionally. At the same time, other influencing factors such as age and health of the operator or age and condition of the machinery can change the productivity level. The »real« learning curve is often not quantifiable. For this reason the learning curve is often interpreted as the relation between productivity and time or cycles of work.

The learning curve describes the performance level reached by an individual operator. However, the operator and the harvester with its configurations form a unit. Only this unit has reached the measured performance. Therefore the only possibility is to set the focus on the whole harvester system, as separating the influence of the operator and the harvester is not possible.

The difference between less trained and well–experienced operators can be very high. Some analyses of learning curves in the forestry sector have already been described for forest workers (Garland 1990), forwarders (Harstella 2004), wood transportation (Björheden 2000), harwarders (von Bodelschwingh 2003), cable-yarder teams (Dodd and McNeel 1996) and cable yarder operators (Stampfer 1999). Stampfer et al. 2002 found in the case of helicopter yarding (K-Max) that the difference in performance between an experienced and an inexperienced operator was nearly 2:1. Nonetheless, analyses of learning curves of harvester operators are rare. Parker et al. (1996) studied a forwarder operator and a harvester operator. The greatest increases of performance are within the first 30 days but the level of performance fluctuates greatly. Heinimann (2001) analysed a harvester driver and found that with small diameter trees the performance increases by 50% within one year.

The learning curve effect can be divided into different segments. Stampfer (1999) suggested two phases: the first phase is called the learning phase where the operator continuously increases his performance.

The second one is the working phase where the operator is working at a relatively constant performance level. Therefore, Jacke 2000 suggested a division into three phases. The first model is equivalent to the description of the so called »inexperienced« and »experienced« operator that is used by most authors (Purfürst 2009, Purfürst and Erler 2006). Current learning curve analyses are often restricted to the first phase e.g. the training in the simulator. The second phase is often disregarded and consistent performance level of the operator is assumed.

90

Learning Curves of Harvester Operators (89–97)

The length of the learning curve is discussed differently. The reaching of the »experienced« level is often described as a number of productive machine hours and it differs between 1000 and 1500 Productive Machine Hours (PMH) (Gabriel 2005, Jacke and Wagner 2002, Wagner 2004). Other authors describe it as a time span between 8 and 12 months (Calabrese 2000), or in hundreds of PMH (Loschek et al. 1998) or as a number of harvested trees (Heinimann 1998).

The duration of the learning curve can be reduced through a structured training program (Kirk et al. 1997). Even the sleeping time of the harvester operator has an influence on the learning time and therefore on the learning curve (Dinges and Kribbs 1991).

The learning curve effect is not restricted to new operators. Von Bodelschwingh and Pausch (2003) describe the training effect of an »experienced« driver on a machine other than his own and discover that he reaches a working plateau after 20 days. Hoellerl (2005) found that for a harvester driver with the change of control (control stick) about four weeks acclimatization is needed to reach his previous level.

Concerning the shapes of the learning curve, there is general consensus in the literature that it first raises and then flattens. Jacke (2000) describes that a learning curve can be represented through a number of exponential – functions, but also suggested that it is hard to describe complex movement patterns.

Björheden (2000) instead uses a simple e-function for the analysis of mechanized skidding. Dodd and MCNeel (1996) divide the compensation function into two linearly rising lines, with the transition between the two at 100 working days. Calabrese (2000) describes the learning effect as a multistage process.

Currently there is a lack of knowledge to explain the learning curve and the impact of the learning process. This study examines duration, shape, differences and costs of learning curves of harvester operators and the influence of human factors described in a qualitative and quantitative manner.

2. Material and methods – Materijal i metode 2.1 Environmental conditions and operators – Okoli{ni uvjeti i voza~i The experimental study sites were located in Germany with the focus on East Germany and Bavaria, south Germany. For comparability, only pine–dominated stands were selected for the analysis. The harvesting system included thinning of marked trees in young stands with a CTL single-grip harvester. For most stands this was the first thinning and extracCroat. j. for. eng. 31(2010)2

Learning Curves of Harvester Operators (89–97)

tion corridors were built. Distances between working corridors were 20 to 24 meters. Only sites with a slope less than 10% were considered. Over 90% of the harvested trees were pines (Pinus sylvestris L.).

Other tree species included Spruce (Picea abies (L.) H.Karst. – 6.1%), Larch (Larix decidua Mill. – 0.6%), Birch (Betula pendula Roth – 1.7%) and hardwood (0.7%).

Most data were collected through the data logging systems of the harvester and evaluated by additional information from time studies on sites. In total the information of 52 operators was recorded but only 32 were considered for the analysis due to the following rules: Þ Machinery (only three types of harvesters for small diameter thinning were selected: John Deere 1070, Valmet 901 and Ponsse Beaver), Þ Tree stands (spruce, larch and hardwood dominated stands were not considered), Þ Harvesting system (only thinning was considered rather than clear-cut or wind throw-areas), Þ Operating sites (number of the working areas had to be more than 15 per operator), Þ Other factors such as: e. g. incomplete information about the operator.

There are differences in the operators’ educational and practical background. Some of the operators had taken a harvester education program or course at a forestry academy. These courses varied from one day to seven weeks. More than half of the operators (56%) had three years of training as a forest worker. Some (28%) had completed a different trade such as mechanics, carpentry or butchery and learnt harvesting on the job-site. The hours spent operating the harvester varied as well from relative beginners to well experienced operators that had been working in this field for seven years and longer.

2.2 Logging documents – Proizvodni dokumenti The study was based on logging documents from the harvester computer. With automatic harvester data logging the StanForD-Standard was used and the production information was stored in defined files (»prd«, »pri«, »drf«, »stm«) depending on the system (Skogforsk 2007). Additional information on times, dates, harvesting data, operators and software was also included. For example, the types of time are Effective time, G15-time, Move-time, Runtime, Work time and Repair time. The G15-time, which is defined as hours of effective machine time including downtime not exceeding 15 minutes per occasion, was used for all of the analyses.

Croat. j. for. eng. 31(2010)2

F. T. Purfürst

2.3 Data analysis – Analiza podataka A program written by the author analyses the huge number of StanForD files, by parsing different variables. A big problem is the partly inconsequent realization of the StanForD-Standard. The variances in collecting software types and versions installed in harvesters made it very difficult to analyse the data automatically. Nonetheless, the parsed StanForD data was written into a database. As the size and duration of the operation per stand differs, the information had to be harmonized and weighted, based on the time variable. These data were analyzed with standard statistical programs independent from the real stand-size. The analysis of the production data is based on the stand, weighted by time (days).

The information about stems, times and harvested volume were used to create performance information for every stand and operator for a specific date. To compare the operator it is necessary to find a reference performance. The choice of the type reference is difficult and based on the data and the use (Purfürst 2009). In this study a relative mean was used as a representative for the whole population and was calculated with one logarithm regression.

Only the influence factor of the tree volume was considered. Afterwards every value was divided by this regression value to calculate the relative performance. The value of 1 is equivalent to the mean performance.

P0 P0 Pr = = 0.684 *in ( tvol ) + 3 .543 (1) Pm e Where: Pr Relative productivity, m³/h Po Actually observed productivity, m³/PMH Pm Model productivity, m³/PMH, tvol Volume/tree, solid cubic meter The level of performance in this study is defined as the arithmetic mean of the relative performance of 60 days. After several attempts it was established that the period of 60 days is a sensible compromise.

It is long enough to reproduce several harvest stands and short enough to include other performance–influencing factors such as the learning curve.

The learning curve indicates a significant increase in performance as a function of time, which under a certain level of tolerance does not fall again. The challenge is to determine if it is solely due to fluctuations in performance by non-temporal (external) factors or to an actual increase associated with the ability of the harvester operator. The process of learning ends, but usually not with the achievement of the average power (Performance Level = 1), but stops beyond. It is therefore important to determine the end

91

F. T. Purfürst

Learning Curves of Harvester Operators (89–97)

Figure 1 Example of a learning curve, 60-days performance level and sigmoid model (Operator No. 22) Slika 1. Primjer krivulje usavr{avanja, {ezdesetodnevna razina u~inkovitosti i sigmoidni model (voza~ br. 22)

of the learning phase. Because of the strong differentiation of the data it is not possible in a purely mathematical way. Therefore, the end of the learning phase will be determined from a combination of visual and mathematical properties as the first (large) maximum is determined by the end of the increase in benefit levels over time. Fig. 1 shows an example of driver No. 22. It is clear that by the end of summer 2004, the subject has already reached a local maximum, but after that the learning increases again.

Thus, this is only a scattering and not the real end of the learning phase. The next maximum, which was reached at the beginning of February 2005, was characterized by a subsequent decrease in benefit levels.

To describe the learning curve, a sigmoid model was used and solved with a nonlinear regression: PL max − a PL(t) = (2) +a 1 + b * e – c *t And PLmin =

PL max + b * a 1+b

(3)

Where: PL(t) Performance level over time PLmax Maximum performance level a, b, c Variable t Working days Start-parameter for regression solution: PLmax = 1.4, a = 1.0, b = 1.0, c = 0.001

92

3. Results – Rezultati 4.5 million stems from 3351 stands and 32 operators were analyzed, which represents approximately 0.65 million cubic meter (m³) of harvested wood.

The mean volume is 0.147 m³/tree. The arithmetic mean production is 9.8 m³/PMH (geometric mean: 8.93 m³/PMH). Nearly 70% of the variation in productivity can be explained with the tree-volume ratio. Data from the logging documents were recorded during December 2003 and September 2006.

The logging documents of 16 of 32 operators reveal a learning curve. There is a large variation in statistical values. Table 1 shows the facts of the learning curves of these harvester operators.

13 of the 16 drivers listed in Table 1 show at their maximum a performance level (PLmax) above the average (PL = 1). The number of days before overtaking the average levels of performance is highly differentiated with a range between 68 and 286 days (mean: 193 days). The duration of time before reaching the end of the learning phase differs too. The range is between 155 and 488 days (arithm. mean: 255 days, median 227 days). It can be generally assumed that the duration of the learning period is approximately 8 months, but the standard deviation is high.

At the beginning of the study these operators started at a performance level between 0.33 and 0.76, wherein the average is 0.56. Half of operators’ interquartile range begin their career between 51% and Croat. j. for. eng. 31(2010)2

Learning Curves of Harvester Operators (89–97)

F. T. Purfürst

Table 1 Facts of the learning-phase for individual operators Tablica 1. Podaci o usavr{avanju pojedinih voza~a Operator – Voza~

2 5 8 10 11 12 13 14 15 16 17 18 20 22 24 28 Arithmetic mean Median 25%-quantile 75%-quantile PLStart PL(1) PLEnd Increase PL PLMax

PLStart

PL(1)

0.39 0.33 0.51 0.76 0.61 0.59 0.50 0.52 0.62 0.60 0.69 0.63 0.51 0.51 0.67 0.58 0.56 0.59 0.51 0.62

Days – Dana 281 – 75 232 68 385 286 – 88 155 105 274 263 168 135 – 193 168 105 274

PLEnd

1.14 0.68 1.43 1.29 1.11 1.05 0.86 0.86 1.22 1.11 1.28 0.98 1.34 1.40 1.05 0.76 1.10 1.11 0.95 1.28

Days – Dana 281 237 192 355 140 488 232 127 130 212 221 460 350 279 155 183 255 227 176 331

Increase PL – Pove}anje PL Overall – Sve per day – po danu % % 292 0.23 206 0.15 280 0.48 170 0.15 182 0.36 178 0.09 172 0.16 165 0.27 197 0.46 185 0.24 186 0.27 156 0.08 263 0.24 275 0.32 157 0.25 131 0.10 200 0.24 183 0.24 169 0.15 220 0.28

PLMax

1.14 0.85 1.43 1.29 1.13 1.05 1.03 0.99 1.47 1.40 1.38 1.10 1.34 1.40 1.05 0.76 1.17 1.12 1.05 1.39

– Performance Level at the beginning of the learning phase – Razina u~inkovitosti na po~etku usavr{avanja – Reaching the Performance Level of 1 (mean overall) – Dosezanje razine u~inkovitosti 1 (prosjek) – Performance Level at the end of the learning phase – Razina u~inkovitosti na kraju usavr{avanja – Increasing of the Performance Level – Pove}anje razine u~inkovitosti – maximum reached Performance Level – Najve}a razina u~inkovitosti

61% of the mean performance and most operators double their performance (200%) by the end of the learning phase.

The performance level achieved by the end of the learning phase varies between 0.68 and 1.43. With large differences, the mean value is 110% (median: 111%) and the inter-quartile ranges between 0.95 (25% percentile) and 1.28 (75% percentile).

There are no significant correlations between the duration of the learning phase, final level of performance (p=0.571) and maximum level of performance (p=0.940). It is therefore not possible to prove statistical significant correlations between the types of learning curves based on the data used in this investigation. The increase in performance throughout the learning phase of the regression line is on averCroat. j. for. eng. 31(2010)2

age with 0.24 percentage points per day, and the values turn out very differently. The increase per day correlated significantly with the achieved levels of performance at the end of the learning phase (p=0.002) and with the maximum level of performance achieved (p=0.002) which is on average 117%.

To describe the shape of the learning curves a sigmoid model was used. After tests, it is assumed that this model describes the learning curves with sufficient accuracy. This is confirmed by the residuals, which have approximately normal distribution and homoscedasticity. However, learning curves can be very individual. Fig. 2 shows four typical types of adjustment of the sigmoid model. For example: the function of operator No. 20 has a clear sigmoid shape

93

F. T. PurfĂźrst

Learning Curves of Harvester Operators (89â&#x20AC;&#x201C;97)

Figure 2 Different types of learning curves of four operators (sigmoid model) Slika 2. Razli~ite vrste krivulja usavr{avanja za ~etiri voza~a (sigmoidni model)

with a convex (from down) followed by a concave gradient. Other only have a convex gradient, for example operator No. 16.

The average cost, time and effort for the learning phase of a new operator can be calculated. On the basis of the logging documents, a mean production loss of about 330 PMH can be expected in the use of a new harvester operator. This corresponds to a re-

94

duced productivity of approximately 24% in the first 8 months. This means nearly two month of productivity and costs per harvester of ca. 45,000 . Possibly increased wear or repair costs and training course fees are not included. Significant correlation between the performance level of the operator and the nonproductive times such as repair time could not be recognized.

Croat. j. for. eng. 31(2010)2

Learning Curves of Harvester Operators (89–97)

4. Discussion – Rasprava After analyzing the logging data and documents, it has been calculated that on average an operator reaches the end of the learning phase in the plateau phase of work after nine months. The inter-quartile range is about 6–11 months. This is slightly lower than the previously described figure of 8–12 months (Calabrese 2000) and confirms the previously assumed, but rarely empirically investigated figure of 1000–1500 PMH (Gabriel 2005, Jacke and Wagner 2002, Wagner 2004). For 16 operators a learning curve effect could be demonstrated, in which their performance roughly doubled. These results are higher than the 50% performance increase over a year described by Heinimann (2001). He does not, however, make any statements about the existing experience of the operator at the start of data acquisition. Moreover, it seems appropriate to describe the end of the learning curve with the time required.

In this study learning curves were represented with the model of the sigmoid function, which proved suitable for describing the increase in productivity of a harvester operator. Therefore, the four parameters to be determined correspond well to the actual course of the performance level. The suggestion that a sigmoid function cannot adequately portray a multi-stage learning process (Calabrese 2000) could not be confirmed. The results also showed that a simple e-function (Björheden 2000) or a composite linear function (Dodd and McNeel 1996) could not reflect the actual course in an adequate way.

A significant correlation between the performance level and the repair time could not be detected in this study. This is in conflict with the results of Kirk et al.

(1997). The explanation could come from the way the data were analysed. This study is focused on the operator rather than on different stages of learning curves as described by Kirk et al. (1997).

This study was focused on dense, first thinning stands where differences in the operators’ skill levels are especially emphasized (Kärhä et al. 2004). However, an effective driver can operate efficiently in all phases of the work cycle (Ranta 2004). In this analysis intra-daily fluctuations and the impact of shift work were not considered. These two factors may result in fatigue and can have an important influence on individual performance (Nicholls et al. 2004). By using stem-based logging data these intra-daily fluctuations can be considered. Other factors such as motivation and physical condition can affect changes in performance over time. These factors, however, are not detailed in this study. The influence of the weather and seasons could be observed in this study but are not yet verified.

Croat. j. for. eng. 31(2010)2

F. T. Purfürst

Using long term logging documents (historical data) will always result in a lack of information about the stand and the actual conditions and events that happen onsite. Additionally, the difficulties with the inconsequent realization of the StanForD-standard and deficits in calibration of harvester measurements can affect the variation of the data. However, an advantage of using historical logging documents is that there is no influence of the observer as in time studies. Therefore, data may misrepresent the overall performance to some degree but the tendency can be generalized.

In the future more tasks performed by the harvester operator will be automated, (Note: either delete or explain both). The performance of the whole harvester system will increase (Löfgren 2004). This automation involves a risk of creating a boring job and reducing the operator’s alertness (Gellerstedt 2002). The boredom may decrease the performance over a long period more than the automation increases the performance. In contrast, it is also possible that additional tasks are requested of the operator and the workload does not really decrease. All of this can have an influence on the operator’s learning curve. A lot of further research is suggested for this field.

5. Conclusions – Zaklju~ci This study indicates that the operator has a decisive influence on the harvester performance. For a large number of operators the learning effect could be demonstrated – performance was roughly doubled within the mean period of 8 months. The differences and variation between the individual operators are large and the training phase can be quite expensive. However, once you have good experienced operators – keep them. They are your most valuable assets.

Acknowledgements – Zahvale This research was supported by the German Research Foundation DFG ER 131/5.

6. Literature – Literatura Björheden, R., 2000: Learning Curves in Tree Section hauling in Central Sweden. International Journal of Forest Engineering 12(1): 9–18.

Von Bodelschwingh, E., 2003: The new VALMET 801 combi first operational test results under central European conditions. In: Austro2003: High Tech Forest Operations for Mountainous Terrain, Austria, CDROM.

95

F. T. Purfürst

Von Bodelschwingh, E., Pausch, R., 2003: Der Valmet 801 Combi – erster Praxis Einsatz in Deutschland (VALMET 801 combi – first operational test results), AFZ-Der Wald 17: 858–860.

Calabrese, D., 2000: Canadian Switches to Mechanized, Cut-to-Length, TimberLine <http://www.timberlinemag. com/articledatabase/view.asp?articleID=222> (Accessed: 11 April 2010).

Dinges, D., Kribbs, N., 1991: Sleep, Sleepiness, and Performance. In: Performing while sleepy: Effects of experimentally – inducet sleepiness., Wiley, Chichester; New York, Chapter 4, 97–128.

Dodd, K., McNeel, J., 1996: Factors Affecting Productivity of newly Established Thinning Operations. International Journal of Forest Engineering 7(2): 43–50.

Gabriel, O., 2005: Fahrer für Stendal (Harvester operator for Stendal). Forst & Technik 17(3): 26–27.

Garland, J., 1990: Assessing gains from woodsworker training, International Journal of Industrial Ergonomics 5: 227–242.

Gellerstedt, S., 2002: Operation of the Single-Grip Harvester: Motor-Sensory and Cognitive Work. International Journal of Forest Engeneering 13(2): 35–47.

Gellerstedt, S., Liden, E., Bohlin, F., 2005: Gesundheit und Leistung bei mechanisierter Waldarbeit. (Health and Performance in Mechanized Forest Operations), Swedish University of Agricultural Sciences, 45 p.

Harstela, P., 2004: The competence of the forest-machine operator and tacit knowledge. In: Simulator-based training of forest machine operators, Joensuu, Finland.

Heinimann, H. R., 1998: Produktivität und Einsatzbedingungen verschiedener Harvestertypen – eine statistische Auswertung aufgrund von Leistungsnachweisen (Productivity and operating conditions of various harvester types – a statistical analysis based on logging documents), Technical report, Department Wald und Holzforschung, Professur forstliches Ingenieurwesen, ETH Zürich.

Heinimann, R., 2001: Lernkurveneffekt eines Harvesterfahrers (Lerning curve effect of harvester operators). Unpublished Technical report, Department Wald und Holzforschung, Professur forstliches Ingenieurwesen, ETH Zürich.

Höllerl, H., 2005: Auf die Größe kommt es an – oder doch nicht? (The size is important – or not?). Forst & Technik 17(3): 16.

Jacke, H., 2000: Lernerfolg in der komplexen Steuerung von Forstmaschinen (Learning in the complex control of forestry machines). Forst & Technik 12(5): 4–7.

Learning Curves of Harvester Operators (89–97)

Kärhä, K., Rönkkö, E., Bumse, S., 2004: Productivity and Cutting Costs of Thinning Harvesters. International Journal of Forest Engineering 15(2): 39–48.

Kirk, P., Byers, J., Parker, R., Sullman, M., 1997: Mechanisation Developments within the New Zealand Forest Industry: The Human Factors. International Journal of Forest Engineering 8(1): 75–80.

Löfgren, B., 2004: Simulator as a research tool. In: Simulator-based training of forest machine operators, Joensuu, Finland.

Loschek, J., Jirikowski, W., Pröll, W., Sperer, S., Tresniowski, S., 1998: Holz in der Durchforstung (Wood in thinning), Kooperationsabkommen Forst-Platte-Papier, Wien, 120 p.

Nicholls, A., Bren, L., Humphreys, N., 2004: Harvester Productivity and Operator Fatigue: Work Extended Hours.

International Journal of Forest Engineering 15(2): 57–65.

Parker, R., Kirk, P., Sullman, M., 1996: Learning curves of mechanised harvester and forwarder operators. LIRO Report 21(29): 1–6.

Purfürst, F. T., 2009: Der Einfluss des Menschen auf die Leistung von Harvestersystemen (Human Influence on harvester productivity). PHD-Thesis, Technische Universität Dresden, Dr. Hut Verlag, Munich, 320 p.

Purfürst, F. T., Erler, J., 2006: The precision of productivity models for the harvester – do we forget the human factor.

In: Precision Forestry in Plantations, Semi–Natural and Natural Forests, Proceedings of the International Precision Forestry Symposium, Stellenbosch University, South Africa, 465–475.

Purfürst, F. T., Erler, J., 2010: Human Influence on Productivity of Harvester Operations. Manuscript.

Ranta, P., 2004: The competence of the forest-machine operator and tacit knowledge. In: Simulator-based training of forest machine operators, Joensuu, Finland.

Skogforsk, 2008: Standard for Forest Data and communications – Main document. Technical report, Skogforsk, Sweden.

Stampfer, K., 1999: Lernkurveneffekte bei Forstmaschinenfühern (Learning curves of forest machinery operators).

Arbeit im Wald – Östereichische Forstzeitung 110(12): 1–2.

Stampfer, K., Gridling, H., Visser, R., 2002: Analyses of Parameters Affecting Helicopter Timber Extraction. International Journal of Forest Engeneering 13(2): 61–68.

Wagner, T., 2004: From the eighties up to the future – the development of forest machine simulators – A story of success? In: Simulator-based training of forest machine operators, Joensuu, Finland.

Jacke, H., Wagner, T., 2002: Einsichten aus einem virtuellen Wettbewerb – Teil 2 (Results from a virtual competition – part 2). Forst & Technik 14(2): 4–6.

96

Croat. j. for. eng. 31(2010)2

Learning Curves of Harvester Operators (89–97)

F. T. Purfürst

Sa`etak

Krivulje usavr{avanja voza~a harvestera Jednozahvatni harvesteri mogu dose}i visoku razinu proizvodnosti, ali su istodobno skupi {umski strojevi.

Brojni razli~iti ~imbenici utje~u na njihovu proizvodnost te su oni dosad ve} prepoznati i opisani u literaturi. Jedan je ~imbenik ~esto zapostavljen kada je rije~ o strojnom radu u {umarstvu, a to je ~ovjek. Razlika izme|u slabije obu~enih i iskusnijih radnika mo`e biti zna~ajna.

Poveznica izme|u proizvodnosti i iskustva naziva se krivulja u~enja. Ona opisuje razinu radnoga u~inka u usavr{avanju (u~enju) tijekom vremena. Mo`e se izra~unati kao kvocijent rezultata usavr{avanja i vremena potrebnoga za obavljanje radnih zadataka. Neposredno mjerenje i pra}enje razine iskustva za voza~e harvestera jo{ je uvijek problem. U dana{nje vrijeme samo se proizvodnost, vrijeme trajanja rada te neki utjecajni ~imbenici mogu neposredno izmjeriti, {to je problem pri spoznaji krivulje u~enja i utjecaja postupka usavr{avanja. Ovo istra`ivanje prou~ava trajanje, oblike, razli~itosti i tro{kove krivulja u~enja kod voza~a harvestera, te utjecaj ljudskih ~imbenika opisanih kakvo}nim ili koli~inskim varijablama.

Lokacije za provedbu ovoga istra`ivanja bile su u Njema~koj, a odabrane su proredne borove sastojine. Sustav pridobivanja drva uklju~ivao je sje~u dozna~enih stabala i sortimentnu metodu izradbe drva jednozahvatnim harvesterom (CTL). Na osnovi dugoro~nih proizvodnih datoteka iz informacijskoga sustava vozila (StanForD) prikupljeni su podaci o u~inku za 32 voza~a u vremenskom razmaku od 3 godine (obuhva}ena je 3351 sastojina i 0,65 mil. m³ izra|enoga drvnoga obujma). [esnaestero voza~a bili su po~etnici i njihove krivulje u~enja bile su ra{~lanjene i oblikovane kao sigmoidne funkcije. Prisutne su bile i razli~itosti u predznanju izme|u pojedinih voza~a.

Nekolicina voza~a zavr{ila je programe ili te~ajeve osposobljavanja za rad harvesterom na {umarskim u~ili{tima.

Ve}ina voza~a na po~etku svoje karijere ostvaruje izme|u 50 i 60 % prosje~noga radnoga u~inka koji se udvostru~uje (200 %) na zavr{etku procesa usavr{avanja. Za opisivanje krivulje usavr{avanja primijenjena je sigmoidna krivulja. Na osnovi testova ustanovljeno je da spomenuti model opisuje krivulje u~enja zadovoljavaju}om precizno{}u. Razlike i varijabilnosti krivulja u~enja izme|u pojedinih voza~a velike su i poop}avanje mora biti pa`ljivo provedeno. Ipak, mo`e se pretpostaviti da vrijeme usavr{avanja (u~enja) iznosi 8 mjeseci, ali uz veliku standardnu devijaciju. Na osnovi dokumentiranih izvje{taja o radu strojeva mo`e se o~ekivati manja proizvodnost kod voza~a po~etnika u iznosu od prosje~no 330 proizvodnih sati rada stroja (PMH). To se poklapa sa smanjenjem proizvodnosti od 24 % u prvih 8 mjeseci rada i razumijeva gotovo dvomjese~nu prosje~nu proizvodnost i tro{kove u visini od 45 000 .

U vremenu trajanja procesa u~enja zavr{na razina radne u~inkovitosti (p = 0,571) i najve}a razina radne u~inkovitosti (p = 0,940) ne pokazuju zna~ajnu povezanost. Zbog toga je nemogu}e dokazati statisti~ki zna~ajnu povezanost izme|u tipova krivulja u~enja u ovom istra`ivanju. Prema regresijskoj krivulji faze usavr{avanja pove}anje je u~inkovitosti u prosjeku 0,24 % po radnom danu, no i te su vrijednosti podlo`ne promjenama Ovo istra`ivanje pokazuje da voza~ ima presudan utjecaj na razinu u~inkovitosti rada harvestera. Za ve}inu je voza~a dokazan utjecaj usavr{avanja (u~enja). Razli~itosti i varijacije izme|u pojedinih voza~a velike su i usavr{avanje mo`e biti iznimno skupo. Kakogod, u slu~aju da postoji kvalitetan iskusan voza~ – valja ga zadr`ati, jer su takvi voza~i najvredniji ~imbenik radnoga procesa.

Author’s address – Autorova adresa:

Received (Primljeno): June 17, 2010 Accepted (Prihva}eno): October 19, 2010 Croat. j. for. eng. 31(2010)2

Frank Thomas Purfürst, PhD. e-mail: thomas.purfuerst@forst.tu-dresden.de Institute of Forest Utilization and Forest Engineering Technische Universität Dresden Dresdener Str. 24 01737 Tharandt GERMANY

97

Original scientific paper – Izvorni znanstveni rad

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory Esmael Ghajar, Akbar Najafi, Sattar Ezzati Abstract – Nacrtak Efficient allocation of resources is an essential principle in forest management. An important case in resource allocation is when the available resources are not sufficient to service all requests. One of the key elements in forest management is to minimize the total costs of the unallocated requests. With respect to high capital cost of forest operation machinery, it is necessary to reduce expenses of one cubic meter of wood extraction by appropriate Skidding Machines Allocation (SMA). Fuzzy set theory as a soft methodology and practical decision support system was used to handle uncertain variables and vague range of logs volume and physiographic conditions to develop models.

The aim of this research is to present a decision support method to determine the economical activity zone of forest operation machines so that this allocation would result in the highest net profit for forest managers. To achieve this goal, all skid routes in the study area were divided into work units with 75 m width and 200 m length whereupon 379 units were formed collectively. Within each unit, the related mathematical productivity models were applied to estimate one cycle time and cost of machines. The effective factors of these models included Skidding Distance (SD), Volume of Logs per Cycle (VLC), and Number of Logs per Cycle (NLC). Three separate fuzzy inference models were developed to predict the skidding cost of each machine in the units, and then proper machines were allocated. 70% of data was used as training and the rest was feed to the models for validation and test process through the generation of fuzzy models. Membership functions and fuzzy rule bases were created with the help of scientific knowledge, experts’ viewpoints and existing machine productivity models. The results showed that the application of the presented approach helps forest managers to recognize the desirable conditions for skidding machines to reduce the total costs of skidding. In addition, SMA fuzzy rule-based models reflect how to integrate expert knowledge with engineering system design. To present an illustrative example, the models were applied to allocate three commonly used Skidders, i.e. Timberjack 450c, HSM 904, and Zetor, in a mountainous forest, whose inventory data were known and harvesting was planned for the next period. The results showed that the Zetor was the most economical option in »Very short« and »Short« (< 300 m) distances at all levels of NLC and VLC, while HSM 904 was the most cost effective machine at »Medium«, »Long« and »Very long« distances (300 m to 900 m) at all levels of NLC and VLC with the exception of »Long« distance units (around 700 m), with »Low« NLC (2 pieces) and »Low« VLC (around 4 m3), as well as the units with »Very long« distance and »Low« and »Medium« volume (less than 5.5 m3), where Timberjack 450C was the most adequate machine from the economic point of view.

The result of this application also showed a spatial variability in skidding costs by different machines based on skid route conditions. The implemented method can be very helpful in where and how to use skidders to gain maximum profit from forest operations.

Keywords: Log extraction, Time study, Resource allocation, Fuzzy rule-based modeling, Skidding cost, Forest harvesting

Croat. j. for. eng. 31(2010)2

99

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99â&#x20AC;&#x201C;110)

Fig. 1 Diagram of steps of the SMA models including approximate reasoning to assess machines location suitability based on input variables (SD: Skidding distance, VLC: Volume of load per cycle, and NLC: Number of logs per cycle).

Slika 1. Planiranje rada skidera pomo}u neizrazite teorije (ulazni podaci: srednja udaljenost privla~enja, obujam tovara/tura, broj komada obloga drva u tovaru) 100

Croat. j. for. eng. 31(2010)2

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

1. Introduction – Uvod Forest planning and management are a complicated proposition in many mountainous regions where there are varieties of physiographic factors. A basic requisite for effective management of felling site in a wooded area is the knowledge of the suitability of the machine to be used in various operations. (Curro and Verani 1990). Machine assignment is an important task during the tactical phase of forest management planning, in which managers can accommodate field activities to obtain maximum efficiency. They have to know where and which skidders must be applied to remove logs from stump to road with the lowest cost. To achieve this aim, initially, forest managers have to know machine productivity model and cost to identify economic zones of each machine. Machinery allocation could be done based on these information and company limitations.

In real world, skidder allocation is usually done based on experiences and budget, while the main aim of this approach is log extraction in a given time, however there is no guarantee for the lowest skidding cost. In traditional logic, inputs are assumed to be unambiguously defined attributes and all inferences are based on clear threshold of class inputs, while most concepts in forest operations cannot be defined precisely or have clearly characterized boundaries in space and time. The analyses of such data entail a technology that applies both intuition (expert knowledge) and engineering heuristic to enhance model-based system designs. To combine knowledge, techniques and methodologies from various sources, an intelligent system that could consider the vague data for inference is needed. In fuzzy logic proposed by Zadeh (1965) by using the elements of everyday language for representing the desired system behavior, the need for rigorous mathematical modeling is circumvented. In fact the theory of fuzzy sets provides a more realistic mathematical representation of the perception of truth than traditional, two-valued logic and Boolean algebra (Hasan et al.

2008). Most former studies were focused on assessment of cost, production and environmental effects of machinery (Adebayo et al. 2000, Ledoux and Huyler 2001) while skidding machine allocation had received scant attention from researchers.

The main objective of this paper was to design three fuzzy models to estimate economical activity zone of three types of commonly used skidders.

These models could serve as decision support systems during the tactical planning phase of forest management planning. To provide an example of model output an analysis was conducted on approximately 7323 ha of mountainous forest of northern Croat. j. for. eng. 31(2010)2

E. Ghajar et al.

Iran. The final result of these models is demonstrated as a map in which economical zones of each skidder are shown in detail.

2. Methods –Metode The following steps were made to build the Skidding Machine Allocation (SMA) models: Þ Identifying work units and determining cost effective input variables, Þ Planning a fuzzy system including linguistic values, membership functions, fuzzy rule bases and method of defuzzification for each machine, Þ Evaluating models performance by NeuroFuzzy training and validating, Þ Computing crisp output skidding cost (SC) for each of the three machines in each unit.

2.1 Identifying work units – Utvr|ivanje radnih jedinica To determine work units assigned to a machine, a buffer zone of 75 m width and 200 m length was delineated on both sides of skid routes. These buffer zones were assumed as work units of skidders (Fig. 2). The work units map was created using ArcGIS ver. 9.3.

2.2 Cost effective input factors – ^imbenici koji utje~u na tro{ak strojnoga rada The factors affecting positively skidding costs were calculated or estimated in each unit. To establish factors that affect time and consequently the cost of wood skidding from each unit to the landing (road), mathematical models of machine productiv-

Fig. 2 Forming work units Slika 2. Radne jedinice 101

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

Table 1 Linguistic values used in the SMA models Tablica 1. Pretvorba vrijednosti u jezi~ne varijable neizrazitoga modela za planiranje rada skidera Variables – Varijable

Linguistic values – Jezi~ne varijable

Skidding distance, m Srednja udaljenost privla~enja, m

Very short Vrlo mala

Short Mala

Medium Srednja

Long Velika

Very long Vrlo velika

0 – 200

200 – 400

400 – 600

600 – 800

>800

Number of logs per cycle Broj komada obloga drva u tovaru, kom.

Volume of load per cycle, m Obujam tovara, m3

3

Low – Mali

Medium – Srednji

High – Veliki

2

2.5

3

Low – Mali

Medium – Srednji

High – Veliki

3.8

5.8

6.9

ity were used. These models were developed based on time study technique, and then the effective factors in productivity of machinery were identified accordingly. Multiple linear regression models were used to predict skidding cycle times. The number of logs per cycle and volume of load in each unit were estimated using inventory data.

The following mathematical models of productivity were calculated in previous time studies carried out in mountainous forests.

F(t) = 4.142 + 1.988N + 0.01769D + 1.093V

(1)

Where: F(t) time of one cycle hauling, min D distance of loading, m N number of logs per cycle V volume of load, m3 L winching length, m (25 m) Skidding distance, number of logs per cycle and volume of load had a significant influence on the time of wood extraction with all machines. The reality fuzzy inference model was based on these effective factors.

Timberjack 450c (Jourgholami 2008) F(t) = 1.3789 + 0.0537D + 0.039N + 0.032L + 0.39V (2) Zetor (Najafi 2005) F(t) = 1.873 + 0.02494D + 2.499N

(3)

HSM 904 (Najafi 2005)

2.3 Defining linguistic variables and values – Utvr|ivanje jezi~nih varijabli i vrijednosti In fuzzy logic, prototypical linguistic rules of a fuzzy model are formulated by use of linguistic variables instead of quantitative variables. For instance if the »skidding distance« (SD) is a linguistic variable, its corresponding term set would be T (SD) =

Table 2 Part of rule bases applied for building the overall models of Timberjack 450C & Zetor Tablica 2. Dio modela neizrazitoga skupa za skidere tipa Timberjack 450C i Zetor

1 2 3

24 25

102

Distance Srednja udaljenost privla~enja Very low Vrlo mala Very low Vrlo mala Very low Vrlo mala

IF – Ako Number Broj komada obloga drva u tovaru Low Mali Low Mali Low Mali

Volume Obujam tovara Low Mali Medium Srednji High Veliki

Long Velika Very long Vrlo velika

High Veliki High Veliki

High Veliki High Veliki

DoS Stupanj podr{ke pravila

THEN – Tada Skidding Cost of Zetor Tro{ak strojnoga rada za skider Zetor

Skidding Cost of Timberjack Tro{ak strojnoga rada za skider Timberjack 450C

1.00

Very low – Vrlo nizak

Low – Nizak

1.00

Very low – Vrlo nizak

Low – Nizak

1.00

Very low – Vrlo nizak

Low – Nizak

High Visok High Visok

High Visok High Visok

1.00 1.00

Croat. j. for. eng. 31(2010)2

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

E. Ghajar et al.

{very short, short, medium, long, very long}. Each term in T (SD) is delineated by a fuzzy set in the universe of discourse, here, e.g. U = [0. 1950]. We considered »very low« fuzzy set as the skidding distance less than 200 m, »low« as the skidding distance between 200 to 400 m, »medium« as the skidding distance between 400 to 600, »long« as the skidding distance between 600 to 800 and »very long«, the last set, as the skidding distance longer than 800 m. The linguistic values of other variables specified by the fuzzy sets are shown in Table 1.

2.4 Description of membership functions – Stupnjevi pripadnosti Each linguistic value is described by a membership function (Fig. 3). The triangularly shaped membership function was adapted for the models. The class boundaries were fuzzy and determined based on the maximum and minimum value of each variable, scientific knowledge and technical experts’ views and experiences. Defining membership functions caused soft thresholds of variable classes compatible with most concepts in forest management, in contrast to classical mathematical models which define hard thresholds. The combination of membership functions used to generate the rule bases in each model.

2.5 Determining the fuzzy rules – Utvr|ivanje pravila neizrazite teorije Fuzzy IF-THEN linguistic rules which reflect the knowledge of experts and the previous research results have the general form: ëIF x1 is A1 and x2 is A2 and xm is Am then y is Bû

Fig. 3 Membership functions of input linguistic variables Slika 3. Stupanj pripadnosti jezi~nih varijabli za srednju udaljenost privla~enja drva, obujam tovara i broj komada obloga drva u tovaru Croat. j. for. eng. 31(2010)2

Where: x1,…,xm are linguistic input variables with linguistic values A1,…,Am, and linguistic output variables with linguistic value B, respectively.

The production rules included all combinations of input variables as antecedent part and skidding cost as consequent part that represented different degree of suitability of each machine. With Timberjack 450C and Zetor models, there were three linguistic variables e.g. skidding distance, number of logs and volume of load per cycle that had 5, 3 and 3 linguistic values (Equation 1 and 2), respectively. Fuzzy rules included the combination of linguistic values of skidding distance, number of logs and volume of load per cycle. Then, 45 rules were produced, and 15 rules were automatically omitted because there were no »2.5 logs« per cycles. The linguistic value »Medium« of the variable »Number of logs per cycle«

103

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

skidding cost. Assuming that the threshold value is the turning point when hardening the rule, we choose the membership functions so that they take a value of 0.5 at the threshold used by the corresponding crisp rule (Muthu et al. 2008). So fuzzification of any production rule yields a membership function in output variable. In this case we had »High, High, and High« for antecedent part and »High« for inference part of this rule (Table 2).

Such calculation was conducted for all rules of each of the three models.

2.6 Defuzzification – Pretvaranje neizrazitoga kona~noga zaklju~ka

Fig. 4 Membership functions of output skidding cost and methodology of converting crisp values to fuzzy values Slika 4. Pretvorba »~vrstih« vrijednosti tro{ka strojnoga rada u neizrazite vrijednosti was omitted and all of its combinations were omitted accordingly. Also there was no unit with »High« number of logs per cycle and low volume, so 5 rules were omitted. Consequently, the rule base of Timberjack 450c and Zetor models included 25 rules. In HSM 904 model according to its time prediction model (Equation 3), two factors out of three factors (e.g. Skidding distance and Number of logs per cycle) were identified as the effective factors, so its rule base included 10 rules.

Fuzzy rules could be derived from both experts reasoning and linguistic expressions and from the relationships between the system variables (Borri et al. 1998). In this study fuzzy rule bases in the models were generated using time study and hourly productivity cost of skidding machines. The time of one cycle hauling and consequently the skidding cost of each skidder for all selected work units were calculated. As a result we had three skidding costs (SC) against three machines per unit. As a detailed example of an IF – THEN rule, the skidding time and cost of one cycle by Timberjack 450C from a unit with skidding distance of 700 m (long), the number of logs 3 (high) and volume of load 7.63 (high) were calculated as 30.83 minutes and 24.78 (based on prices in 2009). These calculations were based on the center of fuzzy sets.

According to the defined output membership functions and considering their threshold values (Fig. 4), this calculated value corresponded to a »High«

104

After generating complete sets of rules for each model, fuzzy variables were combined with respect to their respective membership functions to provide an estimate of the cost of one cycle hauling from a unit on skid routes to the landing. In fact the combination of fuzzy variables presents how their membership functions should be aggregated. Two subjects are usually discussed in aggregation of fuzzy variables. The first is relative importance of fuzzy variables. All criteria and indicators must be assigned with their corresponding weights to reflect their significance (Mendoza and Prabhu 2000), and since we used a mathematical productivity model of machines, in these models relative weights of variables were assigned. The second is the way of combining membership functions.

Two common combination procedures of fuzzy sets, named the »minimum« and the »maximum« operators were first suggested by Zadeh (1965) used to aggregate fuzzy variables.

The minimum operator corresponds to the intersection of fuzzy sets and the maximum operator represents the union of these sets. If the fuzzy sets A and B in universe U are as the following ordered pairs: A = {mA(x),x}, ∀x ∈ U

(4)

B = {mB(x),x}, ∀x ∈ U

(5)

Then the membership function mC(x), as the intersection C = A ∩ B defined by: mC(x) = ∀x : mA ∩ B = min{mA(x),mB(x)}

(6)

And the membership function mD(x), as the union D = A ∪ B defined by: mD(x) = ∀x : mA ∪ B = max{mA(x),mB(x)}

(7)

The developed set of rules should be combined with these conjunction operators. Standard max. (union) »or« and min. (intersection) »and« operators are preferred (Babuska 1998). So to determine apCroat. j. for. eng. 31(2010)2

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

E. Ghajar et al.

Table 3 Part of sample data used in neurofuzzy training Tablica 3. Dio uzorka podataka u teoriji neizrazitoga skupa Unit number Broj radne jedinice

SD, m Srednja udaljenost privla~enja, m

NLC Broj komada obloga drva u tovaru, kom.

VLC, m3 Obujam tovara, m3

1

100

2

5.59

16

11.73

8.53

2

100

3

7.63

20.21

15.41

8.9

3

300

2

4.88

18.76

16.71

19.15

4

500

2

5.44

22.91

21.7

29.99

5

500

3

7.53

27.18

25.38

30.36

6

700

2

5.44

26.45

26.69

40.73

7

900

3

5.97

32.55

35.36

51.59

proximate reasoning, the standard min/max operators were used. Finally to compute crisp output of the skidding cost (SC) (defuzzification) fuzzy – Mamdani reasoning (Mamdani 1981) and »Center Of Gravity« (COG) method were used.

SC of HSM 904, $ SC of Timberjack 450C, $ Tro{ak strojnoga rada za Tro{ak strojnoga rada za skider HSM skider Timberjack 450C

SC of Zetor, $ Tro{ak strojnoga rada za skider Zetor

2.7 Evaluating the model performance – Procjena modela The system performance could be evaluated using a large set of input–output data, and parameters

Fig. 5 Location map of study area Slika 5. Podru~je istra`ivanja Croat. j. for. eng. 31(2010)2

105

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

of the system could be fine tuned in order to achieve a low »generalization error« (Azadi et al. 2009). In our case, 379 work units were indentified, and then units data were collected to enter the models as crisp inputs. In such a data–rich situation, 70% of data was used as training data to fit the models, and then 15% of data was assigned as validation data to estimate the prediction error for model selection. Finally 15% of data was applied to assess the generalization error of the final model chosen. Training data, chosen among all designed units, included all combinations of input variables in the study area. It was used in neurofuzzy training of machine fuzzy models so that input data included collected variables estimated in the units and related cost assigned as output variables in the training stage (Table 3).

3. Application – Primjena The intrinsic motivation toward the problem studied in this research was to find the economical location of skidders in the existing skid routes, to decrease harvesting costs and increase the machine

performance in the forest of Mazandaran Wood & Paper Industries (MWPI).

MWPI is the largest wood and paper company in Iran which manages about 180,000 ha of hyrcanian commercial forests in the north of Iran.

In order to conduct this research, four districts (7323 ha) under MWPI management forest were selected (Fig 5).

4. Results and discussion (output of models) – Rezultati i rasprava After assessment of model performance using training data, each of the three models was applied for all 379 work units in the study area. The output variable of skidding cost (SC) resulted from the combination of the fuzzy values of input factors (SD, VLC, and NLC) according to the If–Then rules. The results were obtained as defuzzified discreet values.

Three crisp cost values of one cycle hauling against three skidders were obtained in each work unit when fuzzy model of skidding cost was applied

Fig. 6 Allocation map of extraction machines in study area Slika 6. Razmje{tanje vozila na istra`ivanom podru~ju prema neizrazitoj teoriji 106

Croat. j. for. eng. 31(2010)2

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

E. Ghajar et al.

Table 4 Results in some units after running the models for whole data Tablica 4. Rezultati u nekim radnim jedinicama nakon provo|enja teorije neizrazitoga skupa Unit number Broj radne jedinice 165 189 227 256 270 330 379

NLC SD, m Srednja udaljenost Broj komada obloga drva u tovaru, kom. privla~enja, m 700 3 100 3 700 2 900 2 300 3 500 3 900 2

VLC, m3 SC of Zetor, $ SC of Timberjack 450C, $ SC of HSM 904, $ Obujam tovara, Tro{ak strojnoga rada za Tro{ak strojnoga rada Tro{ak strojnoga rada za skider Zetor za skider HSM skider Timberjack 450C m3 7.84 32.404 32.17 40.58 7.74 24.431 14 8.55 8 30.627 25.99 39.91 4.48 29.544 32.17 45 5.79 22.467 21.36 17.63 7.32 27.129 25.74 28.9 5.44 30.14 32.17 45

Selected machine Odabrano vozilo HSM 904 Zetor HSM 904 Timberjack 450c Zetor HSM 904 Timberjack 450c

* All values are in USD, adjusted to price level for Year 2009.

for each machine (Table 4). As a result the skidder corresponding to the lowest skidding cost was allocated to each unit.

To create a visual result of extraction machines allocation model, mapping process of defuzzified crisp outputs was conducted using ArcGIS ver. 9.3.

Then the final skidding machines allocation map was produced (Fig. 6).

A general comparison of total skidding cost for different allocation strategies was shown in Figure 7.

The results showed that the application of a proper combination of machines based on skidding route properties resulted in lower total skidding costs than when using just one machine for the entire project

Fig. 7 A general comparison of Total cost of skidding between SMA model and investigated machines in the study area Slika 7. Usporedba ukupnih tro{kova strojnoga rada prema vrsti skidera pomo}u neizrazite teorije Croat. j. for. eng. 31(2010)2

area. SMA model application decreased the total skidding costs by about 8% compared to the case when only Zetor was used for wood extraction. This improvement was 13% and 22%, respectively, for Timberjack and HSM 904. So, it could be proposed to forest managers to apply machines in their respective economic zones through skidding routes if several skidding machines are available.

Frequency distribution of extraction machines depending on the skid route distance in the study area was illustrated in Figure 8. The HSM 904 was the most economical machine in 44.19% of work area, while Timberjack 450C and Zetor were allocated in 33.67% and 22.13% of work area, respectively.

The models showed that in all units located at »Very short« distance, the lowest skidding costs were observed by Zetor. This fact could be explained by the fact that Zetor was a steel tracked skidder with wide arch. Expanded wide arch allows the Zetor to increase skidding performance maneuver to set chokers and winching elements of skidding cycle. So this machine is the first priority of managers at all »Very short« distance skid route units to meet minimum costs during skidding activities.

The differences in frequent distribution among the three machines were tested at different levels of skid route distance (Table 5).

The results showed that Zetor was still the most economical at »Short« skidding distances (down to 300 m) too, its costs were lower than others in 71% of units located in short distance skid route. Timberjack 450C was allocated to the units with the VLC fewer than 5 m3 in this distance class owing to high coefficient of VLC in the productivity model (Equation 1).

As a result Timberjack 450C is an appropriate machine for thinning or lightening operations in which young stands are harvested. Equation 3 showed that performance of HSM 904 was not dependant on the

107

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

Table 5 Comparisons of machines in skidding distance classes Tablica 5. Usporedba statisti~kih parametara ovisno o udaljenosti privla~enja SD Sred. udalj. priv.

Parameters Statisti~ki parametri Chi-Square – Hi kvadrat Df – Stupanj slobode p-Value – p vrijednost

Short – Mala

Medium – Srednja

Long – Velika

Very Long – Vrlo velika

53.846 2 0.000

64.800 1 0.000

1.756 1 0.185

2.380 1 0.123

VLC. In contrast, high coefficient of NLC represents a high sensitivity to the number of pieces per cycle.

For this reason the cost of HSM 904 was the least when the NLC was 2 and VLC was more than 5 m3. It could be advised to apply this skidder in mature stands with »Low« NLC and »Medium« or »High« VLC. The results emphasized that despite the fact that Zetor was a steel tracked skidder and could not move fast, it still was the first selection at »Short« distance because of its ability to set chokers and winching elements of skidding cycle and lower hourly production cost. Preference of »Zetor« in »Short« and »Very short« distances was expected because of its low speed and low hourly production cost in comparison with two other machines.

The HSM 904 became the economical machine in the »Medium« distance (average distance of 500 m) and it was used significantly more frequently than Zetor and Timberjack 450C. Timberjack 450C was less costly than HSM 904 in the »Medium« distance only when VLC was less than 6 m3 and NLC was 3.

In the units with long distance (around 700 m),

Timberjack 450C and HSM 904 were assigned to 57 % and 43 % of units, respectively. Regarding final results of 379 units, it was observed that HSM 904 was the most cost effective machine in the »Long« distance when NLC was 2 and VLC was more than 5.5 m3, as well as when NLC was 3 and VLC was more than 7 m3. Under other conditions of this distance Timberjack 450c was the most cost effective and consequently it was the selected machine.

HSM 904 was used more frequently than Timberjack 450C at the »Very long« distance (around 900 m). It was assigned to 59% of units while Timberjack 450C was allocated in 41% of units where their VLC was less than 5.5 m3.

The differences between machines in frequency at »Long« and »Very long« distances were not significant based on chi-squared test shown in Table 5.

As a general result for three recent classes of distance, HSM 904 was the economical machine at »Medium«, »Long« and »Very Long« distances in skid routes with the exception of »Long« distance units

Fig. 8 Frequency distribution of extraction machines in the study area Slika 8. Udio rada pojedinoga vozila u radnim jedinicama prema srednjoj udaljenosti privla~enja drva 108

Croat. j. for. eng. 31(2010)2

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

with »Low« NLC and »Low« and »Medium« VLC, and also »Very long« distance units with VLC less than »Medium« where Timberjack 450C was less costly than HSM 904 and consequently it was the selected machine.

5. Conclusions – Zaklju~ci The implementation of fuzzy logic seems to be very promising in domains such as forest engineering, environmental modeling and predictive models where imprecise data are dealt with. This paper presents a modeling approach to allocate extraction machines for forest operations through a fuzzy technique using multiple linear regression models of productivity.

The advantages of the SMA models are: Þ The models present a linguistic description of each factor and its corresponding values instead of precise quantitative criteria. This kind of variables description is favored by policy makers of forest management. Þ Application of fuzzy Mamdani–based models to allocate resources makes the result of research consistent and easy to be understood by managers. Þ Flexibility of presented models permits any reform, should the number of variables or their membership functions be changed as a consequence of new scientific researches.

In the current research the input variables were chosen based on mathematical productivity models of skidding machines. In case of a large number of input variables, optimization of fuzzy rule bases using artificial intelligence could help to overcome problems of input variable selection and obtain reliable results. Our results were obtained under conditions including no constraints in skidder allocation models. Linear programming techniques can be used as exact modeling methods to obtain more accurate and realistic results when there are some cost or time constraints in using machinery. More applied research is required in the future to persuade forest managers to apply fuzzy models in practice.

Croat. j. for. eng. 31(2010)2

E. Ghajar et al.

6. References – Literatura Adebayo, A. B., Han, H. S., Johnson, L., 2007: . Forest Products Society. Forest Products Journal 57(6): 59–69.

Azadi, H., Van den Berg, J., Shahvali, M., Hosseininia, G., 2009: Sustainable rangeland management using fuzzy logic: A case study in Southwest Iran. Agriculture Ecosystem Environment. In press Babuska R., 1998: Fuzzy modeling for control. International Series in Intelligent technologies. Kluwer Academic Publishers, 260 p.

Borri, D., Concilio, G., Conte, E., 1998: A fuzzy approach for modeling knowledge in environmental systems evaluation. Computers, Environment and Urban Systems 22(3): 299–313.

Curro, P., Verani, S., 1990: On the maximum skidding output of the Timberjack 380 forest tractor. International Journal of Forest Engineering. 1(2): 35–39.

Hasan, M., 2008: Model for predicting rainfall by fuzzy set theory using USDA scan data. Agricultural water management 95, p. 1350 – 1360. Mining, Inference and Prediction.

Springer, New York.

Jourgholami, M., 2008: Productivity and cost of wheeled skidder in Hyrcanian forest. International Journal of Natural and Engineering Sciences 2(3): 99–103.

Ledoux, B., Huyler, K., 2000: Cost Comparisons for three Harvesting Systems Operating in Northern Hardwood Stands.USDA, Forest Service.Research, NE – 715.

Mamdani, E. H., Gaines, B. R., 1981: Fuzzy Reasoning and its Applications. Academic Press, London.

Mendoza, G. A., Prabhu, R., 2000: Multiple criteria decision making approaches to assessing Forest sustainability using criteria and indicators: a case study. Forest Ecology and Management 131(1-3): 107–126.

Muthu, K., Petrou, M., Tarantino, C. Blonda, P., 2008.

Landslide Possibility Mapping Using Fuzzy Approaches.

IEEE Transactions on Geoscience and Remote Sensing 46(4): 1253–1265 Najafi, A., 2005: Optimal engineering and economical programming in forest operation. PhD thesis. Tehran University, 113 p.

Zadeh, L., 1965: Fuzzy sets. Information and Control 8: 338–322.

109

E. Ghajar et al.

Skidding Machines Allocation (SMA) Using Fuzzy Set Theory (99–110)

Sa`etak

Planiranje rada skidera (SMA) pomo}u neizrazite teorije U~inkovito planiranje rada jedno je od glavnih zadataka u gospodarenju {umama, posebice kada postoje}a sredstva za rad nisu jednako dostupna u svim radnim jedinicama. Smanjenje ukupnih tro{kova rada svakako je bitna sastavnica gospodarenja {umama posebice u radno nedostupnim ili te`e dostupnim podru~jima. S obzirom na visoke tro{kove rada specijaliziranih {umskih vozila, u ovom slu~aju skidera, potrebno je smanjiti jedini~ne tro{kove pridobivanja drva odgovaraju}im planiranjem rada skidera te njihovim smje{tanjem u prostoru (Skidding Machines Allocation – SMA). Neizrazita teorija (fuzzy set theory) i sustav za potporu odlu~ivanja primijenjeni su za obradu nestalnih varijabli te raspona podataka kao {to su obujam obloga drva i terenski uvjeti radili{ta, tj. udaljenost privla~enja drva. Svrha je ovoga istra`ivanja prikazati sustav za potporu pri odlu~ivanju kojim bi se utvrdilo ekonomski najisplativije podru~je rada pojedinih vrsta skidera, {to }e u kona~nici smanjiti ukupne tro{kove te pove}ati dobit u pridobivanju drva. Da bi se postigao taj cilj, istra`ivano je podru~je podijeljeno u radne jedinice od 75 m {irine i duljine od 200 m do 900 m. Unutar svake radne jedinice matemati~ki su izra~unati modeli proizvodnosti te je procijenjeno vrijeme radnoga ciklusa i tro{ak strojnoga rada vozila koji su zajedno sa srednjom udaljenosti privla~enja drva (SD), obujmom tovara (VLC) te brojem komada obloga drva u tovaru (NLC) smatrani kao ulazne varijable modela SMA. Razvijena su tri odvojena neizrazita sustava za predvi|anje jedini~nih tro{kova privla~enja drva po radnim jedinicama koji su omogu}ili ekonomski najisplativije smje{tanje vozila u prostoru. U neizrazitom sustavu ulazni su podaci pretvoreni u jezi~ne varijable i tada se odre|uje stupanj pripadnosti varijable prema postavljenim pravilima. Primijenjena je Madamijeva metoda najmanjih i najve}ih vrijednosti koja za svako pravilo odre|uje stupanj pripadnosti varijable pojedinomu zaklju~ku, {to u kona~nici dovodi do stvaranja neizrazitoga skupa podataka odre|enoga stupnjem pripadnosti pojedine varijable. Modeli su primijenjeni na tri naj~e{}e kori{tena skidera u iranskom {umarstvu: Timberjack 450C, HSM 904 i Zetor, u planinskim {umama provincije Mazandaran.

Istra`ivanja su pokazala da je skider proizvo|a~a Zetor najisplativija ina~ica na vrlo maloj i maloj srednjoj udaljenosti privla~enja (< 300 m) neovisno o vrijednostima obujma tovara i broju komada obloga drva u tovaru, dok je skider HSM 904 najmanje tro{kove rada postigao na srednjoj, velikoj i vrlo velikoj srednjoj udaljenosti privla~enja (od 300 m do 900 m), neovisno o vrijednostima obujma tovara i broja komada obloga drva u tovaru uz iznimku pri velikoj srednjoj udaljenost privla~enja (oko 700 m), s malim brojem komada obloga drva u tovaru (2 komada) i malim obujmom tovara (oko 4 m3). Timberjack 450C je najisplativije vozilo na vrlo velikoj srednjoj udaljenosti privla~enja (oko 900 m) pri malom i srednjem obujmu tovara (manje od 5,5 m3). Istra`ivanje je pokazalo da se kori{tenjem modela neizrazitih skupova pri planiranju rada skidera mogu smanjiti tro{kovi rada vozila, a time i jedini~ni tro{kovi pridobivanja drva. Primjenom modela smanjuju se tro{kovi rada za 8 % u odnosu na slu~aj da se na istra`ivanom podru~ju koristi samo vozilo proizvo|a~a Zetor.

Klju~ne rije~i: privla~enje drva, planiranje rada, metoda neizrazitih skupova, tro{kovi privla~enja drva, sje~a drva

Authors’ address – Adresa autorâ:

Received (Primljeno): April 14, 2010 Accepted (Prihva}eno): October 11, 2010

110

Esmael Ghajar, BSc. e-mail: ismael.ghajar@modares.ac.ir Asst. Prof. Akbar Najafi, PhD. e-mail: a.najafi@modares.ac.ir Sattar Ezzati, BSc. e-mail: s_ztt@yahoo.com Tarbiat Modares University Faculty of Natural Resources Department of Forestry Imam Ave, Noor, Mazandaran Province P.O. Box 64414–356 IRAN Croat. j. for. eng. 31(2010)2

Original scientific paper – Izvorni znanstveni rad

Industrial Round-Wood Losses Associated with Harvesting Systems in Russia Yuri Gerasimov, Alexander Seliverstov Abstract – Nacrtak A field-based study was performed to broaden our knowledge of industrial round-wood (IRW) losses associated with most applicable motor-manual (MM) and fully mechanized (FM) harvesting systems in Russia. Observations were made for five harvesting systems, namely cut-to-length (MM CTL and FM CTL), full-tree (MM FT and FM FT), and treelength (MM TL), during the felling, skidding/forwarding, processing, sorting, and loading operations. Damage to IRW was examined in 15 logging companies in Karelia. There were about 23 400 measured logs in 17 harvesting sites during summer and winter seasons. The damages detected were broken down into four groups: mechanical damage, processing defects, contamination with dirt, and deviations from the desired log dimensions. The results were then compared with the effective quality requirements in a given logging company, and the IRW volume loss was determined in terms of the reject rate and value loss per unit volume in the context of a harvesting system. Mechanical damage (torn and loosened grain, cuts in stemwood, and gouges made by grapples), processing defects (branches, log end splits and cracks) and contamination with dirt were the most frequent types of damage. The MM CTL and FM CTL systems provided the minimum losses (the reject rate was 2% of observed logs and 0.5–0.6 per m3 of total industrial wood). The FT systems resulted in somewhat lower but still acceptable quality (MM: 4% and 1.1; FM: 3% and 0.9). The quality of wood harvested with the MM TL system turned out to be the lowest (5 % and 1.4), especially in the summer season. The total annual losses in IRW value at the companies studied were estimated as 1.0 per IRW m3 or 1.8 million.

Key words: volume loss, value loss, sawlog, pulpwood, pine, spruce, birch, cut-to-length, tree-length, full-tree

1. Introduction – Uvod The use of fully mechanized (FM) systems is increasing in wood harvesting operations globally. The increase in the use of mechanized harvesting systems has led to problems of log damage, including butt pull, log splitting during handling, and bucking and crushing of the log. Damage to a harvested log can occur during the felling, delimbing, bucking, skidding or forwarding, piling, loading, and hauling functions of wood harvesting. Several studies have reported log damages during harvesting operations. Greene and McNeel (1987, 1989) and Faust and Greene (1989) studied log damage by feller-bunchers with shear and saw heads. Volume losses of up to 4.5% were found when using a shear head, while losses of only 0.25% were found when using a saw head. Egan (1999) and Unver and Acar (2009) studied the log Croat. j. for. eng. 31(2010)2

damage caused by ground based skidding. The contamination of the logs with dirt was highlighted.

Brunberg et al. (2006), Jonsson and Hannrup (2007), and Nuutinen et al. (2010) studied the log damage caused by the harvester. They found that feed rollers equipped with steel spikes caused the greatest damage, leading to a 3% reduction in value.

According to Hamish et al. (2006), on average mechanical log-making systems lose 18% of the potential value compared to 11% for motor-manual (MM) systems. However, Connell (2003) and Spinelli et al.

(2010) reported that some mechanized harvesting operations have reduced the incidence of log damage due to mechanical handling. In addition, log damage and value loss associated with tree-length (TL) and full-tree (FT) harvesting systems were reported by Wang et al. (2004). The MM TL system

111

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Table 1 Basic data about conditions for sites and number of measured logs during summer and winter seasons (MM – moto-manual, FM – full mechanized) Tablica 1. Osnovni podaci o uvjetima na radili{tima te broju izmjerene oblovine tijekom ljetne i zimske sje~e (MM – djelomi~no mehanizirani sustavi, FM – potpuno mehanizirani sustavi) Technology – Tehnologija rada

Total logs sampled, pieces Ukupan broj trupaca, kom.

Pine Spruce Birch Bor Smreka Breza

Conditions – Radni uvjeti

1

10

2

11

10

12

11

13

12

15

15

15

16

1

17

14

FM FT

MM FT

14

112

Husqvarna 254XP Husqvarna 254XP TDT-55A

>5 >5 >5

Husqvarna 254XP TDT-55A LP-30B PL-1 Husqvarna 254XP TDT-55A LP-30B PL-1 Timberjack 850 ML-136 Hitachi Zaxis 230 Timberjack 850 Timberjack 460D Hitachi Zaxis 230LC

16 19 16 >5 16 19 16 >5 7 23 5 3 0.5 5

Pulp-wood – Celuloza

9

Sawlog – Trupci

5

Pulp-wood – Celuloza

8

Sawlog – Trupci

5

Pulp-wood – Celuloza

7

Sawlog – Trupci

3

Other – Ostalo

6

Birch – Breza

1

Spruce – Smreka

5

Pine – Bor

1

7 1 12 1 >5 1 2.5 1.5 2,5 0.5 4 3.5 1 1 1 0.5 5 2.5 1 8 14 >5 30 >5 >5 >5

Stem volume, m3 Obujam stabla, m3

4

Husqvarna 254XP John Deere 1110D Husqvarna 254XP John Deere 1110D Husqvarna 254XP John Deere 1010D John Deere 1270D John Deere 1410D John Deere 1270D John Deere 1410D Volvo EC210BLC Timberjack 1110D John Deere 1070D John Deere 1010D John Deere 1270D John Deere 1410D Timberjack 1270D Timberjack 1410D Valmet 911.3 Valmet 840.3 Husqvarna 254XP Axe TDT-55A Husqvarna 256XP Axe TDT-55A

0.257

30

20

50

–

300

300

300

300

300

300

0.313

10

30

40

20

300

300

–

–

300

300

0.300

20

20

40

20

–

–

300

300

–

–

0.303

40

50

10

–

300

300

300

300

–

300

0.130

70

–

30

–

–

–

–

–

–

300

0.356

10

80

10

–

300

300

300

300

–

300

0.328

10

50

20

20

300

–

300

300

–

–

0.270

60

20

20

–

300

300

300

300

–

300

0.266

40

10

50

–

300

300

300

300

–

300

0.270

60

20

20

–

300

300

300

–

–

300

Winter Zima

0.641

20

30

30

20

300

300

300

300

300

300

Summer/Clay loam Ljeto/glinovita ilova~a

0.267

10

40

30

20

300

300

300

300

300

300

Summer/Clay loam Ljeto/glinovita ilova~a

0.230

30

10

60

–

300

300

300

300

–

–

Winter Zima

0.254

50

30

20

–

300

300

300

300

300

300

Summer/Clay loam Ljeto/glinovita ilova~a

0.254

50

30

20

–

300

300

300

300

300

300

Winter Zima

0.276

50

40

10

–

300

300

300

300

–

300

Summer/Silt loam Ljeto/pjeskovita ilova~a

0.234

40

50

10

–

300

300

300

300

–

300

Season/soil Godi{nje doba/tlo

9

Experience, years Radno iskustvo, god.

3

Equipment used Kori{tena oprema

8

System – Sustav

2

MM CTL

7

FM CTL

Company – Poduze}e

1

MM TL

Site – Radili{te

Species composition, % Udio vrsta, %

Winter Zima Summer/Clay loam Ljeto/glinovita ilova~a Summer/Clay loam Ljeto/glinovita ilova~a Winter Zima Winter Zima Winter Zima Summer/Silt loam Ljeto/pra{kasta ilova~a Summer/Silt loam Ljeto/pjeskovita ilova~a Summer/Silt loam Ljeto/pjeskovita ilova~a Summer/Clay loam Ljeto/glinovita ilova~a

Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

with chain-saws and skidders and the FM FT system with feller-bunchers and skidders were studied. Volume losses of up to 6.1% and 1.1% were found and value losses of up to 6.0 US $/m3 and 1.5 US $/m3 were found when using MM TL and FM FT systems, respectively. The majority of the value loss was caused during the felling function when using an MM TL system.

The world’s best harvesting operations using the modern cut-to-length (CTL) machinery – many of them in Nordic countries – are currently losing 4–5% of the wood value of forests at harvest (Murphy 2005). However, wood harvesting operations in many countries, such as Russia, using a number of different harvesting systems, such as MM FT, FM FT, MM CTL, FM CTL, and MM TL, have shown losses of 11–18% of the wood value at harvest (Marshall et al.

2006). Certainly, the influence of wood quality on industrial round-wood (IRW) value cannot be ignored

Y. Gerasimov and A. Seliverstov

when comparing different technologies. This is determined by evaluating it in accordance with the quality specifications in the customer contracts as well as other quality requirements. To remain competitive, logging companies should minimize the wood loss at the time of harvest by using a more advanced harvesting technology. The major objective of this study was to identify the major sources of damage to IRW arising from applied harvesting systems in Russia to minimize this damage loss.

2. Methods and data – Metode i podaci The Republic of Karelia was selected as the study region, because its territory is very representative in terms of a wide range of harvesting methods, systems, and equipment used and in terms of nearly all harvesting technologies being employed in different natural conditions typical for northwest Russia. The

Fig. 1 Measurement of logs in MM CTL harvesting (chainsaw + forwarder) Slika 1. Izmjera oblovine pri djelomi~no mehaniziranom pridobivanju drva sortimentnom metodom (motorna pila i forvarder) Croat. j. for. eng. 31(2010)2

113

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111â&#x20AC;&#x201C;126)

study was performed in 2007â&#x20AC;&#x201C;2009 and involved 15 logging companies that provide approximately 35% of the total harvesting volume in Karelia (2.2 million m3 per year). The selected companies perform harvesting operations across the whole territory of Karelia in different conditions and apply MM CTL, FM CTL, MM FT, FM FT, and MM TL harvesting systems using both Russian and foreign machinery (Gerasimov and Sokolov 2009).

A common approach was used for field data collection directly at harvesting sites in the actual working conditions (Table 1). The harvested stands were not managed or thinned before the final felling. A typical study stand was mixed in terms of tree age and species. The tree species composition included spruce (31% on average), pine (35%), birch (28%), and aspen (6%). The average stem volumes of the harvesting sites varied between 0.13 and 0.64 m3 with the average value 0.29 m3. The average growing stock of stands in the studied regions was 150 m3 per ha

with tree density of approximately 520 trees per ha.

The typical soils in the test areas were silt loam, clay loam, and sandy loam. The harvesting sites were on flat terrain.

According to the methodology used, the required number of logs to be measured equals 300 both for each species and for each assortment per harvesting site separately in the winter and summer seasons.

The total number of observed logs was 23 400, and the number of observed harvesting sites was 17 including 7 in winter and 10 in summer (Table 1).

All the measurement results were registered on checklists using a data collector. In CTL harvesting systems, the logs were measured at the felling site and at the road-site landing (Fig. 1 and 2).

In TL harvesting, the logs were measured both at the felling site before skidding and at the road-site landing after skidding and piling. In addition, the logs were measured at the bucking and sorting line of the central processing yard (Fig. 3).

Fig. 2 Measurement of logs in FM CTL harvesting (harvester + forwarder) Slika 2. Izmjera oblovine pri potpuno mehaniziranom pridobivanju drva sortimentnom metodom (harvester i forvarder) 114

Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Y. Gerasimov and A. Seliverstov

Fig. 3 Measurements during MM TL (chainsaw + skidder) Slika 3. Izmjera pri djelomi~no mehaniziranom pridobivanju drva deblovnom metodom (motorna pila i skider) In FT harvesting, the logs were measured at the road-site landing and logs at the cross-cutting and sorting line of the central processing yard (Fig. 4 and 5).

The IRW damage evaluation was based on a number of damage indicators, which are regulated by relevant national standards and forest industry specifications, as follows (Fig. 6). Þ mechanical damage, which occurs during skidding, sorting, piling, and transportation of timber; there are the following types of damage: torn and loosened grain, barked stem, cuts (damage by chainsaw, skidder cable), and gouges made by grapples, Þ processing defects, including unprocessed branches and defects caused by improper tree-felling and cross-cutting, namely: log end (butt and top) splits, cracks, log splitting, and snipes, Croat. j. for. eng. 31(2010)2

Þ contamination with dirt, Þ deviation of IRW dimensions, including loglength allowances and tolerances, as well as the grades and the maximum butt and minimum top diameters of assortments.

The results were then compared with the effective quality requirements in a given logging company and the percentage of rejected logs was determined. The obtained estimates for all the measured parameters were integrated into one indicator – the so-called reject rate.

The quality requirements for IRW of various species, grades, and end-uses (sawlog, pulpwood, etc.) were determined in the contract between the logging company and the IRW buyer, that is, in the technical specifications. The specifications include the following: tree species, harvesting schedule, dimensions, requirements for processing and quality requirements, such as compliance with relevant na-

115

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Fig. 4 Measurements during MM FT (chainsaw + skidder + delimber) Slika 4. Izmjera pri djelomi~no mehaniziranom pridobivanju drva stablovnom metodom (motorna pila, skider, procesor)

tional standards, for example the national technical specifications for export (TU 13-2-12-96, TU 13-2-1-95), the national standards for the domestic market (GOST 9462-88, GOST 9463-88), or other technical specifications used in trade contracts, or by using internal log quality specifications of the given logging company in the case where the logs were to be used within the company. Where contamination of dirt was not acceptable according to the contractual specifications, logs were rejected if more than 15% of the log side surface area or 50% of the log end was contaminated.

Logging companies also develop their own additional specifications for grading and piling logs, defining the length and diameter of piles, as well as the most preferable log length. Quality requirements of the measured logs of various species and end-uses are shown in Tables 2 and 3.

If a log complies with both the quality and dimension requirements, it is accepted. If a log does

116

not comply with the above-mentioned requirements, it is rejected or transferred to another grade according to its quality: sawlog to pulpwood, pulpwood to fuel wood.

The IRW damage was analysed in terms of value losses. The losses in IRW value in the context of a logging company and a harvesting system were estimated as follows: L = Rpsl ´ Ppsl(Cpsl – Cppw – C') + + Rssl ´ Pssl ´ (Cssl – Cspw – C') + + Rbsl ´ Pbsl ´ (Cbsl – Cpbw – C') + + Rppw ´ Ppsl ´ (Cppw – Cpfw – C') + + Rspw ´ Pspw ´ (Cspw – Csfw – C') + + Rpbw ´ Pbpw ´ (Cpbw – Cbfw – C') Where: L value losses due to IRW damage at harvest, /m3 R average reject rate of the IRW assortment Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

P proportion of the assortment production in the total volume of industrial wood at the given logging company C average EXW (ex works) price of the IRW assortment at the road-site or central processing yard, /m3 C’ additional expenses associated with load, unload and transport of rejected wood, /m3 psl, ssl, bsl indexes for pine, spruce, and birch sawlogs ppw, spw, bpw indexes for pine, spruce, and birch pulpwood fw index for fuel wood A log pricing system was developed based on IRW market prices from the Karelia timber market reports (Timber Prices 2010). A monetary value was assigned to the IRW based on its tree species, assortment, and delivery terms.

Y. Gerasimov and A. Seliverstov

3. Results – Rezultati To summarize, Table 4 shows the distribution of reject rates by assortment in 17 studied harvesting sites.

The results include all harvesting systems and apply to both the winter and summer seasons. A rejected log sometime had two or more types of damages. In that case the log was rejected with several reasons. The difference between results in Table 4 and Tables 5 – 7 (original tables by damage type for coniferous sawlogs, birch veneer logs, and pulpwood) shows a common fact that a rejected log had more than one type of damage.

The MM TL system caused the highest reject rate both for sawlogs (7% of observed logs in winter and 9–10% in summer) and for pulpwood (3% in winter and 7–8% in summer). The lowest reject rate for sawlogs was provided by the FM CTL system (3%). The

Fig. 5 Measurements during FM FT harvesting (feller buncher + skidder + delimber) Slika 5. Izmjera pri potpuno mehaniziranom pridobivanju drva stablovnom metodom (feller buncher, skider, procesor) Croat. j. for. eng. 31(2010)2

117

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111â&#x20AC;&#x201C;126)

Fig. 6 Typical IRW damages regulated by relevant national standards and log quality specifications Slika 6. Tipi~na o{te}enja oblovine propisana va`e}im nacionalnim normama 118

Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Y. Gerasimov and A. Seliverstov

Table 2 Quality requirements for saw and veneer logs in domestic and export markets Tablica 2. Zahtijevi za kakvo}om pilanskih i furnirskih trupaca za doma}e i inozemno tr`i{te Sawlogs – Pilanski trupci Damage type Vrsta o{te}enja

1. Mechanical damage – Mehani~ka o{te}enja

Pine – Bor

Birch veneer logs Brezovi furnirski trupci

Spruce – Smreka

Export Izvoz

Domestic Doma}e tr`i{te

Export Izvoz

Domestic Doma}e tr`i{te

Export Izvoz

TU 13-2-12-96 Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

TU 13-2-12-96 Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

TU 13-2-8-96

2. Processing defects – Gre{ke pri izradbi drva Branches Ostatci grana

TU 13-2-12-96 l < 10/20 mm 4 d <50/60 mm 4

GOST 9463-88

TU 13-2-12-96 l < 10 mm 4 d <50 mm 4

GOST 9463-88

TU 13-2-8-96

Log end splits, cracks Raspucano ~elo, pukotine

Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

TU 13-2-8-96

Log end splinters Raspukline

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4

Butt trimming Obrada ~ela

Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

Not acceptable 4 Nije prihvatljivo 4

GOST 9463-88

TU 13-2-8-96

3. Contamination with dirt – Zabla}enost drva

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

4.9; 5.5 (0 / +6) 4.0 (0 / +6) 4.3; 4.6; 6.1 (+5 / +8)

5.0; 5.5; 6.0; 6.1 (0 / +10) 6.1; 4.0; 3.1 (+3 / +5) additional – dodatno 4.0; 4.3 (+3 / +10)

5.5 (+3 /+6) 5.5 (0 / +6) 4.05 (0 / +6)

5.0; 5.5; 6.0; 6.1 (0 /+10) additional – dodatno 4.0; 4.3; 5.2 (+3 /+10)

3.3; 6.0 (0 / +10) 4.4; 5.0 (0 / +5) 3.3 (0 / +5)

Maximum diameter of the butt end without bark, cm Najve}i promjer na debljem kraju bez kore, cm

55.0* 34.0

75 42.0*

55.0* 40.0* 14.9

75 52.0* 36.0 56.0

65.0* 55.0* 50.0*

Minimum diameter of the butt end without bark, cm Najmanji promjer na debljem kraju bez kore, cm

18.0* 15.0 15.0*

16.0 14.0 11.0

18.0* 17.0* 16.0* 12.0

16.0 14.0

25* 18.0*

4. Dimension non-compliance – Neodgovaraju}e dimenzije

Length, m (allowance, cm) Duljina, m (nadmjera, cm)

4 * l d

– Quality requirements in contractual specifications – Ugovorni zahtjev za kakvo}om drva – Diameter over bark – Promjer s korom – Max. acceptable branch length – Najve}a prihvatljiva duljina ~aprlja (grane) – branch diameter – promjer grane

lowest reject rate for pulpwood (2%) was registered with the MM CTL system (Table 4).

The study of the quality of IRW harvested with the MM CTL system demonstrated that log end splits and cracks (up to 3% of observed logs), as well as cuts by chainsaws and gouges by forwarders’ grapples during loading operations (up to 2%), were the most common types of processing defects (Tables 5-7). The reject rate was about 5% in winter and 4% in summer for coniferous sawlogs and about 1% for pulpwood regardless of the harvesting season (Table 4).

Croat. j. for. eng. 31(2010)2

The FM CTL system, in both winter and summer, was mostly associated with the following types of defects (Tables 5 and 7): unprocessed branches (2% of observed logs), log end splits and cracks during felling and bucking (2%), and log surface damage.

The latter appeared in the form of damage by delimbing and feeding mechanisms of the harvester head during delimbing, that is, torn and loosened grain (2%). This damage was accompanied by barked stems or even lost layers of stemwood. Logs damaged by harvester head saws (cuts) or forwarders’ grapples were rare (about 1%).

119

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Table 3 Pulpwood quality requirements for domestic and export markets Tablica 3. Zahtjevi za kakvo}om celuloznoga drva za doma}e i inozemno tr`i{te Pine – Bor

Damage type Vrsta o{te}enja

Spruce – Smreka

Export Izvoz

Birch – Breza

Domestic Doma}e tr`i{te

Export Izvoz

Domestic Doma}e tr`i{te

Export Izvoz

GOST 9463-88; TU 13-2-10-96

TU 13-2-1-95; TU 13-2-10-96; TU 13-2-11-96. l <20mm 4

GOST 9463-88; TU 13-2-10-96

Not acceptable 4 Nije prihvatljivo 4

1. Processing defects (branches) – Gre{ke pri izradbi (grane)

GOST 9463-88; TU 13-2-10-96

GOST 9463-88

GOST 9463-88; TU 13-2-10-96

2. Contamination with dirt – Zabla}enost drva

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

Not acceptable 4 Nije prihvatljivo 4

3. Size non-compliance – Neodgovaraju}e dimenzije 2.4; 3.6; 4.8; 6.0 (–5/+15); 4.0; 5.5 (–5/+15); 1.2 (–2 /+ 2); 4.0; 5.5 (0 / +10); 2.4 (–2 /+ 2); 3.0; 4.0; 6.0 (–10 / +10) 3.6 and 4.8 (–15 / 15); 4.0 and 5.5 (–10 / +10); 2.4 and 3.6 (+3 / +5)

Length, m (allowance, cm) Duljina, m (nadmjera, cm)

3.0; 4.0; 6.0 (0 / +10)

3.0–6.0 (–20 / +20)

3.0; 4.8. 6.0 (0 / +10)

Maximum diameter of the butt end without bark, cm Najve}i promjer na debljem kraju bez kore, cm

60.0

40.0

40.0

60.0; 50.0; 36.0

60.0

Minimum diameter of the butt end without bark, cm Najmanji promjer na debljem kraju bez kore, cm

8.0; 6.0

6.0

8.0*

16.0; 6.0

16.0; 6.0

4 – Quality requirements in contractual specifications – Ugovorni zahtjev za kakvo}om drva * – Diameter over bark – Promjer s korom l – Max. acceptable branch length – Najve}a prihvatljiva duljina ~aprlja (grane)

Table 4 Reject rates of round-wood (% of observed logs) at the studied harvesting sites by harvesting system and season Tablica 4. Stopa odbacivanja oblovine (% od oblovine uzorka) na istra`ivanim radili{tima s obzirom na sustav pridobivanja drva i sezonu radova System Sustav

Pine sawlog Borovi trupci

Spruce sawlog Smrekovi trupci

Pine pulpwood Borova celuloza

Spruce pulpwood Smrekova celuloza

Birch veneer logs Brezovi furnirski trupci

Birch pulpwood Brezova celuloza

Winter Zima

Summer Ljeto

Winter Zima

Summer Ljeto

Winter Zima

Summer Ljeto

Winter Zima

Summer Ljeto

Winter Zima

Summer Ljeto

Winter Zima

Summer Ljeto

MM CTL

4.0

3.7

5.3

4.0

1.0

1.3

1.0

1.3

2.7

3.0

1.7

1.3

FM CTL

2.7

3.3

3.0

3.3

2.0

2.0

1.7

1.7

n/a

n/a

2.0

2.1

MM TL

7.3

9.0

7.0

10.3

2.7

8.3

2.7

8.0

6

8.3

2.3

7.3

MM FT

4.7

7.7

6.3

7.3

2.1

6.0

1.8

6.3

3.3

7.0

1.7

5.0

FM FT

5.3

5.0

5.0

5.3

2.3

2.0

2.7

2.3

n/a

n/a

2.3

2.3

When harvester operators followed all work requirements and instructions, the reject rate was less than 3% for coniferous sawlogs harvested with the FM CTL system, and less than 2% for coniferous pulpwood, regardless of the season. The FM CTL system also ensured efficient cross-cutting of the stems with the required length allowance, normally +(0–4) cm, which maximized the amount of received

120

IRW assortments, unlike the MM CTL system, where the allowance was mostly +(5–10) cm.

For the MM TL and MM FT systems, regardless of the season, the following types of damage were typical (Tables 5-7): torn and loosened grain (2–3% of observed logs) and cuts in stemwood and gouges made by grapples (2–3%). Less frequent were unprocessed branches (1%) and log end splits and cracks Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Y. Gerasimov and A. Seliverstov

Table 5 Losses of coniferous sawlogs volume (% of observed logs) at the studied harvesting sites by harvesting system and damage type (PSL – pine, SSL – spruce) Tablica 5. Gubitci obujma pilanskih trupaca ~etinja~a (% od oblovine uzorka) na istra`ivanima radili{tima s obzirom na sustav pridobivanja drva i vrstu o{te}enja (PSL – bor, SSL – smreka) Mechanical damage – Mehani~ka o{te}enja System Sustav

Torn Pokidana drvna vlakanca Winter Zima

Processing defect – Gre{ke pri izradbi drva

Cuts, gouges Zarezan pla{t, odlupljeno drvo

Summer Ljeto

Winter Zima

Unprocessed branches Neokresane grane

Summer Ljeto

Winter Zima

Contamination with dirt Zabla}enost drva

Splits, cracks Raspukline, pukotine

Summer Ljeto

Winter Zima

Summer Ljeto

Summer Ljeto

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

PSL

SSL

MM CTL

2.0

2.3

0

0

0

0

1.7

2.0

0

0

0

0

2.3

2.7

2.0

1.3

0

0

FM CTL

0.7

0.7

1.5

1.7

1.3

1.7

1.0

1.0

1.7

2.1

1.7

1.7

1.7

2.1

1.5

1.7

0

0

MM TL

2.0

2.7

2.7

3.0

2.3

2.7

2.0

2.1

0.9

1.0

1.1

1.3

1.0

0.9

1.0

1.3

8.1

7.3

MM FT

2.0

2.7

2.0

2.7

2.3

2.0

1.7

1.7

1.0

1.3

1.1

1.3

1.0

1.1

1.3

1.0

5.0

4.3

FM FT

3.0

3.1

1.7

1.3

1.3

1.7

1.3

1.3

1.3

1.3

1.3

1.7

2.1

1.9

1.7

1.3

0

0

Table 6 Losses of birch veneer logs volume (% of observed logs) at the studied harvesting sites by harvesting system and damage type Tablica 6. Gubitci obujma brezovih furnirskih trupaca (% od oblovine uzorka) na istra`ivanim radili{tima s obzirom na sustav pridobivanja drva i vrstu o{te}enja Mechanical damage – Mehani~ka o{te}enja System Sustav

Torn Pokidana drvna vlakanca Winter Zima

Processing defect – Gre{ke pri izradbi drva

Cuts, gouges Zarezan pla{t, odlupljeno drvo

Summer Ljeto

Winter Zima

Unprocessed branches Neokresane grane

Summer Ljeto r

Winter Zima

Summer Ljeto

Splits, cracks Raspukline, pukotine Winter Zima

Summer Ljeto

Contamination with dirt Zabla}enost drva Summer Ljeto

MM CTL

0

0

1.8

1.7

0

0

1.7

1.7

0

MM TL

2.7

3.0

2.7

2.0

1.0

1.1

1.1

1.3

8.3

MM FT

1.7

1.7

2.3

2.0

1.3

1.3

1

1.3

6.0

Table 7 Losses of pulpwood volume (% of observed logs) at the studied harvesting sites by harvesting system and damage type (PPW – pine, SPW – spruce, BPW – birch) Tablica 7. Gubitci obujma drva za celulozu (% od oblovine uzorka) na istra`ivanim radili{tima s obzirom na sustav pridobivanja drva i vrstu o{te}enja (PPW – bor, SPW – smreka, BPW – breza) Unprocessed branches – Neokresane grane System Sustav

Winter – Zima

Contamination with dirt – Zabla}enost drva

Summer – Ljeto

Summer – Ljeto

PPW

SPW

BPW

PPW

SPW

BPW

PPW

SPW

BPW

MM CTL

0.7

0.7

0.7

0.7

0.7

1.0

0

0

0

FM CTL

1.7

1.3

1.7

1.7

1.7

2.0

0

0

0

MM TL

0.9

1.0

0.9

1.0

1.3

1.0

9.1

8.3

8.0

MM FT

1.0

1.0

0.7

1.0

0.9

0.9

5.0

5.7

4.0

FM FT

1.3

1.7

1.3

1.3

1.7

1.7

0

0

0

(1%). In summer, contamination with dirt was also found (up to 9% for the MM TL and 6% for the MM FT). For spruce and pine sawlogs, the following reject rates were registered (Table 4): 6–7% for spruce and 5–7% for pine in winter, and 7–10% for spruce Croat. j. for. eng. 31(2010)2

and 8–9% for pine in summer. The maximum reject rate was registered for the sawlogs intended for the export market. For birch pulpwood, this figure reached 2% in winter and up to 7% for the tree-length system and 5% for the MM FT system in summer. For

121

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Table 8 Losses of industrial round-wood volume and value at the studied companies by harvesting system Tablica 8. Gubitci obujma i vrijednosti obloga drva za istra`ivana poduze}a s obzirom na sustave pridobivanja drva Annual harvest – Godi{nji etat, 1000 m3 System Sustav

Company Poduze}e

Total Ukupno

Fuel round-wood Ogrjevno drvo

Industrial round-wood Tehni~ka oblovina

Volume loss Gubitak obujma %

1000 m3

Value loss Gubitak vrijednosti /m3

1000

MM CTL

2, 3, 7, 8, 9

363.0

56.2

306.8

1.8

5.64

0.51

156.6

FM CTL

1, 2, 3, 4, 5,6

503.0

64.8

438.2

2.3

9.98

0.65

286.2

MM TL

1, 5, 10, 11, 12 13, 14

935.4

155.6

779.8

5.0

39.14

1.38

1074.9

MM FT

15

67.1

6.7

60.4

4.2

2.52

1.04

62.7

FM FT

1, 14

318.2

31.8

286.4

3.3

9.56

0.86

247.4

2186.7

315.1

1871.6

3.6

66.84

0.98

1827.8

Total Ukupno

pine and spruce pulpwood, the reject rates were up to 3% and 2% in winter, respectively, and 3% in summer.

FM FT harvesting in both winter and summer was mostly associated with the following types of timber defects (Tables 5 and 7): cuts in stemwood and gouges made by grapples (2% of observed logs), log end splits and cracks (2%), torn and loosened grain (2–3%), and unprocessed branches (2%). The reject rate for spruce and pine sawlogs (Table 4) was about 5%, regardless of the season. For birch, pine, and spruce pulpwood, this figure was about 3% in winter and up to 2% in summer.

The seasonality of harvesting operations has a negative impact on the quality of harvested wood; this pertains to the MM CTL, MM TL, and MM FT systems.

In 15 studied companies, volume losses of IRW (in terms of the reject rate as a percentage of total IRW on average per year) were found by harvesting system as follows: MM CTL: 1.8%; FT CTL: 2.3%; MM TL: 5.0%; MM FT: 4.2%; and FM FT: 3.3% (Table 8).

The total average volume loss of IRW in the studied companies was 3.6% or around 67 000 m3 of IRW per year. Value losses of IRW (in terms of value loss per unit volume of IRW) were found by harvesting system as follows: MM CTL: 0.51; FT CTL: 0.65; MM TL: 1.38; MM FT: 1.04; and FM FT: 0.86. The total average value loss of IRW in the studied companies was 0.98 per m3 of IRW or around 1.8 million per year.

4. Conclusion and recommendations –Zaklju~ci i preporuke The analysis of the obtained results indicates that CTL harvesting can ensure the highest quality of harvested wood (reject rate below 3% of observed

122

logs) in all the studied companies, with different species composition. The FT harvesting systems demonstrated acceptable IRW quality (reject rate about 3–5%). The quality of wood in TL harvesting was low (reject rate over 6%), particularly in summer (reject rate up to 10%).

Over 50% of the harvesting sites in Russia are on wet and soft soil terrain and the proportion of sandy soils is small in Russian forests in comparison with loams and clays (Gerasimov and Katarov 2010). The MM TL system causes the highest reject rate due to the fact that debranched tree-length logs are bunched and skidded by a cable skidder in this type of terrain, which leads to contamination with dirt and other damages. The MM FT has the same reason for high reject rate, particularly in summer season, but branches in some extend protect stemwood from damages. The FM FT is largely free from this disadvantage due to bunching by a feller-buncher and skidding by a grapple skidder. Regardless of the season, the CTL systems show the lowest rejection rate due to using forwarding (round-wood is carried out on a trailer) instead of skidding (logs are dragged out of the forest over soft soils). Therefore, the selection of the harvesting system has to be adapted to the most common soil terrains in order to reduce wood losses.

Mechanical damage (torn and loosened grain, cuts in stemwood, and gouges made by grapples), processing defects (branches, log end splits and cracks), and contamination with dirt were the most frequent types of damage. On the whole, damage to IRW in terms of volume loss did not differ much from what is obtained with FT systems in the USA (Wang et al.

2004) and CTL systems in Finland and Russia (Eronen et al. 2000, Syunev and Seliverstov 2006).

Certainly, the improvement of harvesting operations is needed for the reduction of IRW losses even Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

in the same harvest system. Loggers (operators and lumberjacks) need to pay more attention to value, rather than volume alone: this could be accomplished by the development of a payment system and harvesting instructions for utilizing the forest resources better by not damaging valuable logs. The seasonality could be taken into account: the reject rate is higher for the MM CTL system in winter and for MM TL and MM FT systems in summer. Bed logs under piles should be used for IRW piling at roadsite landings, depending on the dirt conditions. Operators need to perform the maintenance of harvesting machines properly (e.g. adjustment of the delimbing and feeding mechanisms of harvester heads, sharpening the delimbing knives, cleaning the rollers to remove bark and timber residue, etc.). A harvester head must match both the base machine and the site conditions (species composition, tree size).

The development of new guidelines and corresponding training to minimize IRW damage occurring during wood harvesting are needed as well (Syunev et al. 2008). However, prior to specializing in operating sophisticated machines, as a harvester, a forwarder, and a feller-buncher, an operator is required to have a relevant vocational education.

The potential improvement of rejection rate was roughly estimated from the best practices in the studied logging companies in comparison with the common practices. If all the discovered shortcomings typical for FM CTL and FM FT harvesting were eliminated, it should be possible to decrease the reject rates by approximately 20% and 25%, respectively. It should be noted that bucking optimization of the FM CTL harvesting system allows an increase in the amount of received IRW assortments. Improvements made to the MM CTL system would enable the reject rate to be reduced by approximately 15%. In MM TL and MM FT, the potential reductions in the amount of damaged logs could reach 20% and 15%, respectively.

The IRW damage in terms of value loss per unit volume in the studied companies may not seem important. This is especially true when looking at the lack of difference between FM CTL and MM CTL systems. However, the switch from the traditional motor-manual TL to CTL gives an average savings value of 0.8 per m3 of industrial wood, or around 100 000 per year for an average size logging company. With the initial investment cost in CTL machines of several hundred thousand euros (a forwarder costs over 200 000, a harvester over 300 000), the switch from motor-manual TL and FT systems to an FM CTL system might be worth it in the long run, but the switch to an MM CTL system might be justified in the medium run. Additional analysis Croat. j. for. eng. 31(2010)2

Y. Gerasimov and A. Seliverstov

is needed before the system selection is made, taking into account that the efficiency of a particular harvesting system depends on a number of criteria. The economic benefits, which are the most applicable in practice, are evaluated by productivity and costs (Adebayo et al. 2007, Konovalov and Seliverstov 2008). Special attention has been paid to comfortable and safe working conditions in felling operations.

This will make harvesting work more attractive to young people and employment in a logging company more desired (Gerasimov and Sokolov 2009).

Environmental criteria and terrain conditions include dirt damage, damage to undergrowth or remaining trees, and so on (Syunev et al. 2009).

This study has been focused on various quality requirements and wood harvesting practices at the logging companies in Karelia. This fact might limit the application of obtained results in other regions in Russia. Moreover, further research is needed to determine the influence of different quality requirements (for the domestic market, export, individual customers) and bucking on IRW volume and value losses. Improper bucking might not damage the log in a physical sense, but could damage the potential value gained from bucking correctly (Wang at al.

2004, Marshall et al. 2006). Taking into account natural and production conditions in Russia, it is necessary to improve the design of the harvester head delimbing and feeding mechanism, in order to ensure its higher efficiency in the processing of trees with crooked trunks and tapering of large branches of deciduous trees. More in-depth analysis of birch veneer log degradation in fully mechanized CTL and FT systems is also needed due to substantial increases in the demand for veneer products.

Acknowledgements – Zahvala The work was carried out for the project »Wood Harvesting and Logistics», financed by the European Union through the Finnish Funding Agency for Technology and Innovation (TEKES), and for the project »Comparison of Wood Harvesting Methods in the Republic of Karelia», funded by the EU-Russia programme »Euregio Karelia Neighbourhood«.

5. References – Literatura Adebayo, A. B., Han, H., Johnson, L., 2007: Productivity and cost of cut-to-length and whole-tree harvesting in a mixed-conifer stand. Forest Products Journal 57(6): 59–69.

Brunberg, T., Hofsten, H., Jonsson, M. 2006: Kartlägging och värdering av dubbskador (Stud damage to logs – research and evaluation). Skogsforsk Resultat 2006(18), 4 p.

123

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Connell, M. J., 2003: Log presentation: log damage arising from mechanical harvesting or processing. CSIRO Forestry and Forest Products. Victoria, 62 p.

Egan, A. F., 1999: Residual stand damage after shovel logging and conventional ground skidding in an Appalachian hardwood stand. Forest Products Journal 49(6): 88–92.

Eronen, J., Asikainen, A., Uusitalo, J., Sikanen, L., 2000: Control of log end checks during bucking with a modified single-grip harvester. Forest Products Journal 50(4): 65–70.

Faust, T. D., Greene, W. D., 1989: Effects of felling head type on tensile strength of southern pine dimension lumber. Forest Products Journal 39(11–12): 82–84.

Gerasimov, Y., Katarov, V., 2010: Effect of Bogie Track and Slash Reinforcement on Sinkage and Soil Compaction in Soft Terrains. Croatian Journal of Forest Engineering 31(1): 35–45.

Gerasimov, Y., Sokolov, A., 2008: Ergonomic characterization of harvesting work in Karelia. Croatian Journal of Forest Engineering 30(2): 159–170.

GOST 2292-88, 1990: State standard. Roundwood: marking, transportation, methods of measurement and acceptance. Moscow. Gosstandart, 12 p.

GOST 9463-88, 1990: State standard. Roundwood of coniferous species. Moscow. Gosstandart. 14 p.

Greene, W. D., McNeel, J. F., 1987: Productivity, costs, and levels of butt damage with a Bell Model T feller-buncher.

Forest Products Journal 37(11–12): 70–74.

Greene, W. D., McNeel, J. F., 1989: Potential costs of shear damage in a southern pine chip-n-saw mill. Forest Products Journal 39 (5): 12–18.

Jonsson, P., Hannrup, B., 2007: Virkesvardestest 2006 – virkesskador (Timber-value tests 2006 – timber damage and defects). Skogsforsk Resultat 2007(7), 4 p.

Konovalov, A. P., Seliverstov, A. A., 2008: Logging technologies: an assessment of techno-economic factors. Forestry Expert 1(2008): 76–81.

Marshall, H., Murphy, G. E., Boston, K., 2006. Evaluation of the economic impacts of length and diameter measurement error on mechanical processors and harvesters operating in pine stands. Canadian Journal of Forest Research 36: 1661–1673.

McNeel, J. F., Czerepinski, F., 1987: Effect of felling head design on shear-related damage in southern yellow pine.

Southern Journal of Applied Forestry 11(1): 3–6.

Murphy, G. E, 2005: Technology Aids Value Recovery. Focus on Forestry 18(2): 12.

Murphy, G., Twaddle, A. A., 1985: Techniques for the assessment and control of log value recovery in the New Zealand forest harvesting industry. In: Proceedings of the

124

9th Annual Meeting of Council on Forest Engineering, Mobile, AL, September 29 – October 2, 1985.

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.

Pickens, J. B., Lee, A., Lyon, G. W., 1992: Optimal bucking of Northern hardwoods. Northern Journal of Applied Forestry. 9(4): 149–152.

Sessions, J., 1988: Making better tree-bucking decisions in the woods. Journal of Forestry (10): 43–45.

Spinelli, R., Magagnotti, N., Nati, C., 2010: Comparison between mechanized and manual log-making in Italian poplar plantations. In: Forest Engineering Meeting the Needs of the Society and the Environment. Proceedings of 43 International Symposium on Forestry Mechanisation, July 11–14, 2010, Padova, Italy, 8 p.

Syunev, V., Seliverstov, A., 2006: The influence of assortment of logging on the quality of wood raw material. In: Actual Problems of Forest Industry. Volume 14. Bryansk, BSETA, 68–71 p.

Syunev, V., Sokolov, A., Konovalov, A., Katarov, V., Seliverstov, A., Gerasimov, Y., Karvinen, S., Välkky, E., 2009: Comparison of wood harvesting methods in the Republic of Karelia. Working Papers of the Finnish Forest Research Institute 120, 117 p.

Syunev, V., Sokolov, A., Seliverstov, A., Konovalov, A., Katarov, V., Gerasimov, Y., Välkky, E., Karvinen, S., 2008: Training needs analysis for operators of harvesters. Finnish Forest Research Institute, 11 p. <http://www.idanmetsatieto.info/ rus/cfmldocs/document.cfm?doc=show&doc_id=1200> (Accessed 1 July 2010).

Timber Prices, 2010: Average contract prices for major products of forest and woodworking industries, established in the Republic of Karelia. Official site of the Government of Karelia <http://www.gov.karelia.ru> (Accessed 1 July 2010).

TU 13-2-12-96, 1996: Technical specifications. Coniferous saw logs delivered to Finland. Khimki. TsNIIME, 20 p.

TU 13-2-1-95, 1995: Technical specifications. Pulpwood delivered to Finland. Khimki. TsNIIME, 15 p.

Unver, S., Acar, H. H., 2009: A damage prediction model for quantity loss of skidded spruce logs during ground base skidding in north eastern Turkey. Croatian Journal of Forest Engineering 30(1): 59–65.

Wang, J., LeDoux, C., Vanderberg, M., McNeel, J., 2004: Log damage and value loss associated with two ground-based harvesting systems in central Appalachia.

International Journal of Forest Engineering 15(1): 61–69.

Croat. j. for. eng. 31(2010)2

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

Y. Gerasimov and A. Seliverstov

Sa`etak

Gubitci obloga drva u sustavima pridobivanja drva u Rusiji Terenska su istra`ivanja provedena radi pro{irivanja spoznaja o gubitcima obloga drva za naj~e{}e kori{tene djelomi~no (MM) i potpuno (FM) mehanizirane sustave pridobivanja drva u Rusiji. Pra}enja su obuhvatila pet sustava pridobivanja drva s obzirom na sortimentnu (MM CTL i FM CTL), deblovnu (MM TL) i stablovnu (MM FT i FM FT) metodu izradbe drva, tijekom sje~e, privla~enja/izvo`enja, izradbe, uhrpavanja (razvrstavanja) i prihvata (utovara) drva. Istra`ivanje je provedeno u Republici Kareliji zbog {irokoga raspona razli~itih sustava pridobivanja drva i uporabe vozila na tom podru~ju, {to je ujedno i primjerni uzorak za podru~je sjeverozapadne Rusije, u razdoblju od 2007. do 2009. godine te je uklju~ivalo 15 poduze}a za pridobivanje drva koja godi{nje sijeku i izra|uju oko 35 % od ukupnoga etata u Kareliji (2,2 milijuna m3/god.). Odabrana poduze}a pridobivaju drvo u cijeloj Republici Kareliji te primjenjuju razli~ite sustave pridobivanja drva (MM CTL, FM CTL, MM FT, FM FT i MM TL) i koriste vozila ruskih i stranih proizvo|a~a.

Istra`ivanje je obuhvatilo 23 400 komada obloga drva na 17 radili{ta, a osnovni podatci o uvjetima na radili{tima te broju izmjerene oblovine tijekom sje~a prikazani su u tablici 1. Pri sortimentnoj metodi pridobivanja drva trupci su mjereni u sje~ini i na pomo}nom stovari{tu (slike 1 i 2). Pri deblovnoj metodi pridobivanja drva trupci su mjereni u sje~ini prije privla~enja drva, na pomo}nom stovari{tu nakon privla~enja i uhrpavanja drva te naknadno na glavnom stovari{tu prilikom trupljenja i razvrstavanja drva (slika 3). Pri stablovnoj metodi pridobivanja drva trupci su mjereni na pomo}nom stovari{tu te na glavnom stovari{tu prilikom prerezivanja i razvrstavanja drva (slike 4 i 5).

Gubitci su obloga drva procijenjeni na temelju broja o{te}enja. Oni su propisani odgovaraju}im dr`avnim normama i propisima {umarske industrije, kako slijedi (slika 6): a) mehani~ka o{te}enja, b) gre{ke pri izradbi drva, c) zabla}enost drva i d) odstupanja od odgovaraju}ih veli~ina (dimenzija). Rezultati su potom uspore|eni s va`e}im zahtjevima kakvo}e u pojedinim poduze}ima te je dobiven postotak odbijenih trupaca. Zahtjevi kakvo}e za trupce i celulozno drvo prema raznim vrstama drve}a te ciljano tr`i{te prodaje prikazani su u tablicama 2 i 3.

Dobivene procjene za sve mjerene veli~ine povezane su u jedan pokazatelj, tzv. stopu odbacivanja. Ako je trupac u skladu i s kakvo}om i s veli~inama (dimenzijama) tra`enih zahtjeva, ozna~uje se kao prihvatljiv. Nadalje, ako trupac nije u skladu bilo s kakvo}om bilo s izmjerenim veli~inama, odbija se ili se upu}uje za drugu namjenu, na primjer trupci postaju celulozno drvo, celulozno drvo postaje ogrjevno drvo.

Gubitci su obloga drva prikazani kao gubitci tr`i{ne vrijednosti drva pomo}u formule 1, dok su cijene drva preuzete s drvnoga tr`i{ta Republike Karelije. Tablica 4 prikazuje stopu odbacivanja oblovine na 17 istra`ivanih radili{ta pri svim sustavima pridobivanja drva tijekom zimske i ljetne sje~e drva. Odbijeni trupac mo`e sadr`avati dva o{te}enja i vi{e njih, {to je i prikazano u tablicama 5–7. Najvi{a se stopa odbacivanja oblovine javlja pri deblovnoj metodi i djelomi~no mehaniziranom sustavu pridobivanja drva. Tako je stopa odbacivanja za pilanske i furnirske trupce iznosila 7 % pri zimskoj sje~i i 9–10 % pri ljetnoj sje~i drva, dok je za celulozno drvo iznosila 3 % pri zimskoj sje~i i 7–8 % pri ljetnoj sje~i drva. Najni`a je stopa odbacivanja za pilanske i furnirske trupce pri potpuno mehaniziranom sustavu pridobivanja drva i sortimentnoj metodi te iznosi 3 %, dok je za celulozno drvo najni`a stopa odbacivanja pri djelomi~no mehaniziranom sustavu pridobivanja drva i sortimentnoj metodi te iznosi 2 % (tablica 4). Gubitci obujma i vrijednosti obloga drva za istra`ivana poduze}a s obzirom na sustave pridobivanja drva prikazani su u tablici 8.

Analiza dobivenih rezultata pokazuje da sortimentna metoda pridobivanja drva osigurava najvi{u kakvo}u oblovine u svim ispitivanim poduze}ima pri razli~itim vrstama drve}a. Stablovna metoda pridobivanja drva upu}uje na zadovoljavaju}u kakvo}u oblovine odnosno stopu odbacivanja trupaca, dok se deblovna metoda pokazala kao najmanje zadovoljavaju}a metoda, posebice pri ljetnoj sje~i drva. Mehani~ka o{te}enja drva (iskidana drvna vlakna, o{te}enja debla, rezovi drva hvatalima), gre{ke pri izradi drva (djelomi~no okresane grane, raspucano ~elo i pukotine na drvu) te zabla}enost drva bila su naj~e{}a o{te}enja na oblovini.

Zbog privla~enja oblovine skiderom za vrijeme male nosivosti tla i velike vlage u tlu deblovna metoda s djelomi~no mehaniziranim sustavom pridobivanja drva ima najvi{u stopu odbacivanja trupaca {to zbog zabla}enja, {to zbog drugih gre{aka. Pri stablovnoj metodi s djelomi~no mehaniziranim sustavom pridobivanja drva isti je razlog visoke stope odbacivanja, me|utim grane na deblu ipak pru`aju kakvu-takvu za{titu debla prilikom privla~enja. Sortimentna se metoda, bez obzira na vrijeme sje~e, pokazala kao najbolja jer se drvo izvozi na

Croat. j. for. eng. 31(2010)2

125

Y. Gerasimov and A. Seliverstov

Industrial Round-Wood Losses Associated with Harvesting Systems ... (111–126)

forvarderu i nema izravni doticaj s tlom. Odabir odgovaraju}ega vozila za privla~enje drva treba se zasnivati na terenskim uvjetima, odnosno na trenuta~nom stanju tla.

Da bi se smanjili gubitci obloga drva pri raznim sustavima pridobivanja drva, potrebno je obratiti vi{e pozornosti na vrijednost i kakvo}u drva, a ne samo na koli~inu odnosno drvni obujam. Stoga je potrebno razviti platni sustav te upute za sje~u drva, {to }e omogu}iti bolju iskoristivost drva te izbje}i o{te}ivanja vrijednih sortimenata. Tako|er treba voditi ra~una o vremenu sje~e (vi{a stopa odbacivanja za MM CTL pri zimskoj sje~i, te za MM TL i MM FT pri ljetnoj sje~i). Da bi se izbjeglo dodatno zabla}ivanje drva prilikom uhrpavanja na pomo}nom stovari{tu, potrebno je koristiti se potpornim trupcima. Radnici trebaju redovito odr`avati vozila (pode{avanja harvesterske i procesorske sje~ne glave, o{trenje no`eva, ~i{}enje valjaka i dr.). Vozila trebaju odgovarati sastojinskim uvjetima te svakako treba voditi ra~una o tome da su radnici osposobljeni za rad na pojedinim vozilima odnosno za vrhunske na~ine pridobivanja drva.

Gubitci obloga drva u smislu gubitaka nov~ane vrijednosti po jedinici obujma u pojedinim poduze}ima mo`da se ne ~ine zna~ajnim, posebice ako se usporede gubitci pri sortimetnoj metodi potpuno i djelomi~no mehaniziranoga sustava pridobivanja drva. Ipak, primjena sortimentne metode nasuprot tradicionalnoj deblovnoj metodi donijet }e u{tedu od 0,8 /m3 odnosno oko 100 000 /god. prosje~nomu poduze}u. Po~etna ulaganja od nekoliko stotina tisu}a eura za prelazak iz MM TL u FM CTL isplativa su za du`e razdoblje, dok je prelazak iz MM TL u MM CTL isplativiji u kra}em razdoblju. Potrebna su daljnja istra`ivanja koja }e tako|er uklju~iti proizvodnost sustava, tro{kove, ergonomske uvjete i za{titu na radu, te svakako sastojinsku i okoli{ni pogodnost.

Key words: gubitak obujma, gubitak vrijednosti, pilanski trupci, celulozno drvo, bor, smreka, breza, sortimentna, deblovna i stablovna metoda izradbe drva

Authors’ address – Adresa autorâ: Yuri Gerasimov, PhD. e-mail: yuri.gerasimov@metla.fi Finnish Forest Research Institute Joensuu Research Centre Box 68 FIN-80101 Joensuu FINLAND

Received (Primljeno): August 12, 2010 Accepted (Prihva}eno): November 22, 2010

126

Alexander Seliverstov, PhD. e-mail: alexander@psu.karelia.ru Petrozavodsk State University Forest Engineering Faculty A. Nevsky av., 58 185030, Petrozavodsk Republic of Karelia RUSSIA Croat. j. for. eng. 31(2010)2

Original scientific paper – Izvorni znanstveni rad

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails: Towards a Complex Approach Petr Hruza, Ilja Vyskot Abstract – Nacrtak Marked trails in forests usually make use of forest roads. Nowadays, the selection of a suitable forest road is predominantly, if not solely, based on the technical aspects of the road: the running surface and the available facilities enhancing their suitability for outdoor activities.

However, the quality of the surrounding forest stands and their suitability for recreation – their social-recreation potential are not considered. We hypothesize that it is possible and desirable to evaluate forest roads using another criterion – the potential of the social-recreation function of forest roads according to the social-recreation suitably of surrounding forest stands. Degrees of the social-recreation value of a forest road were calculated as the weighted average of the degrees of values of individual surrounding forest stands and values of the significance/weight according to the length of individual forest stands along the road. The data were processed graphically using the ArcGIS geographic information system, in which it is possible to carry out the presented procedure. Forest roads with the highest degree of social-recreation potential were chosen to plan a trail and the direction of the trail was marked in the maps. The results demonstrate a new possible complex approach to trail designing; it represents an interactive way to choose suitable forest roads, with respect to both their technical quality and also the highest achievable recreational effect of the adjacent forest stands.

This brings us closer to the achievement of a complex approach to trail designing: to be able to evaluate not only the technical aspects of roads but other criteria as well and use this ability to achieve the highest possible recreational effect of forest roads. If the interest of the community in the recreation in a particular forest area increases, we have provided the bases for the planning of new forest trails and roads, suitable surfaces, resting and parking places in purposeful locations.

Keywords: forest road, trail designing, recreation

1. Introduction – Uvod As the society and technology are evolving, people are migrating to larger settlements and their requirements concerning the way they spend their leisure time are changing. Especially outdoor activities are gaining in popularity and these are mostly performed in forests. Forests, as the information on the status of forests in the Czech Republic tell us (MZE, 2010), represent 33.7% of its entire area and they form an opposite of urban areas. In the Czech Republic, based on the Act on Forests (1995), everybody is entitled to enter forests and move there freely, pick forest fruits and wood residuals for their own disposal. However, the citizens are not allowed to move Croat. j. for. eng. 31(2010)2

outside forest roads and marked tracks on bicycles, skis or sledges. Consequently, the increasing sporting population has to put up with forest roads only.

For that reason, the selection of locations suitable for these activities and the associated quality of environment, in our case forest ecosystems, remains an issue. Poto~nik (1996) remarked that besides its traditional productive task, forest roads should also perform several non-productive tasks, as a consequence of the modern way of life: escape from industrialisation, noise, ect. to unspoiled nature. Also Poto~nik et al. (2005b) mentioned that in conditions of public free and unlimited access to forests, the standard of forest road maintenance is higher. Along with this they con-

127

P. Hruza and I. Vyskot

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127–135)

sider that unlimited access to forests means higher personal standard and higher quality of modern life, and that hence the higher cost of wood extraction is not the most important condition of multi-purpose and natural forest management.

As early as in 1994, Keller (1994) pointed out that as the interest of the public is rising, it is necessary to quantify forest functions for the purposes of Swiss forest policy. He defined other forest functions, but he referred to them as non-productive. He also presented possible geographical information systems as suitable instruments for a graphical representation of forest functions and space analysis. Authors from the Finnish Forest Research Institute (Engelmark 1998) remarked that the forest ecosystem cannot be any longer looked upon as a source of wood material only; it is vital that the approach of forestry changed to multiple-use forestry. They emphasize that this kind of research is certainly needed as the economic role of timber, pulp and paper production is currently changing, and perhaps its relative importance is decreasing in relation to other demands, such as tourism and environmental concerns. Other authors (Brändli and Ulmer 2001) discuss the opposing demands for the productive use and the recreational use of the forest ecosystem. They point out the increase in popularity of active outdoor recreation, including trekking, mountain biking, and also climbing and mountaineering. They suggest that the solution is to control and limit recreational activities to certain locations. This would reduce the risk of the recreational forest function conflicting with its other functions. Also Bell (2005) emphasizes the increasing interest in good quality natural environment. He notes that with the falling price of wood it is possible to make use of this interest in and popularity of forests and according to him the value of the recreational forest function should be expressed financially. He discerns the high efficiency of investments – with relatively low expenses for technical facilities, the value of the forest rises considerably. The Danish authors (Larsen et al. 2008) point out that the recreational value of forests does not depend on technical facilities solely but that it is also in direct relation to biodiversity.

A new approach to forestry was introduced by Vyskot (Vyskot et al. 2003). He accepts the philosophy of equal importance of all forest functions for the community, not only wood production but also what he calls the social-recreation function. However, the equal importance of forest functions does not mean that the forest stands are able to fulfil them all to the same degree. The ability of forests to perform functions is highly differentiated and can be expressed by the degree of their real potentials. The

128

acceptance of the multifunctional nature of forests inevitably leads to the necessity of a complex approach to the evaluation of stands. The degree of the real potential of the forest social-recreation function can be used to decide upon the recreational use of a forest. So far, the decision-making process concerning the spending of leisure time has been based on the quality of services and technical recreation – related qualities. However, it is necessary to use the possibilities naturally provided by the forest ecosystem.

Therefore, the objective of this study is to evaluate forest roads according to the values of adjacent stands with regard to their social-recreation function. The hypothesis is that such evaluation is possible, plausible and will bring benefits to the field of planning of marked trails, making such planning highly interactive, flexible and purposeful.

2. Materials and methods – Materijal i metode The basis of the study is the evaluation of forest roads with the criterion of social-recreation function of the surrounding forest stands. For the evaluation of the surrounding forest stands we used the method acknowledged by the Ministry of Environment of the Czech Republic (Vyskot et al. 2003); it defines the social-recreation function as the ability of forest ecosystems to produce effects providing human and social satisfaction of physical and psychological needs (the optimization of organism’s physiological processes).

The real potentials are then used to classify the scale of suitability for recreation and to quantify the effects of all systemized ecosystems in Czech forests.

The suitability of stands for recreation is evaluated by direct and indirect ecosystem criteria. The direct criteria category consists of two criteria: the real species composition expressed through a stand type, and the forest type, which is based on data on precipitation, temperature, soil type and terrain type.

When the evaluation of one location is conducted, as in our case, it is the direct criteria which are of the more determining nature as the indirect criteria for individual stands within one location can be very similar (the same geographical features).

The indirect criteria are numerous and varied, and they are based on climatic, hydrological, terrestrial, geopedological, and physiological and biological data, which can be acquired from database sources.

They include air temperature in the growing season (affecting the overall resulting impression of the stand); physiological climatic optimum (the interaction of air pressure, temperature, wind); the number of summer days (based on temperature); the number Croat. j. for. eng. 31(2010)2

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127–135)

P. Hruza and I. Vyskot

Table 1 Colour coding of individual degrees of real potentials of social-recreation forest function according to Vyskot (Vyskot et al. 2003) Tablica 1. Mogu}nost kori{tenja {umskih sastojina u dru{tveno-rekreacijske svrhe Functional interval of the degree, 0–100% Kakvo}ni udio ~imbenika, 0–100 %

Real potential of the function Prikladnost {umskih sastojina za rekreaciju

Degree – Razina

<10

Unsuitable – Neprikladna

0

11–30

Very low – Vrlo niska

1

31–45

Low – Niska

2

46–55

Average – Prosje~na

3

56–70

High – Visoka

4

71–90

Very high – Vrlo visoka

5

>91

Outstanding – Izvanredna

6

of rainy days (based on precipitation); the number of snow days (based on snow cover); the duration of sunshine (number of hours of sunshine); the altitude (diversity of terrain), the terrain accessibility (slopes, surfaces, obstacles); the soil depth (as a result affecting the variability of the environment); the site bearing capacity (the capacity of a site to bear the load of recreating population); »physiological« biodiversity of trees; herb layer – species diversity; and herb layer – cover. The degrees are expressed in direct proportion with the available data, with the exception of the number of rainy days where the proportion is inverse. As a result all the criteria affect human feelings, impressions and the sense of well-being while staying in the actual stand.

In the study, the real potentials of the social-recreation function of forest stands adjacent to forest roads in the example area were evaluated and individual degrees were marked with a colour so that they could be visualised in stand maps (Table 1).

The vector layer representing the boundaries between forest stands – the stand map created during the preparations of Regional Plans for Forest Development (ÚHÚL 2001) – was used as the basis for visualising the degrees of the social-recreation forest function. The resulting map of the social-recreation function potentials was then used for the degree evaluation of the road network in the area. Subsequently, the vector layer of forest hauling roads was projected onto this map.

The resulting degrees of the social-recreation potentials of the forest stands were then used for the degree evaluation of the road network in the area.

The forest roads were evaluated in sections between crossings with other roads; thanks to this, the direction of the designed trail can be changed on demand.

Degrees of the social-recreation value of a forest road were calculated as the weighted average of the Croat. j. for. eng. 31(2010)2

Colour code Boja prikaza

degrees of the values of individual surrounding forest stands and the values of the significance/weight according to the length of individual forest stands along the road, according to the following formula: n

∑w x=

i=1

i

× xi (1)

n

∑w i=1

i

Where: wi are the values of degrees of individual surrounding forest stands xi are the values of significance/weight, which means the length of individual forest stands along the road.

This means that the degrees of potentials were related to the length of the boundary between a forest stand and the forest road. The weighted average was calculated for both sides as forest roads often form boundaries between stands.

Fig. 1 Area of study Slika 1. Podru~je istra`ivanja 129

P. Hruza and I. Vyskot

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127â&#x20AC;&#x201C;135)

Fig. 2 Graphical visualisation of forest road evaluation Slika 2. Razredba terena prema kakvo}i {umskih cesta za razonodu ljudi 130

Croat. j. for. eng. 31(2010)2

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127–135)

Then, the sections of forest roads with the highest degree of the social-recreation potential were chosen to plan the trail.

The data were processed graphically using the ArcGIS geographic information system, in which it is possible to carry out the presented procedure.

3. Area of study – Podru~je istra`ivanja The area used for the proposal for interactive planning of trails is the floodplain forest of the @idlochovice Forest Enterprise of the Forests of the Czech Republic, state enterprise (Fig. 1).

The @idlochovice Forest Enterprise is located in the southern part of the South Moravian region and its floodplains have been preserved in the southern part of the Thaya-Svratka Dell (Dyjsko-svratecky úval), which stretches along the Morava and the Thaya Rivers. Long, quite straight, broad, and shallow valleys are typical for the area. The current level of the area accessibility is very good due to the flatland terrain, with the average density of forest hauling roads being 16 m/ha, their surface using mostly bituminous binder (ÚHÚL 2001). The forest road network in the area in question was built gradually throughout the history and it has been updated when needed. It means that besides historic roads, forest roads built in the second half of the last century are used. They are mostly classified as the highest class of forest roads. A part of the area is located in the Lednice-Valtice Cultural Landscape, which has been an UNESCO site since 1996. All in all, this environment is suitable for recreation and free-time activities of the South Moravian region inhabitants.

4. Results – Rezultati The hypothesis that it is possible to evaluate forest roads from other perspectives than technical has been confirmed. The degrees of the social-recreation forest roads evaluation were calculated. The graphical visualisation of the evaluation is presented in the geographic information system ArcGIS in Fig. 2.

The evaluation of forest road sections was carried out for forest hauling roads, as these are most often used as trails for hiking, cycling or cross country skiing. The degrees of forest road sections were added to the database of forest hauling roads as an independent field and visualised graphically. For the actual design of the trail, forest roads with the highest degree of the recreational function were chosen and the direction of the trail was marked in the maps (Fig. 3).

The presented solution allows for a qualified interactive designing of trails, making use of the forest Croat. j. for. eng. 31(2010)2

P. Hruza and I. Vyskot

roads, which are most suitable and of the highest quality for recreation; moreover, it makes the planning of accompanying recreational facilities possible and purposeful.

The forest trail was designed so as to lead through the section of forest roads with the highest social-recreation degree (Fig. 4).

The results demonstrate a possible complex approach to trail designing; it represents an interactive way to choose suitable forest roads, with respect to both their technical quality and also the highest achievable recreational effect of the adjacent forest stands. We have provided the foundations for planning new forest hauling roads, suitable surfaces, and resting and parking places in purposeful locations, which can be easily applied, if the interest of the community in the recreation in this particular forest area increases. This is the practical implication and the main benefit of the forest road evaluation from the social-recreation function perspective.

5. Discussion – Rasprava The development of outdoor activities is becoming a priority for some areas, and planning and building of facilities for hikers, bikers and others has to be based on a concept. Gül et al. (2006) point out that it is necessary to create a recreation management plan for these areas. The authors also mention the need to control the movement of tourists (hikers, bikers, etc.) and to build some facilities for them in designated places (including picnicking, sports facilities and playgrounds, camping sites, walking paths, food and local outlets). However, this requires a qualified selection of trails and locations where the facilities should be built. In his paper Poto~nik (2006) deals with the management strategy in a preserved forest area. The strategy proposed the regulation of parking lots and introduction of alternative public transport, fees, providing wide tourist plateau and providing visitors with information about the natural environment, protected natural and forest areas, the national park itself, code of behaviour in the natural environment.

The methodology presented is a useful tool for the above mentioned selection of new facilities. It allows maximum utilization of the social-recreation function of forest stands and efficient investments in the promotion of recreation in the particular area. In this way, it is possible to mark the trails so that the hikers (or bikers, etc.) use the forest roads with the highest degree of the social-recreation function and also make sure that the trails have the necessary facilities. This does not only mean minor recreational constructions, such as resting places, but also an overall technical design of a forest road. The design of the surface

131

P. Hruza and I. Vyskot

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127â&#x20AC;&#x201C;135)

Fig. 3 Degree evaluation of individual forest road sections Slika 3. Stupanj vrijednosti pojedinih dijelova {umskih cesta 132

Croat. j. for. eng. 31(2010)2

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127â&#x20AC;&#x201C;135)

P. Hruza and I. Vyskot

Fig. 4 The designed recreation trail Slika 4. Projektirana rekreacijska staza Croat. j. for. eng. 31(2010)2

133

P. Hruza and I. Vyskot

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127–135)

cross-section affects the road appearance and its aesthetics. This concerns mainly the length of fill and cut slopes and the type of surface (Akay et al. 2007).

Other authors (Lugo and Gucinski 2000), who study relations between forest roads and forest ecosystems, confirm their close relationships; they use the terms road ecology and technoecosystem. A suitable technical design of roads on selected trails enhances the attractiveness for recreation provided that the recreation carrying capacity of the forest ecosystem is not exceeded. Daniels and Marion (2006) state that the management of backcountry recreation areas involves a balance of actions taken to meet environmental and social objectives. In their study, a group of European scientists (Bartczak et al. 2008) prove that e.g. the willingness of people to pay for a stay in the forest is continuously increasing. Authors Poto~nik et al. (2005a) present the level of over-standard, which means the level which is higher than the minimum requirements for normal forestry service of a forest road. Support should be provided from the state, local communities and social groups interested in a particular forest road. Forestry itself could not support higher standards without extra financial support.

Also for this reason, it is recommendable to approach the social-recreation function of forest roads competently and to harmonize it with the economic needs of wood production. The methodology presented provides a solution for outdoor recreation in the forest environment with the highest degree of suitability for recreation.

6. Conclusions – Zaklju~ci The evaluation of forest roads according to the functional criteria of the social-recreation function of forest stands for trail designing in the forest environment brings qualitative changes to the issue of multifunctional use of forest roads.

The functional criteria enable us to harmonize the recreational and technical aspects of forest roads used for trails and to achieve the highest social-recreation effect. The harmonization of various aspects found in forest evaluation is the complex approach we are aiming at. The technical-economic approach changes into the technical-economic-environmental approach, which contributes to the improvement of the quality of leisure time spent in forests.

7. References – Literatura Akay, A. E., Pak, M., Yenilmez, N., Demirbag, H., 2007: Aesthetic Evaluations of Forest Road templates, International Journal of Natural and Engineering Sciences 1(3): 65–68.

134

Bartczak, A., Lindhjem, H., Navrud, S., Zandersen, M., Zylicz, T., 2008: Valuing forest recreation on the national level in a transition economy: The case of Poland. Elsevier.

Forest Policy and Economics 10 (7-8): 467–472.

Bell, S., 2005: Forest recreation: New opportunities and challenges for forest managers. <http://www.sumins.hr: 8080/2007-izv.10/16_bell_engl.pdf> (Accessed 20 June 2009).

Brändli, U. B., Ulmer, U., 2001: Recreational Function. <http://www.wsl.ch/staff/urs-beat.braendli/Forest_Recreation_Function.pdf> (Accessed 30 June 2009).

Daniels, M. L., Marion, J. L., 2006: Visitor Evaluations of Management Action at a Highly Impacted Appalachian Trail Camping Area. Springer. Environmental Management 38(6): 1006–1019.

Engelmark, O., 1998: Multiple-use forestry in the Nordic countries. Elsevier. Forest Ecology and Management 102: 339–340.

Gül, A., Örücü, M. K., Karaca, Ö., 2006: An Approach for recreation Suitability Analysis to Recreation Planning In Gölcük Nature Park. Springer. Environmental Management 37(5): 606–625.

Keller, M., 1994: Consideration to quantify forest functions by means of a GIS. <http://libraries.maine.edu/Spatial/gisweb/spatdb/egis/eg94190.html> (Accessed 22 June 2009).

Larsen, F. W., Petersen, A. H., Strange, N., Lund, M. P., Rahbek, C., 2008: A Quantitative Analysis of Biodiversity and the Recreational Value of Potential National Parks in Denmark. Springer. Environmental Management 41(5): 685–695.

MZE, 2010: <http://eagri.cz/public/web/mze/lesy/ ?fullArticle=1 > (Accessed 11 November 2010).

Poto~nik, I., 1996: The multiple use of the forest roads – relative importance of the particular. In. Kosir, Bo{tjan, (ed.).

Izzivi gozdne tehnike: zbornik posvetovanja (proceedings). Ljubljana: Gozdarski in{titut Slovenije: 95–103.

Poto~nik, I., Pentek, T., Pi~man, D., 2005a: Impact of traffic characteristic on forest roads due to forest management.

Croatian Journal of Forest Engineering 26(1): 51-57.

Poto~nik, I., Yoshioka, T., Miyamoto, Y., Igarashi, H., Sakai, H., 2005b: Maintenance of forest road network by natural forest management in Tokyo University Forest in Hokkaido. Croatian Journal of Forest Engineering 26(2): 71–78.

Poto~nik, I., 2006: Road Traffic in Protected Forest Areas – Case Study in Triglav National Park, Slovenia. Croatian Journal of Forest Engineering 27(2): 115–121.

Lugo, A. E., Gucinski, H., 2000: Function, effects, and management of forest roads. Elsevier. Forest Ecology and Management 133(3): 249–262. ÚHÚL, 2001: Oblastní plán rozvoje lesa – Pøírodní lesní oblast ~. 35 (Regional Plan of Forest Development – Natural forest area no. 35). Ústav pro hospodáøskou úpravu lesa (The Forest Management Institute): 149. Brandýs nad Labem 2001.

Croat. j. for. eng. 31(2010)2

Social-Recreation Evaluation of Forest Roads and their Suitability for Trails ... (127–135)

Vyskot, I., et al., 2003: Quantification and Evaluation of Forest Functions on the Example of the Czech Republic.

Ministry of Environment of the Czech Republic, p. 194.

Prague 2003.

P. Hruza and I. Vyskot

Zákon o lesích, 1995: <http://www.uhul.cz/legislativa/ 289.php> (Accessed 12 November 2010).

Sa`etak

Odre|ivanje dru{tveno-rekreativne kakvo}e {umskih prometnica Danas se odgovaraju}a {umska cesta ili put za rekreaciju ili boravak u prirodi prete`no, ako ne i isklju~ivo odabire na osnovi tehni~kih svojstava same ceste, odnosno njezine {irine (povr{ina za slobodno kretanje stanovni{tva) te na osnovi dodatnih objekata koji su na raspolaganju u {umi. Me|utim, kakvo}a okolnih {umskih sastojina i njihova prikladnost za razonodu, tj. njihov dru{tveno-rekreativni potencijal, ne uzimaju se u obzir. Stoga je svrha ovoga rada bila da se ocijene {umske ceste na temelju vrijednosti okolnih sastojina s obzirom na njihove dru{tveno-rekreativne mogu}nosti. Hipoteza je da }e takva procjena na uvjerljiv na~in pridonijeti pobolj{anju planiranja budu}ih planinarskih staza i rekreacijskih putova.

Za ocjenu okolnih {umskih sastojina primijenjena je metoda koju priznaje Ministarstvo za{tite okoli{a Republike ^e{ke (Vyskot i dr. 2003). Ona odre|uje dru{tveno-rekreativne funkcije {ume kao sposobnost {umskih ekosustava za stvaranje pozitivnih dru{tvenih u~inaka koji kod ljudi ispunjavaju odre|ene fiziolo{ke i psiholo{ke potrebe. Izra|ena je podjela {umskih ekosustava na temelju njihove pogodnosti za rekreaciju stanovni{tva te je potom analizirana mre`a {umskih cesta po pojedinim podru~jima. [umske su ceste ocjenjivane na temelju koli~ine me|usobnoga kri`anja, {to stanovni{tvu omogu}uje promjenu smjera kretanja po {umi. Podaci su obra|eni pomo}u ra~unalnoga programa ArcGIS. Podru~ja s najvi{om ocjenom izabrana su za planiranje rekreacijskih putova i staza ~iji je smjer ozna~en na karti. Predstavljeno je rje{enje interaktivnoga projektiranja rekreacijskih staza (ali i mogu}ih prate}ih turisti~kih objekata) kori{tenjem postoje}e mre`e {umskih cesta koje su unaprijed odre|ene najprikladnijama za razonodu ljudi.

Hipoteza da je mogu}e procijeniti {umske prometnice iz drugih, netehni~kih gledi{ta potvr|ena je. Rezultati pokazuju mogu} nov pristup pri projektiranju planinarskih staza odabirom prikladnih {umskih cesta s obzirom na njihova tehni~ka svojstva i {umske sastojine koje one otvaraju. Osigurali smo temelje za planiranje novih rekreacijskih staza uzimaju}i u obzir postoje}u mre`u {umskih prometnica, povr{inu prikladnu za odmor te parkirna mjesta na svrhovitim polo`ajima unutar {ume. Ako se interes zajednice pove}a, lako se mo`e ostvariti. Vrednovanje {umskih cesta prema dru{tveno-rekreativnoj djelatnosti donosi pokazatelj vi{ekriterijskoga kori{tenja {umskih cesta koje je odavno poznato, ali tek treba biti priznato u praksi.

Klju~ne rije~i: {umska cesta, projektiranje rekreacijskih staza, rekreacija stanovni{tva u {umi

Authors’ address – Adresa autorâ:

Received (Primljeno): April 14, 2010 Accepted (Prihva}eno): October 11, 2010 Croat. j. for. eng. 31(2010)2

Petr Hruza, PhD. e-mail: petrhr@mendelu.cz Prof. Ilja Vyskot, PhD. e-mail: ilja.vyskot@mendelu.cz Mendel University Faculty of Forestry and Wood Technology Department of Landscape Formation and Protection Lesnicka 3 613 00 Brno CZECH REPUBLIC

135

Original scientific paper – Izvorni znanstveni rad

Noise Pollution in Forest Environment Due to Forest Operations Igor Poto~nik, Anton Poje Abstract – Nacrtak Noise is a disturbing and unpleasant sound and refers to subjective definition of sound. A sound can have a series of different physical features. However, it becomes noise when it has negative physiological or psychological impact on a human being, e.g. causes health impairments and behavioral disorders. In the animal kingdom the high levels of noise may interrupt natural cycles, such as animal eating habits, coupling, and migration paths, or even cause the extinction of animal species living in noise polluted environment. Undoubtedly, modern forest operations cause noise in the forest. The goal of this research is to study the level of noise pollution as well as stand and terrain conditions influencing noise spreading in forest environment. It was established that the total chain saw noise power equals the wind noise at the distance of 140 m, whereas the sound levels up with that of forest silence at 252 m. The chain saw noise is similar to background noise at distances of 60–80 m and frequencies below 80 Hz and above 12.5 kHz. Consequently, this means lesser impact on natural environment in these frequency bands. The hypothesis was not confirmed, i.e. that vertically screened forest attenuates noise spreading more successfully than vertically nonscreened forest: the difference emerges due to sound reflections in vertically screened forest, causing less sound absorption. However, the differences were confirmed at the distance of 80 m regarding noise attenuation in different seasons: winter – summer (difference of 11.92 dB), spring – summer (difference of 6.89 dB), and insignificant between winter and spring.

Key words: noise attenuation, noise frequency, forest operations, noise pollution, forest

1. Introduction – Uvod Be it a tone, sound, rustle, or bang, definitions of noise, implying that a feeling of noise is related to human hearing impression and mood, are quite similar and consistent. Noise is regarded as an unpleasant sound and relates to subjective definition of sound.

A sound can have a series of different physical features. However, it is regarded a noise only when it has negative physiological or psychological impact on a human being. Biological definition of a noise is: every sound that disturbs a human being, causes agitation, interrupts work, and harms health and wellbeing.

Protection against noise in urban areas of modern societies is becoming more and more important, since the number of urban as well as rural population that feels endangered by noise is increasing. It was established that almost 25% of European population is exposed to noise above 65 dB(A) caused by traffic (Berlund and Lindvall 1995).

Croat. j. for. eng. 31(2010)2

The excessive noise can cause health injuries and behavioral disturbances. The unpleasant and unwanted noise can cause a feeling of annoyance, aggressiveness, hypertension, stress, gradual hearing loss, and other injuries depending on exposure and noise level (Berlund and Lindvall 1995). Þ Hearing impairment – it has been proved that exposure to noise causes injuries of inner ear that can lead to hearing loss. Studies in the USA showed that the majority of people have impaired hearing in old age due to the exposure to loud sound, and not as a sole consequence of old age. Þ Cardiovascular impairment – exposure to sound above 70 dB at ordinary work post for eight hours a day causes the increase of blood pressure, leading to artery injuries and stress.

High noise levels can interrupt natural cycles of animals, such as eating habits, coupling, and migration paths. Exposure to noise can cause the extinc-

137

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

tion of animal species in noise polluted environment.

Kaseloo (2006) sums up the findings that grassland birds are sensitive to traffic noise, which is reflected in decreasing population in noise polluted environment: 7 out of 12 grassland bird species moved their nests 20 to 1700 m from the road with traffic load of 5000 vehicles a day, whereas in areas with 50 000 vehicles a day, birds withdrew their nests 65 to 3535 m away from the road. 26 of 43 studied forest birds (60%) reacted to noise by decreasing their population density at the distance of 50 to 1500 m (traffic load of 10 000 vehicles a day) and 70 to 2800 m (traffic load of 60 000 vehicles a day).

English recommendations (Anon. 1995) deal with buffer zones of badger setts according to performance of forest operations. They suggest 20 m zone around the sett entrance and avoiding the use of heavy mechanization due to relatively shallow tunnels (60 cm). On the other hand, they advise at least 100 m distance of construction site from setts in building forest infrastructure (drilling, hard surface mining).

The study conducted by (Fang and Ling 2003) summarized experimental data in a single map, incorporating the relationships between relative attenuation and both visibility and width. This study provides data of use to environment designers. For example, designers can reduce noise by 6 dB(A) via suitable plantings. Also, belts of trees and shrubs could be planted based on 1 m visibility and 5 m width, or 10 m visibility and 18 m width.

The chain saw noise is one of the most important noise sources, depending on the type of motor oil used for chain lubrication (Wojtkowiak et al. 2007).

Measurement results showed that noise levels observed were high and varied with oils used, ranging from 99.6 dB(A) for a vegetable oil to 105.2 dB(A) for a mineral oil.

A chain saw is regarded problematic even when compared to helicopter noise (Delaney et al. 1999) based on the example of reaction of owls. Spotted owls did not flush when the noise level of helicopters was less than 92 dB(A) and the level of chain saws was less than 46 dB(A). Chain saws were more disturbing to spotted owls than helicopter flights at comparable distances. Results indicate that a 105 m buffer zone for helicopter overflights would minimize spotted owl flush response and any potential effects on nesting activity.

On the other hand, Tempel and Gutiérrez (2003) tested the physiological response of 9 non-breeding wild male owls to the sound of a chainsaw operating 100 m from their roost site. The chain saw exposure did not result in a detectable increase of physiologi-

138

cal response, which suggests that spotted owls can tolerate low-intensity human sound in their environment without eliciting a physiological stress response.

Pal (2000) studied the effect of green belt on coalmine noise attenuation. Eight plantation sites in Jharia (JCF) and Raniganj (RCF) coal fields in India were investigated. The maximum total noise attenuation for ¸eq at 50 m depth of the green belt was found to be within 18.8 to 21.1 dB(A) in JCF and 18.7 to 21.0 dB(A) in RCF. Excess noise attenuation (¸eq) exclusively due to green belts in JCF and RCF was 3.3 to 6.0 and 3.6 to 5.7 dB(A), respectively. Excess attenuation for higher frequencies (> 250 Hz) was more (> 4 dB(A)) than that for lower frequencies (£ 125 Hz).

2. Objectives – Ciljevi On the basis of previous studies and research problems observed in-situ, we formed two primary objectives of the study, i.e. research questions: Þ What is the level of noise pollution in forest environment due to forest operations, Þ What is the influence of natural and stand factors on noise spreading in the environment.

When constructing scientific hypothesis, we assumed that vertically screened stands and winter time, due to sound absorption of snow cover, have greater influence on noise attenuation.

3. Methods – Metode The research was conducted in natural fir-beech (Omphalodo-Fagetum) forests of southern Slovenia with prevailing features of high karst and AMSL ranging from 1030 to 1220 m. The wider area of research belongs to the protected area (SPA) of Natura 2000 as special areas of conservation (SAC). In this research dealing with noise spreading due to forest operations, especially due to cutting, we used an approximately 1 year old professional chain saw STIHL MS 460 as the noise source. The chain saw, which was regularly maintained, represents a common type of working means for cutting in the above described conditions. To acquire data about maximum noise during work operations, we conducted preliminary measurements of noise for all cutting operation stages. On the basis of 2 hour long measurements and seven trees cut down, we established that the chain saw makes the loudest noise during notch cutting and cross-cutting operations. Due to easier work execution, we conducted the experiment by cross-cutting beech trunks of approximately 30 cm in diameter. The selection of working opCroat. j. for. eng. 31(2010)2

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

I. Poto~nik and A. Poje

Table 1 Experiment design (Number of measurements (repetitions) by distance from the noise source, season, relief and noise screening) Tablica 1. Postavke pokusa (broj mjerenja /ponavljanja/ u odnosu na izvor buke, godi{nje doba, odmor i zaklon od buke) Season Distance from the noise source, m Godi{nje doba Udaljenost od izvora buke, m

Summer Ljeto

Spring Prolje}e

Winter Zima

5 10 20 40 60 75 80 5 10 20 40 80 5 10 20 40 80

Stands without regeneration Sastojine bez pomlatka Uphill Downhill Flat Uzbrdo Nizbrdo Ravnica 3 3 6 3 3 6 3 3 6 3 3 6 – 2 – 3 – – – 1 6 – – 6 – – 6 – – 6 – – 6 – – 6 – – 6 – – 6 – – 6 – – 6 – – 6

eration was thus partly in accordance with standard measurement of noise level caused by chain saws at woodcutter’s ear (ISO 7182:1996). Cross-cutting of trunks was always conducted on the ground and perpendicularly to direction of noise recording. The side of cross-cutting was selected randomly (left or right), for it is typical of forest production.

We included three factors according to the purpose of the study: screening of noise source, relief and season. They are regarded to have influence on noise spreading in the forest. The influence of noise source screening was studied at two levels: screened and unscreened. The screening corresponded to the presence of regeneration in a stand, which usually screened the noise source already at the distance of 5 m. The influence of relief was analyzed at three levels: downhill, flat and uphill. It has to be mentioned that the noise source was positioned on the slope and skid road with slope cut. The influence of different ground surface and leaf density on trees was established by partial repetition of measurements in different seasons (Table 1): summer, winter and spring. In the described area we tried to select 12 series of measurement points with 10 measure points per series according to the number of factors and Croat. j. for. eng. 31(2010)2

Stands with regeneration Sastojine s pomlatkom Uphill Downhill Flat Uzbrdo Nizbrdo Ravnica 3 3 4 3 3 4 3 3 4 3 3 4 1 1 – 1 1 – 1 1 4 – – – – – – – – – – – – – – – – – – – – – – – – – – – – – –

their levels. The measurement points were situated at the distances of 5, 10, 20, 40, and 80 meters left and right from the noise source. It is possible that the last two measurement points were closer than intended (60 or 75 meters) due to terrain roughness (great slope cut of forest road, slope crest). Due to natural conditions, we managed to conduct 11 out of 12 planned series of measurements; without 1 series in combination of factors: unscreened-level.

Noise measurements were conducted successively from 5 to 80 m distance from the noise source.

At each measurement point, the measurements lasted for 10 seconds (10 × 1 s). Only the loudest 5 second intervals were included in noise analyses, since for 10 second duration of noise we would have to cross-cut the trunk again. To acquire data about background noise, we also recorded »natural« forest noise in times of silence together with other noise sources (overflights) in different recording days and intervals.

While it is generally recognized that the various components of attenuation may be inter-related and not simply additive, investigations have not proceeded as yet to the extent that it is possible to quantitatively express all of the possible inter-relations in

139

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

one encompassing algorithm, but rather approximated as a linear sum of effects (Hansen 2005).

The measurements were conducted with Sound Level Meter Brüel and Kjaer 2250, by which total equivalent noise level in second intervals was recorded as well as equivalent noise level by 1/3 octaves of frequency spectrum; both with F (Fast) time and Z (Zero) frequency balancing (Equation 1). Measurements without frequency balancing were conducted due to the fact that all existing frequency weights – filters (A – D) adjust sound pressure sensation to human hearing, which definitively differs from other species. The noise measurement ranged in frequency spectrum from 12.5 to 20 kHz. The microphone (Brüel and Kjaer ZC 0032) was set approximately 150 cm above the ground. While conducting measurements, the windscreen-foam ball microphone cover was used (Brüel and Kjaer UA 0237).

T + Dt   LZeq(T) = 20 lg  (1 /Dt) ∫ p z2 ( x) dx / p 0  , dB (1) T  

Where: LZeq equivalent continuous sound level T start time averaging time interval Dt Z zero frequency weighting dummy variable of time integration over x the averaging time interval pZ(x) Z frequency weighted instantaneous sound pressure p0 reference sound pressure (20 mPa).

Simultaneously with noise measurements, the measurements were also conducted of meteorological factors (Metrel MI 6401 Poly euro set), air temperature, relative air humidity, and wind speed in 5 minute intervals that could influence noise measurements. The estimated air pressure at AMSL of 1125 m was 885 hPa.

To process the acquired data, we calculated total equivalent value of sound and equivalent value by 1/3 frequency bands from the loudest 5-second intervals. Since the repetition of measurements (Table 1) was not done for all combinations of factors, we conducted the analysis by individual factors to decrease variability due to other factors. Thus, we eliminated the influence of season from the analysis of influence of terrain and screening, whereas the influence of terrain and screening was eliminated from the season analyses. In other words, we only applied the data recorded in summer to conduct analyses of terrain and screening influences, whereas the analyses of season influence used the data of flat relief and without screening.

To correctly conduct the analysis of covariance, it is necessary to consider preliminary conditions of application among which the most important are independence, normal distribution and homogeneity of variance. The independence of measurements was provided, since every measurement was conducted as a real repetition. The test of normality (Kolmogorov – Smirnov test with Lilliefors significance correction) showed that abnormal data distribution was only present at individual distances of covariance (most frequently at the distance of 40 m), whereas the greatest violations occurred in preliminary condition of variance homogeneity (Levene’s test). The reason for this can be in the fact that the influences of different factors in inhomogenous environment accumulate themselves by the distance from the noise source, consequently causing the increase of data variability. Since we believed the increasing of variability to be natural, we used untransformed data for our analyses.

The selected sound source, i.e. chain saw, does not allow the analyses of differences between combined influences of individual factors, since we assume that the variability of sound source can be greater than the established influences of factors. On the other hand, we can analyze how the influential factors affect the noise spreading. Mathematically,

Table 2 Averaged meteorological data by seasons Tablica 2. Prosje~ni meteorolo{ki podaci po godi{njim dobima Season Godi{nje doba Summer – Ljeto Winter – Zima Spring – Prolje}e

140

Days Dani 2 1 1

Air temperature, °C Temperatura zraka, °C Min. Avg. Max. Min. Srednja Maks.

14.5 18.6 22.6 –2.7 –2.1 –1.6 10.8 12.1 14.9

Relative air humidity, % Relativna vla`nost zraka, % Min. Avg. Max. Min. Srednja Maks.

46.3 59.5 71.0 77.5 80.9 85.4 44.9 53.6 62.4

Min. Min. 0.0 0.0 0.0

Wind speed, m/s Brzina vjetra, m/s Avg. Srednja 0.5 0.6 0.7

Max. Maks.

3.3 3.0 2.2

Croat. j. for. eng. 31(2010)2

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

I. Poto~nik and A. Poje

we can say that we are interested in differences of strait line inclinations and not in their deviation. In practice, we conducted the analysis of covariance (ANCOVA), where, considering the influence of factor, the significant interaction between factor and covariance (distance from the sound source) showed us differences between inclinations of straight lines.

When conducting spectrum analysis, we relativized noise attenuation according to the distance from the source. Thus, we compared noise by all frequency bands at the distances above 10 m with the noise measured at 5 meters or we compared noise between two neighboring points. In this way we eliminated differences between potentially different sound power of the source and also frequency structure of the sound. The spectrum analysis was conducted through the means of frequency bands. The statistical analysis of frequency spectrum was conducted by analysis of variance (ANOVA) or by conducting Tamhane T2 post-hoc test (MD = mean difference). The statistical software SPSS 16.0 was used for all statistical processing.

4. Results and discussions – Rezultati s raspravom The comparison between sound spreading in the forest and geometrical sound spreading (As) shows us that sound power in the forest does not attenuate only due to the distance from the noise source but also on account of other factors (Fig. 1).

By inserting the distance from the source in linear regression model (Equation 2), we found out that noise in the forest averagely attenuates by 8.4 dB (95% CI 8.07 – 8.79), if the distance from the noise source is doubled. This means that atmospheric absorption and other factors contribute additional 2.4 dB of noise attenuation, if noise generally attenuates by 6 dB with doubling the distance from the noise source.

LZeq(dB) = 109.765 – 28.016 ´ log(dist) N = 170, R2 = 0.927, P < 0.001

(2)

Where: LZeq equivalent zero weighted sound level dist distance from the noise source in meters.

The noise of chain saw was compared with the noise of a plane overflying the areas of measurement. By applying the model (Equation 1) and average values of the plane noise (68.9 dB), we found out that the total plane noise power was equal to the chain saw noise at the distance of 28.6 meters. In the same way, we also calculated two distances of balance with noise caused by wind and with »noise« of Croat. j. for. eng. 31(2010)2

Fig. 1 Attenuation of sound by the distance from the noise source according to geometrical decreasing and some other noise sources in the forest Slika 1. Smanjenje razine buke s promjenom udaljenosti od izvora buke prema geometrijskomu smanjenju i nekim dodatnim izvorima buke u {umi forest silence. Thus, the total chain saw noise power at the distance of 140 m equals the wind noise, and at the distance of 252 m it matches forest silence noise.

The results are comparable to recommendations about 105 m buffer zone (Delaney et al. 1999), where the chain saw noise is obviously more disturbing than that of a helicopter. However, the physical presence of human being in nature as a disturbing element was not analyzed.

In fact the equivalence with forest silence noise is never achieved because the silence represents the background noise and as such the limit of influence of the chain saw noise.

On the basis of our findings, we should not assert that the noise source at the calculated distances »disappears« in the background noise or that it becomes undetectable since the applied sources have very different frequency spectra (Fig. 2). Since low frequencies prevail in background and wind noise, whereas chain saw noise is of high frequency, it would be still possible to detect chain saw noise in higher frequency bands despite the equalization of total noise power.

The frequency spectrum of noise changes with increasing distance from the noise source. The most

141

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137â&#x20AC;&#x201C;148)

Fig. 2 Frequency spectrum at 80 meters from the noise source and frequency spectrum of natural noise sources Slika 2. Frekvencijski spektar na 80 metara od izvora buke i frekvencijski spektar prirodnih izvora buke

Fig. 3 Frequency spectrum of chain saw noise depending on the distance from the noise source Slika 3. Frekvencijski spektar buke motorne pile ovisno o udaljenosti od izvora buke prominent differences are recorded for frequencies below 100 Hz where influence of background increases according to the distance from the source, as shown in Fig. 3.

The already mentioned general noise attenuation by the distance (5 m to 80 m) from the source (8.4 dB)

142

is not constant throughout the whole frequency spectrum as it could be concluded from Fig. 3. If we compare noise attenuation at the double distance from the noise source by frequency bands (Fig. 4), we can see that in general the noise up to 80 Hz attenuates less as predicted by geometrical sound spreading; at 200 Hz Croat. j. for. eng. 31(2010)2

Noise Pollution in Forest Environment Due to Forest Operations (137â&#x20AC;&#x201C;148)

I. Poto~nik and A. Poje

Fig. 4 Noise attenuation in relation to distance between two neighboring points of measurement and noise frequency Slika 4. Prigu{enje buke u odnosu na razmak izme|u dviju susjednih to~aka mjerenja i na frekvenciju buke it reaches the local maximum (10.3 dB), up to 2.5 Hz attenuates (8.0 dB), and then intensifies up to the maximum at 16 kHz (12.8 dB). The prominence of both noise extremes increases according to the distance between two points, thus reaching the highest level at the double distance; from 40 to 80 meters. This is expected because the resistance of environment (ground, atmospheric absorption, studied factors) accumulates with the increasing distance between two points. Due to great variability of data by frequencies, in general the noise at 16 kHz is significantly louder than noise with the frequency ranging between 250 Hz and 8 kHz, and below 200 Hz. Both maximums are insignificantly different.

Pal et al.(2000) found out that the noise attenuation due to green belt can be assessed from 3.3 to 6.0 dB(A), depending on actual conditions. This implies that it is fairly difficult to obtain general findings of noise pollution because all results come from case study. Also Crocker (1998) concludes that for sound attenuation through foliage and trees, the main effect at low frequencies is to enhance ground attenuation, the roots making the ground more porous. On the other hand, at high frequencies, where dimensions of leaves become comparable with the wavelength, there is also a significant attenuation by scattering.

Noise attenuation at frequencies above 2.5 kHz and comparisons up to 40 meters is linearly dependent on frequency and it also varies from the location of Croat. j. for. eng. 31(2010)2

comparisons (Fig. 5). Therefore, the noise intensification with frequency is the smallest when doubling the distance from 5 to 10 meters from the noise source, and the highest when doubling it from 20 to 40 meters.

In the same order, the significant relation between noise and frequency also increases (R210/5 = 0.05, R220/10 = 0.19, R240/20 = 0.50; all p < 0.001). The order corresponds to the increased influence of atmospheric absorption (Aa) on noise attenuation, since it increases linearly with the distance from the noise source. Dependence between frequency and noise in compared measurements above 40 meters is best described by parabolas (R260/40 = 0.338, p < 0.001; R275/40 = 0.084, p = 0.128; R280/40 = 0.319, p < 0.001) with apex around 12 kHz, meaning that noise at frequencies above 12 kHz attenuates. The reason for different forms of relation between frequency and noise can be ascribed to the influence of background.

If we sum up the results from Fig. 2 â&#x20AC;&#x201C; 4, we find out that the chain saw noise, measured at the distances from 60 to 80 meters and frequencies below 80 Hz and above 12.5 kHz, approaches the level of background noise, and that background noise prevails over other influential factors. Consequently, a decreased influence on natural environment is shown in these frequency bands. Due to the influence of background, the calculated general noise attenuation in forest (8.4 dB) is slightly underrated.

The influence of stand with regeneration on sound spreading was established on the sample of

143

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

Fig. 5 Intensification of noise in relation to distance between two neighboring points and frequencies above 2500 Hz Slika 5. Pove}anje buke u odnosu na razmak izme|u dviju susjednih to~aka mjerenja i na frekvenciju iznad 2500 Hz summer measurements (Table 1). The analysis of covariance showed that when all measurements are included into analysis, the noise spreading in the forest with regeneration was not significantly different from noise spreading in the forest without regeneration (F (1,106) = 2.192; p = 0.142). If we further limit the sample only to flat relief, the noise spreading in the forest with regeneration becomes significantly different (F (1,46) = 7.751, p = 0.008) compared to the forest without regeneration (Fig. 6). In this case, the noise spreading attenuates by 9.1 dB (CI 95% 9.93 – 8.36) in the forest with regeneration, and by 7.7 dB (CI 95% 8.38 – 7.10) in the forest without regeneration (Equation 3).

tion as a barrier between a noise source and an observer should not be overlooked – in many cases if the noise source is not visible, it is less noticeable and thus less annoying, even if the level is not significantly changed.

The analysis of differences in the influence of revegetation was conducted within the frequency

LZeq(dB) = 111.397 – 1.725ns ´ noreg – – 25.705 ´ noreg ´ log(dist) – – 30.374 ´ reg ´ log(dist) N = 50, R2 = 0.962, P < 0.001

(3)

Where: LZeq equivalent zero weighted sound level dist distance from the noise source in meters reg forest with regeneration noreg forest with no regeneration ns not significant (p = 0.456).

On the other hand, some experiences show that vegetation is not generally considered as an effective (traffic) noise barrier, although it does have an effect in attenuating noise at frequencies above 2 kHz (Hansen 2005). The psychological effect of vegeta-

144

Fig. 6 Noise attenuation according to the distance from the noise source and presence of regeneration in a stand Slika 6. Prigu{enje buke u odnosu na udaljenost od izvora i na prisutnost pomlatka u sastojini Croat. j. for. eng. 31(2010)2

Noise Pollution in Forest Environment Due to Forest Operations (137â&#x20AC;&#x201C;148)

I. Poto~nik and A. Poje

Fig. 7 Noise attenuation by frequency bands according to the distance from noise sources and presence of regeneration in the stand Slika 7. Prigu{enje buke po frekvencijskim pojasima ovisno o udaljenosti od izvora buke i o prisutnosti pomlatka na mjernom mjestu range from 630 Hz to 10 kHz, which corresponds to the greatest assessed tree diameters at chest height and average cross section of beech leaf according to wavelength (66â&#x20AC;&#x201C;3 cm). The comparison of relative noise attenuation by frequency bands shows (Fig. 7) that differences caused by revegetation increase by the distance from the noise source. Thus, the analysis of variance showed insignificant differences in relative noise attenuation in the forest with or without regeneration at 10 meters from the noise source (F(1,138) = 0.788, p = 0.376) and significant differences at the distance of 80 meters (F(1,138) = 77.750, p < 0.001). Assuming that the other conditions are the same, we can ascribe the average difference of 7.96 dB to the influence of regeneration. Therefore, we could not confirm the set up hypothesis that the vertically screened forest attenuates noise spreading better than the vertically unscreened one. We assume that the insignificant difference occurred due to sound reflection in the vertically screened forest and caused lesser sound absorption.

On the sample of summer measurements we also conducted the analysis of relief influence on noise attenuation in the forest. Due to the sunny position of the slope and occasional light wind blowing upward the slope we expected noise attenuation downward the slope to be the highest, and the lowest in the opposite direction. However, the analysis of covariance showed that there are no differences in noise attenuation by distance according to relief (F (2,104) = 1.417, Croat. j. for. eng. 31(2010)2

p = 0.247). Otherwise insignificant greater noise attenuation upward the slope could be ascribed to the influence of skid road slope cut acting as reflective object (Fig. 8).

The analysis of season influence on sound spreading complexly encompasses the changes of meteoro-

Fig. 8 Noise attenuation according to the distance from the noise source and relief Slika 8. Prigu{enje buke ovisno o udaljenosti od izvora i o reljefu 145

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137â&#x20AC;&#x201C;148)

Fig. 9 Noise attenuation according to the distance from the noise source and season Slika 9. Prigu{enje buke u odnosu na udaljenost od izvora buke i na godi{nje doba logical (air temperature, air humidity) as well as stand conditions (leafiness, ground). By comparing summer and spring time, the influence of leafy roof of the stand should be evident, for it increases the reflective surface. The comparison between winter and

spring should highlight the influence of the surface (snow cover, ground mainly covered with leaves), air temperature and air humidity, whereas the comparison between summer and winter should show the joint influence of all four factors (air temperature, air humidity, surface, leafiness of stand roof). In the analysis, the sample was limited to flat relief and forest without regeneration.

The results of noise attenuation according to the distance from the source show (Fig. 9) that attenuation is not significantly dependant on the season (F (2,84) =1.575, p = 0.213), which can be ascribed to opposing effect of the mentioned factors on noise spreading. For example, in summer the high air temperature attenuates, whereas leafiness of stand roof increases noise spreading; in winter the low air temperature increases, whereas the snow cover attenuates noise spreading.

The noise analysis by frequency spectrum was expected to show that the forest would have higher stifling level at higher frequencies due to lack of leafiness in winter and spring time, since the sound reflective surface is smaller, and that the influence of snow cover would be evident in winter at lower frequencies. The frequency spectrum of relative noise attenuation according to 5 meters from the noise source shows us, that the differences between seasons are best evident at the distance of 80 m and mainly at the frequencies lower than 500 Hz, i.e. in the areas with background and ground influence

Fig. 10 Noise attenuation by frequency bands according to the distance from the noise source and season Slika 10. Prigu{enje buke po frekvencijskim pojasima ovisno o udaljenosti od izvora buke i o godi{njem dobu 146

Croat. j. for. eng. 31(2010)2

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

(Fig. 10). The analysis of variance in the frequency area below 100 Hz shows us that the background noise and consequently its influence is the lowest in winter and spring, and the highest in summer. The differences in noise are at the distance of 80 meters significant between summer on the one hand and winter and spring on the other (T2, MDSummer/Winter = 11.92 dB, p < 0.001; MDSummer/Spring = 6.89 dB, p = 0.002), and insignificant between winter and spring (T2, MDWinter/Spring = 5.04 dB, p = 0.140).

To establish the influence of ground surface, we analyzed the frequency range between 200 and 500 and found out statistically insignificant influence of seasons on noise attenuation at 80 meters from the noise source (T2, all p > 0.05). Plausible reason for this can be attributed to too wide frequency range. The highest attenuations by seasons pertain to different frequency bands; in winter at 160 Hz, in spring at 250 Hz, and in summer at 400 Hz. If we limit the analysis to the frequency range between 160 Hz and 400 Hz, we confirm different noise attenuation for summer and winter time (T2, MDSummer/Winter = 4.87 dB, p = 0.008), meaning that the snow cover and unleafy stand roof have higher influence on noise attenuation than the influence of forest ground surface and leafy stand roof. The same reason can be attributed to significant differences between summer and winter (T2, MDSummer/Winter = 2.79 dB, p = 0.005) in the frequency range from 630 Hz to 10 kHz. In this way we confirmed the hypothesis that in winter time the noise spreading is less intensive than in other seasons due to influential factors.

I. Poto~nik and A. Poje

5. References – Literatura Anon., 1995: Forest Operations and Badger Setts, Forestry Practice Guide 9, Forestry Practice Division, Edinburgh.

Berlund, B., Lindvall, T., 1995: Community Noise, Center for Sensory Research, Stockholm, p. 233.

Fang, C.-F., Ling, D.-L., 2003: Investigation of the noise reduction provided by tree belts. Landscape and Urban Planning 63(4): 187–195.

Crocker, M. J., 1998: Handbook of acoustics. John Wiley & Sons, Inc., Canada, 1491 p.

Delaney, D. K., 1999. Effects of helicopter noise on Mexican spotted owls. Journal of wildlife management 63 (1): 60–76.

Hansen, C., 2005. Noise control – from concept to application. Taylor & Francis, Oxon, USA, 419 p.

Kaseloo, P. A., 2006: Synthesis of noise effects on wildlife populations. IN: Proceedings of the 2005 International Conference on Ecology and Transportation, Eds. Irwin CL, Garrett P, McDermott KP. Center for Transportation and the Environment, North Carolina State University, Raleigh, NC: p. 33–35.

Pal, A. K., 2000: Noise attenauation by green belts. Journal of Sound and vibration 234(1): 149–165.

Tempel, J. D., Gatiérrez, R. J., 2003: Fecal corticosterone levels in California spotted owls exposed to low-intensity chainsaw sound. Wildlife Society Bulletin 31(3): 698–702.

Wojtkowiak, R., 2007: Measurements of noise resulting from cutting chain movements on a chain-saw bar, lubricated with different oils. Acta Sci. Pol. Silv. Colendar. Rat. Ind. Lignar 6(1): 85–93.

Sa`etak

Zaga|enje {umskoga okoli{a bukom pri izvo|enju {umskih radova Svrha je istra`ivanja da se utvrdi razina zaga|enosti okoli{a bukom pri sje~i i izradi stabala te terenski uvjeti koji utje~u na {irenje buke u okoli{u. U istra`ivanju je kao izvor buke kori{tena redovito odr`avana profesionalna jednogodi{nja motorna pila STIHL MS 460. Ispitivanje je provedeno pri trupljenju bukovih trupaca prosje~noga promjera 30 cm. Odabir radnih operacija bio je odre|en prema standardima za mjerenje razine buke motorne pile i njezina utjecaja na uho operatera (ISO 7182:1996).

U istra`ivanju su obra|ena tri ~imbenika koji imaju utjecaj na {irenje buke u {umi: prigu{ivanje izvora buke, reljef i godi{nje doba. Na prigu{ivanje izvora buke klju~ni utjecaj ima prisutnost pomlatka u sastojini koji uobi~ajeno zaklanja izvor buke ve} na udaljenosti od 5 m. Utjecaj je reljefa promatran s obzirom na {irenje buke nizbrdo, uzbrdo i na ravnom terenu. Mjerne su to~ke postavljene na 5, 10, 20, 40 i 80 metara lijevo i desno od izvora buke. Na svakoj je to~ki mjerenje trajalo 10 sekundi. Samo je najglasniji interval od 5 sekundi uklju~en u daljnju analizu buke jer bi se za trajanje buke od 10 sekundi ponovno morao prerezati trupac. Kako bi se prikupili podaci ostalih izvora buke, tako|er je snimana »prirodna« {umska buka zajedno s ostalim izvorima buke (prelet zrakoplova) u razli~itim danima i u razli~itim intervalima.

Mjerenja su provedena zvukomjerom Brüel&Kjaer 2250, pomo}u kojega je snimana ukupna ekvivalentna razina buke u sekundnim intervalima i ekvivalentna razina buke s 1/3 oktave frekvencijskoga spektra, oboje s vremenskim i frekvencijskim uravnote`enjem (jednad`ba 1). Mjerenje je bez frekvencijskoga uravnote`enja Croat. j. for. eng. 31(2010)2

147

I. Poto~nik and A. Poje

Noise Pollution in Forest Environment Due to Forest Operations (137–148)

provedeno zbog ~injenice da su svi postoje}i frekvencijski filteri (A – D) prilago|eni ljudskomu sluhu. Buka je mjerena u frekvencijskom spektru od 12,5 do 20 kHz. Mikrofon je (Brüel&Kjaer ZC 0032) postavljen otprilke 150 cm iznad zemlje. Za vrijeme mjerenja na mikrofon je postavljena za{titna spu`va (Brüel&Kjaer UA 0237).

Ustanovljeno je da se ukupna razina buke motorne pile izjedna~ava s bukom vjetra na udaljenosti od 140 m od motorne pile, a s prirodnom bukom u {umi na udaljenosti od 252 m. Buka je motorne pile sli~na pozadinskoj buci na 60 – 80 m i frekvenciji ispod 80 Hz, odnosno iznad 12,5 Hz. Iz navedenoga izlazi manji utjecaj buke na prirodni okoli{ u tim frekvencijskim pojasima. Pretpostavka da {uma potpunoga sklopa smanjuje {irenje buke puno usje{nije nego {uma prekinutoga sklopa nije potvr|ena. U {umi potpunoga sklopa odbija se zvuk te se zato manje apsorbira zvuk. Nadalje, razlika je uo~ena na udaljenosti od 80 m s obzirom na smanjenje razine buke u razli~itim godi{njim dobima: zima – ljeto (razlika 11,92 dB), prolje}e – ljeto (razlika 6,89 dB), dok je nezna~ajna razlika izme|u zime i prolje}a.

Smanjenje buke s obzirom na udaljenost od izvora nije zna~ajno pod utjecajem godi{njega doba, {to se mo`e pripisati suprotstavljenim u~incima na ~imbenike koji utje~u na {irenje buke. Tako, na primjer, ljeti visoka temperatura smanjuje, dok razdoblje bez li{}a na kro{njama potpoma`e {irenje buke. Tako|er, zimi niska temperatura pove}ava, dok snje`ni pokrov smanjuje {irenje buke.

O~ekivalo se da }e analiza buke po frekvencijskim spektrima pokazati kako {uma ima ve}i prigu{uju}i u~inak pri vi{im frekvencijama zimi i u prolje}e zbog manjka lisnoga pokrova, kada je povr{ina za odbijanje zvuka manja, a da }e snje`ni pokrov prigu{iti buku pri ni`im frekvencijama.

Frekvencijski spektar relativnoga smanjenja buke na 5 m od izvora pokazuje da je razlika izme|u godi{njega doba najbolje uo~ljiva na udaljenostima od 80 m i frekvancijama ni`im od 500 Hz (slika 10). Analiza varijance pri frekvencijama ispod 100 Hz pokazuje najmanji utjecaj pozadinske buke zimi i u prolje}e, a najve}i ljeti. Usporedba ljeta s jedne strane te zime i prolje}a s druge strane pokazuje zna~ajne razlike u razini buke (T2 MDljeto/zima = 11,92 dB, p < 0,001; MDljeto/prolje}e = 6,89 dB, p = 0,002) i nezna~ajne izme|u zime i prolje}a (T2, MDzima/prolje}e = 5,04 dB, p = 0,140).

Kako bi se utvrdio utjecaj povr{ine tla, ra{~lanjene su frekvencije izme|u 200 i 500 Hz. Utvr|en je statisti~ki bezna~ajan utjecaj godi{njih doba na smanjenje razine buke na 80 metara od izvora (T2, svi p > 0,05), vjerojatno zbog {irokoga raspona frekvencija. Prigu{enje se buke po godi{njim dobima razlikuje s obzirom na frekvenciju: zimi je najve}e na 160 Hz, u prolje}e na 250 Hz i ljeti na 400 Hz. Ako se ograni~i analiza na frekvencijski raspon od 160 do 400 Hz, mo`e se potvrditi razli~ito prigu{enje buke za ljeto i zimu (T2, MDljeto/zima = 4,87 dB, p = 0,008), {to zna~i da snje`ni pokriva~ i kro{nje bez li{}a imaju ve}i utjecaj na smanjenje buke nego povr{ina zemlje i olistale kro{nje. Isti se razlog pripisuje zna~ajnoj razlici izme|u ljeta i zime (T2, MDljeto/zima = 2,79 dB, p = 0,005) u frekvencijskom rasponu izme|u 630 Hz i 10 kHz. Ovo je istra`ivanje prvi korak u ispitivanju zaga|enja prirodnoga okoli{a bukom zbog obavljanja {umskih radova. Potrebno je nastaviti s daljnjim istra`ivanjima na ostalim izvorima buke (`i~arama, traktorima, harvesterima, forvarderima i ostalom). Nadalje, ostaje otvoreno pitanje negativnoga djelovanja buke na divlja~.

Authors’ address – Adresa autorâ:

Received (Primljeno): October 22, 2010 Accepted (Prihva}eno): December 09, 2010

148

Assoc. Prof. Igor Poto~nik, PhD. e-mail: igor.potocnik@bf.uni-lj.si Anton Poje, MSc. e-mail: anton.poje@bf.uni-lj.si University of Ljubljana Biotechnical Faculty Department of Forestry and Renewable Forest Resources Ve~na pot 83 1000 Ljubljana SLOVENIA Croat. j. for. eng. 31(2010)2

Original scientific paper – Izvorni znanstveni rad

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral and Vegetable Oils Alois Skoupy, Radomir Klvac, Seyedmohammad Hosseini Abstract – Nacrtak Similarly to many other countries, the use of biologically degradable oils in forestry is also addressed by the Czech law. Several studies point out several technical problems regarding such regulations. It has not been demonstrated so far whether for example biologically degradable engine oils used for chain saw lubrication from mixture may be the cause of an excessive engine wear or deterioration of the combustion process and hence increased contamination of air inhaled by the operator. An experimental laboratory measurement was taken for the purpose of determining the external characteristics of a common chain saw engine at a brake stand, which enabled exact measurement of differences in engine output, fuel consumption and composition of exhaust gases (CO, CO2, and HC), namely in dependence on the type of oil and the blending ratio. The results of the laboratory tests did not reveal any statistically significant differences between the oils in any of the measured criteria. The theory of workers based on practical experience that some oils may cause clogging of the fine fuel filter in the carburettor and that increased carbon sedimentation occurs in the engine exhaust duct was neither displaced by evidence, nor corroborated. Its refutation or confirmation would only be possible on the basis of a longer service test.

Keywords: biologically degradable engine oil, two-stroke engine, speed characteristic, composition of exhaust gases, laboratory testing

1. Introduction – Uvod In order to reduce negative impacts of leakages from operational fuel charge of forest machines, several countries have taken the step of replacing mineral products by vegetable-based lubricants and fuels. These substances are expected to have much faster degradation characteristics in the natural environment.

Lauhanen et al. (1998), among others, assessed the possible effects of mineral oil on the forest environment in the course of several years. They calculated that during motor-manual felling of 200 m3 of merchantable timber volume per hectare a chain saw consumes 20 litres of lubricating oil, which leaks into the forest environment. Their overall estimate of the amount of such oil spilt into the forest environment in Finland amounts annually to 2 million litres. Wightman et al. (1999) evaluated and compared relative environmental impacts and soCroat. j. for. eng. 31(2010)2

cio-economic costs of several types of oil produced in the Great Britain from rape or on mineral basis in a model case study concerning oils for chain lubrication of chain saws with the use of the LCA (Life Cycle Assessment) and CBA (Cost Benefit Analysis) methods.

Only few authors analysed the impact of the exhaust gases composition of two-stroke engines according to the type of fuel or oil used to lubricate from mixture.

Apparently the adjustment of the carburettor should be carried out in full compliance with the prescribed conditions, and the engine should be kept at nominal running speed by working load and experience of the operators rather than by structural design of the engine, both affecting the quantity and composition of the exhaust gases. Idle speed (as well as maximum speed without loading) increases the proportion of incompletely combusted hydrocarbons – HC and CO (Wojcik and Skarzynski 2006).

149

A. Skoupy et al.

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

Testing two-stroke engines for different types of fuel – aliphatic petrol with synthetic lubricating oil and traditional unleaded petrol in combination with mineral lubricating oil – proved that if a mixture of petrol with mineral oil is used, the content of hydrocarbons in the exhaust gases is 5–10 times higher compared to the synthetic mixture (Magnusson et al.

2000a). However, it is the type of used petrol rather than the lubricant oil that makes a significant difference (Magnusson et al. 2000b).

2. Material and methods – Materijal i metode The first stage of the testing methodology for two-stroke engine lubricating oils used in chain saws and brush cutters was based on a simple comparison of the basic speed characteristics for engines of a specific type in using various oils with different blending ratios. External speed characteristics of the engine were designed in line with the applicable standards ^SN 30 2008 »Automobile engines – Tests at a brake stand« and ISO 7293 »Forest machinery – Portable chain saws – Engine performance and fuel consumption«.

The test compared the behaviour of torque, actual engine output, specific consumption and consumption per hour, engine temperature below the spark-plug and concentration values of CO, CO2 and HC in exhaust gases under specific conditions.

All tests were carried out on a chain saw, whose engine can be characterized according to the following technical documentation: Þ engine type: reciprocating piston, Þ operating mode: ignition engine, Þ working cycle: two-stroke, Þ number of cylinders: 1, Þ arrangement: upright, Þ calibre: 50.0 mm, Þ stroke: 34.0 mm, Þ total cubic capacity: 66.7 cm3, Þ cooling – by air, crankshaft fan, Þ nominal speed: 8500 revs. per minute, Þ idle speed: 2500 revs. per minute, Þ centrifugal clutch switching speed: 3500 revs. per minute, Þ carburettor: floatless, TILLOTSON HS 234 A, Þ basic setting of the pilot jet: 1.0 revolution, Þ basic setting of the main jet: 1.0 revolution.

This engine was selected deliberately because no accurate and repeatable measurements would be attainable for smaller higher-speed engines under the given laboratory conditions (see below).

150

The following machines and apparatuses were used during the actual testing that took place in the testing lab for combustion engines at Brno Technical University: Þ electric eddy-current brake SCHENCK W 40, Þ thermometer THERM 5500 – 3, Þ exhaust gases analyser SUN SGA 9000 – measures the volume of CO, %, CO2, % and HC, ppm, Þ calibrated vessel of 25cm3 for measuring fuel consumption.

The chain saw required several modifications for adjusting it to the brake stand prior to the start of testing, which facilitated its firm fixing on a special clamping table. The following parts were removed from the chain saw: Þ front handgrip, Þ clutch case, Þ chain catcher, Þ centrifugal clutch, Þ chain tensioning pin, Þ oil pump, Þ spring-loaded handgrip part with the fuel tank and rear handgrip, Þ fuel inlet tube.

The chain saw had to be fitted with the following special parts required for its proper functioning: Þ control mechanism of the carburettor throttle valve, Þ GUFERO gasket ring in a special casing used to seal the crankcase.

Several parts of the chain saw had to be adjusted to allow for the installation of scanning sensors of the measuring instruments: Þ exhaust silencer was adjusted for sampling of exhaust gases; a hole was drilled in the silencer to which a tube fitting was soldered, Þ sealing below the ignition coil was modified for the connection of thermometer sensors.

After these modifications were completed, the fastening screws on the tested saw bar of the chain saw were firmly bolted to a specially adjusted test bench by two nuts. The chain saw was also fastened by the openings used for attaching the flexible elements of the antivibration system to prevent possible damage to the saw due to vibrations or other mechanical impacts during the test. The openings were reinforced crosswise by two spacers welded to a steel strip connected to the mandrel. Rotational movement of the chain saw along the crankshaft axis was forestalled by pins fastened in places where the complex chain saw construction (chiefly the engine block) allowed it. The chain saw was then connected to the Croat. j. for. eng. 31(2010)2

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

electric eddy-current brake SCHENCK W 40 by means of a transmission shaft over a fast coupling, expansion axial coupling and Hardy flexible coupling. An external fuel feed and external control of the throttle valve were connected to the chain saw. HOTTINGER sensor was used to measure the retarding force value required for the calculation of the engine torque.

The actual measurements were taken in the range between 4000 and 9000 revolutions per minute (each measurement was performed with a speed increase in steps of 1000 revolutions). During each measurement cycle, individual measurements were carried out when the values became stable. Each cycle was repeated four times. Arithmetic means were calculated for all monitored variables from individual measurement results. These were used for the calculation of the standardized variables (^SN 30 2008) of the corrected actual torque, actual output, corrected actual output, fuel consumption per hour and specific fuel consumption.

The following formulas were used for the calculations: Þ corrected torque Mt kor (Standard: ^SN 30 2008) 1013  273 + ta  0,5  , Nm Mt kor = Mt (1) ×  273 + 20  pa Where: Mt measured torque, Nm pa atmospheric pressure, hPa ta air temperature, °C Þ actual output Pe (Standard: ^SN 30 2008) Mt × n Pe = , kW (2) 9550 Where: Mt measured torque, Nm n engine revolutions, revs. per minute

A. Skoupy et al.

Þ corrected actual output Pe ´ kor (Standard: ^SN 30 2008) 1013  273 + ta  0,5  , Nm Pe ´ kor = Pe (3) ×  273 + 20  pa Þ fuel consumption per hour Gt (Standard: ^SN 30 2008) Vm × dp , k × gh–1 (4) Gt = 3,6 ´ tm Where: Vm volume of the measuring vessel, ml dp fuel density, g/cm3 tm time necessary for consuming the volume Vm, s Þ specific fuel consumption ge (Standard: ^SN 30 2008) Gt ×1000 ge = (5) , g × kw–1 Pe

3. Results – Rezultati The above method was used to determine the values of torque, air intake temperature, CO, CO2 and HC concentration in the exhaust gases and the time required for the consumption of 25 ml of the fuel mixture. The values of the corrected torque, net and corrected outputs, fuel consumption per hour and specific consumption were calculated. The measured and calculated data were recorded in tables as shown in Table 1, where it can be seen that each measurement series was repeated four times for each oil type and blending ratio for the whole speed range.

The results of the Husqvarna 266 SE engine speed characteristics were subsequently subjected to statistical processing. Regression calculations were per-

Table 1 Behaviour of HC concentration in exhaust gases Tablica 1. Izmjerene i izra~unate vrijednosti brzinskih zna~ajki Measurement Mjerenja

kW 1.53 1.57 1.56 1.56

Time Vrijeme s 60.88 61.04 61.11 61.18

kg.h–1 g.kWh–1 1.050 696 1.047 676 1.046 675 1.044 674

1.55

61.05

1.047

Mt

Mt kor

Pe

Pe kor

Nm 3.6 3.7 3.7 3.7

Nm 3.64 3.74 3.73 3.73

kW 1.51 1.55 1.55 1.55 1.54

1 2 4 4 Average 3.7 3.71 Prosjek Used Oil – Uporabljeno ulje Blending ratio – Omjer mje{avine

Croat. j. for. eng. 31(2010)2

gp

mp

680

Air temp.

Temp. zraka °C 19.6 19.5 20.6 20.5 20.1

Air press. Engine temp.

Tlak zraka Temp. motora hPa °C 1,000 140 1,000 140 1,007 140 1,007 140 1,004

140

CO

CO2

HC

% 4.70 4.63 4.58 4.51

% 3.47 3.21 2.81 2.78

ppm 3,150 2,940 2,773 2,791

4.61

3.07

2,914

2T/6 Engine: Husqvarna 266 SE – Tip motorne pile: Husqvarna 266 SE 1:40 Speed: 4,000 revs./min – Brzina: 4000 okr./min.

151

A. Skoupy et al.

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

Further to the above statistical assessment results, it can be said that very good measurement repeatability was achieved and that the engine oils can be evaluated on the basis of the method that compares the speed characteristics.

It can be, therefore, stated that for example the torque value does not correlate with the blending ratio, while it seems that the lower the concentration of oil in the mixture, the lower the actual output. In comparison to the mixture containing the OA M6A oil, the maximum output in the case of the 2T/6 vegetable oil was generally lower, while it was higher when compared to the mixture with the MOGUL oil.

These and other possible dependencies can be established from the correlation matrices in shown in Tables 2 through 4.

The above tables show mutual correlation indexes between the engine speed, engine temperature and air intake temperature based on the type of used motor oils; the behaviours of noxious agent concentrations are assigned to the correlation indexes. The table for the 2T/6 oil also specifies the dependencies on the blending ratio and atmospheric pressure, which underwent significant changes during the measurements of these motive fluids.

It is particularly significant that CO2 concentration depends more on the air intake temperature and on the engine temperature (in the case of the 2T/6 oil also on the air pressure) than on the engine speed.

CO concentration is closely linked to engine temperature with all of the tested oils.

formed for the behaviour of individually measured variables for all six tested fuel mixtures. The behaviour of individual parameters depending on the engine speed was best expressed by the functions: y = a0 + a1 ´ x + a2 ´ x2 + a3 ´ x3 + a4 ´ x4 + a5 ´ x5 (6) or y = a 0 + a1 ´ x + a 2 ´ x 2 + a3 ´ x 3 + a4 ´ x 4

(7)

Only in the case of CO2 concentration during the test of MOGUL TS oil mixture in combination with BA 90 petrol in the ratio of 1:40 the following function proved to be more suitable: y = a0 + a1 ´ ln(x) + a2 ´ ln(x)2 + a3 ´ ln(x)3 + a4 ´ ln(x)4 + a5 ´ ln(x)5

(8)

The specific coefficients of functions of approach, reached correlation indexes and extreme values of functions – i.e. the maxima during specified engine revolutions – were calculated on the basis of this formula. The minimum was expressed only in the case of specific fuel consumption.

The above stated dependencies were calculated from all of the measurement series. Mean error did not exceed 3% with the majority of criteria. Only the measurements of noxious substances emissions in exhaust gases usually showed higher mean error values; the mean error value for the CO2 concentration once reached nearly 10%.

Table 2 Correlation matrix for the OA M6A oil Tablica 2. Me|usobni odnosi zna~ajki za ulje OA M6A OA M6A Speed – Brzina Engine temperature Temperatura motora Intake temperature Temperatura zraka

Speed Brzina 1.0000

Engine temperature Intake temperature Temperatura motora Temperatura radnoga prostora 0.8140 0.0194

CO concentration Koncentracija CO

CO concentration Koncentracija CO

CO concentration Koncentracija CO

0.9155

0.4187

–0.9575

0.8140

1.0000

–0.0868

0.6867

0.7201

–0.7998

0.0194

–0.0868

1.0000

0.1153

–0.5303

–0.0864

Table 3 Correlation matrix for the MOGUL TS oil Tablica 3. Me|usobni odnosi zna~ajki za ulje MOGUL TS Mogul TS Speed – Brzina Engine temperature Temperatura motora Intake temperature Temperatura zraka

152

Speed Brzina 1.0000

Engine temperature Temperatura motora 0.9139

Intake temperature Temperatura zraka 0.0639

CO concentration Koncentracija CO 0.8870

CO concentration Koncentracija CO 0.4352

CO concentration Koncentracija CO –0.8605

0.9139

1.0000

0.0064

0.8277

0.6143

–0.8157

0.0639

0.0064

1.0000

0.1517

–0.5768

–0.3441

Croat. j. for. eng. 31(2010)2

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

A. Skoupy et al.

Table 4 Correlation matrix for the 2T/6 oil Tablica 4. Me|usobni odnosi zna~ajki za ulje 2T/6 2T/6 Speed – Brzina Blending ratio Omjer mje{avine Engine temperature Temperatura motora Intake temperature Temperatura zraka Air pressure Tlak zraka

Speed Brzina 1.0000

Blending ratio Engine temp. Intake temp. Air pressure CO concentration CO2 concentration HC concentration Omjer mje{avine Temp. motora Temp. zraka Tlak zraka Koncentracija CO Koncentracija CO2 Koncentracija HC 0.0000 0.8666 0.0796 0.0000 0.8781 0.4604 –0.8730

0.0000

1.0000

–0.0441

–0.1202

0.0000

0.0144

–0.1589

–0.1874

0.8666

–0.0441

1.0000

0.1045

0.0352

0.7593

0.5782

–0.7549

0.0796

–0.1202

0.1045

1.0000

0.9431

0.1435

–0.5843

–0.3447

0.0000

0.0000

0.0352

0.9431

1.0000

0.0809

–0.6698

–0.3063

Fig. 1 Behaviour of HC concentration in exhaust gases Slika 1. Kretanje koncentracije HC-a u ispu{nim plinovima CO and HC concentrations are dependent on engine speed and – due to close dependence of the engine temperature on its speed – also on the engine temperature.

It is rather surprising that the oil blending ratio does not significantly affect any of the noxious substances concentrations in exhaust gases.

Fig. 1 is the example of a graphic representation of the results, showing that with the increasing engine speed the combustion of the HC motive fluid improves with respect to the HC volume.

4. Conclusion – Zaklju~ci Evaluation according to changes in the external speed characteristics of engines at a brake stand is a Croat. j. for. eng. 31(2010)2

standard assessment method during engine oil testing. Statistical evaluation revealed that a very good measurement repeatability was achieved by way of this method. The conclusions arising from the results specified in this paper are as follows: Þ engine torque value does not correlate with oil blending ratio in the fuel, Þ actual engine output probably drops with decreasing oil concentration in the fuel, Þ maximum output is lower when vegetable oils are used as opposed to mineral oils, Þ CO2 concentration is more dependent on air intake and engine temperatures than on engine speed with all of the tested oils; air pressure is a significant factor in the case of BIO 2T/7 MIX oil, Þ CO concentration is closely linked to engine temperature with all of the tested oils, Þ CO and HC concentrations depend on engine speed and – due to very close dependence of the engine temperature on its speed – also on engine temperature, Þ blending ratio does not substantially affect the concentration of any of the monitored components of exhaust gases.

We believe that despite the achieved accuracy, the results do not correspond to generally held assumptions such as that the decreasing concentration of oil in the fuel causes the content of noxious agents in exhaust gases to fall and the engine output to grow. It is further surprising that the OA M6A oil, a traditional motor oil now only used for veteran cars, seemed to have the best qualities, while the MOGUL TS oil, a specially developed oil for two-stroke engines with high working loads, ranked worst. This was probably caused by the dependence on oil concentration rather than on its quality. It can be, therefore, stated that all of the tested fuel mixtures can be used for a trouble-free operation.

153

A. Skoupy et al.

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

Acknowledgment – Zahvala This paper was prepared in the framework of the research projects of the Ministry of Education of the Czech Republic »Forest and Wood. Support to a functionally integrated forest management«, MSM 6215648902, and of the Ministry of Agriculture of the Czech Republic »Sophisticated model for naturefriendly timber haulage evaluation«, QH71159.

5. Refrences – Literatura Lauhanen, R., Kolppanen, R., Kuokkanen, T., Sarpola, S., Lehtinen, M., 1988: The environmental effects of oils used in forest operations. Teho Helsinki 48(4): 32–34.

Magnusson, R., Nilsson, C., Andersson, K., Andersson, B., Gieling, R., Wiberg, K., Ostman, C., Rannug, U., 2000a: Determination of chemical composition and mutagenicity in particles from chainsaw exhaust. Experimental set-up, stability and results from two different fuels. Environmental technology 21(7): 819–829.

Magnusson, R., Nilsson, C., Andersson, K., Andersson, B., Rannug, U., Ostman, C., 2000b: Effect of gasoline and lubricant on emissions and mutagenicity of particles and semivolatiles in chain saw exhaust. Environmental science & technology 34(14): 2918–2924.

Wightman, P., Eavis, R., Batchelor, S., Walker, K., Bennett, R., Carruthers, P., Tranter, R., 1999: Comparison of Rapeseed and Mineral Oils Using Life-Cycle Assessment and Cost-Benefit Analysis. OCL-Oleagineux Corps Gras Lipides 6(5): 384–388.

Wojcik, K., Skarzynski, J. G., 2006: Emission and composition of exhaust gases by new chain saws produced by Husqvarna and Stihl. Acta Scientiarum Polonorum, Silvarum Calendarum ratio et Industria Lignaria 5(2): 147–157.

Standard: ^SN 30 2008. Automobile engines – Tests at a brake stand.

Standard: ISO 7293:1997. Forest machinery – Portable chain saws – Engine performance and fuel consumption.

Sa`etak

Promjene brzinskih zna~ajki motora motornih pila pri uporabi mineralnih ulja i ulja biljnoga podrijetla U nastojanju da se smanji negativan utjecaj izlijevanja goriva iz spremnika {umskih strojeva vi{e je zemalja po~elo zamjenjivati mineralna goriva i ulja gorivima i uljima biljnoga podrijetla koji se brzo razgra|uju u prirodi.

U Republici ^e{koj, kao i u ostalim zemljama, uporaba je biorazgradivih ulja u {umarstvu propisana zakonima.

Nekoliko studija isti~e tehni~ke probleme vezane uz zakonske propise. Do sada nije istra`ivan utjecaj uporabe biorazgradivih motornih ulja u motornim pilama na pretjerano tro{enje dijelova motora i pogor{anje sagorijevanja koje pove}ano one~i{}uje zrak i radni okoli{ operatera. Zbog toga su obavljena pokusna laboratorijska mjerenja za utv|ivanje zna~ajki uobi~ajenih motora motornih pila na elektri~noj ko~nici, {to omogu}uje precizna mjerenja razlika u radu motora, potro{nji goriva i sastavu ispu{nih plinova (CO, CO2 i HC), ovisno o vrsti ulja i omjeru mije{anja ulja i benzina.

Samo nekolicina autora istra`uje sastav ispu{nih plinova dvotaktnih motora prema vrsti goriva ili ulja koje se u mje{avini koristi za podmazivanje. Prazan hod, jednako kao i najve}i broj okretaja motora bez optere}enja, pove}ava omjer nepotpuno sagorjelih ugljikovodika – HC i CO (Wojcik i Skarzynski 2006). Magnusson i dr.

(2000a) ispitivanjem dvotaktnih motora s razli~itim vrstama goriva (alifatski benzin sa sinteti~kim uljem za podmazivanje i uobi~ajeni bezolovni benzin s mineralnim uljem) utvr|uju da je upotrebom mje{avine benzina i mineralnoga ulja koli~ina ugljikovodika u ispu{nim plinovima 5 – 10 puta ve}a u usporedbi s mje{avinom goriva i sinteti~koga ulja. Pri tome ve}i utjecaj na koli~inu ugljikovodika u ispu{nim plinovima ima uporabljeno gorivo nego samo ulje (Magnusson i dr. 2000b).

Prva je faza u istra`ivanju ulja za podmazivanje dvotaktnih motora motornih pila i motornih kosa bila temeljena na jednostavnoj usporedbi brzinskih zna~ajki motora odre|enoga tipa pri uporabi razli~itih ulja i razli~itih omjera mje{avine.

Brzinske su zna~ajke motora odre|ivane prema zahtjevima va`e}ih normi ^SN 30 2008 »Automobile engines – Tests at a brake stand« i ISO 7293 »Forest machinery – Portable chain saws – Engine performance and fuel consumption«.

Ispitivanjem su uspore|eni zakretni moment motora, trenutna snaga motora, specifi~na i satna potro{nja goriva, temperatura motora na svje}ici i vrijednost koncentracije CO, CO2 i HC u ispu{nim plinovima pod

154

Croat. j. for. eng. 31(2010)2

Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral ... (149–155)

A. Skoupy et al.

utvr|enim uvjetima. Uobi~ajena se metoda ispitivanja zna~ajki motornoga ulja zasniva na promjeni brzinske zna~ajke motora na elektri~noj ko~nici.

Metoda razumijeva mjerenje ovih vrijednosti: zakretni moment motora, broj okretaja motora, atmosferski tlak zraka, temperatura zraka, obujam utro{ene mje{avine goriva i ulja, gusto}a mje{avine goriva i ulja te vrijeme potrebno za utro{ak odre|ene koli~ine mje{avine goriva i ulja. Mjerenja su bila obavljena u rasponu broja okretaja motora od 4000 do 9000 min-1. Svako se mjerenje ~etiri puta ponavljalo te je izra`ena aritmeti~ka sredina za vrijednosti mjerenja. Izrazi 1 do 5 kori{teni su za izra~un osnovnih veli~ina u istra`ivanju. U tablici 1 prikazani su mjereni i izra~unati podaci za razli~ite vrste ulja i omjere mje{avine goriva i ulja. Statisti~kim vrednovanjem rezultata dokazana je vrlo velika mogu}nost ponavljanja mjerenja s ovom metodom.

Rezultati istra`ivanja pokazuju da omjer mje{avine goriva i ulja ne utje~e na vrijednost zakretnoga momenta motora, a stvarna se snaga motora smanjuje sa smanjenjem koncentracije ulja u mje{avini i s uporabom biorazgradivih ulja u mje{avini. Koncentracija je {tetnih ispu{nih plinova u pozitivnoj ovisnosti o temperaturi motora i broju okretaja motora.

Utvr|en je neznatan utjecaj omjera mje{avine na sastav ispu{nih plinova, {to je u suprotnosti s pretpostavkom da se smanjenjem koncentracije ulja u gorivu posti`e smanjenje {tetnih tvari u ispu{nim plinovima i pove}anje korisnosti motora. Nadalje je zanimljivo da ulje OA M6A, uobi~ajeno motorno ulje koje se rabi u starijim automobilima, ima najbolju kakvo}u, dok se ulje MOGUL TS, posebno razvijeno za dvotaktne motore s velikim optere}enjima, pokazalo najlo{ijim. Uzrok je tomu vjerojatno ve}i utjecaj koncentracije nego kakvo}e ulja na dobivene rezultate. Prema navedenomu mo`e se zaklju~iti o pouzdanosti uporabe svih ispitivanih mje{avina goriva i ulja u motornim pilama.

Klju~ne rije~i: biorazgradiva motorna ulja, dvotaktni motor, brzinska zna~ajka, sastav ispu{nih plinova, laboratorijsko ispitivanje

Authors’ address – Adresa autorâ: Assoc. Prof. Alois Skoupy, PhD. e-mail: skoupy@mendelu.cz Asst. Prof. Radomir Klvac, PhD. e-mail: klvac@mendelu.cz Mendel University of Brno Faculty of Forestry and Wood Technology Department of Forestry and Forest Product Technology Lesnicka 37 613 00 Brno CZECH REPUBLIC

Received (Primljeno): October 26, 2010 Accepted (Prihva}eno): November 26, 2010 Croat. j. for. eng. 31(2010)2

Asst. Prof. Seyedmohammad Hosseini, PhD. e-mail: S_hosseini99@yahoo.com Ministry of Science, Research and Technology Shahrak Gharb, Hormoyan Steet 26, Tehran IRAN

155

Original scientific paper – Izvorni znanstveni rad

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network Damir Klobu~ar, Renata Pernar, Sven Lon~ari}, Marko Suba{i}, Ante Seletkovi}, Mario An~i} Abstract – Nacrtak Forest dieback is taking on increasing proportions in many parts of Croatia. To improve the situation, it is of primary importance to acquire timely, accurate and inexpensive information on the scale of forest damage. Such information can be collected for large forest areas with remote sensing techniques. This paper explores the possibility of applying segmentations of color infrared aerial photographs (CIR). Self-organizing artificial neural networks are used to detect damage in beech-fir forests and determine its spatial distribution. The results of the research confirm the benefits of applying neural networks to forest damage detection, since there are no statistically significant differences between damage in the field and damage detected with a neural network.

Key words: forest damage, color infrared aerial photographs, segmentation, neural networks, Croatia

1. Introduction – Uvod An increase in the proportion of snag yield in the prescribed annual cut has a negative effect on sustainable forest management. Growing amount of salvage cutting operations requires that the focus be placed on forest health status and the quantity of snag yield. The primary task is to locate stands of poorer health in order to maintain their vitality and naturalness at an optimal level by applying timely measures (Pernar et al. 2007 b). Snag inventories are generally made with an intensive terrestric method, which requires substantial investments. Over large areas, however, it is much more practical, financially more acceptable and more reliable to apply a remote sensing method. Some common uses of CIR aerial photographs in forestry include site planning, description and mapping, temporal studies or disease assessments (Bütler and Schlaepfer 2004). Their features are particularly suited to making inventories of vegetation damage, especially of forests and forest trees (Kalafad`i} 1987, Pernar 1994, Haara and Nevalainen 2002, Pernar et al. 2007a, b).

Owing to the six inventories of forest damage undertaken in the Republic of Croatia, this method has reached an operative level. The application of CIR aerial photography in the assessment of the forest Croat. j. for. eng. 31(2010)2

condition has proved to be of equal value to terrestrial working methods in terms of accuracy, but much more efficient in terms of speed and objectivity (Pernar and Ku{an 2001). According to Mas and Ramirez (1996), the most widely used method is visual interpretation, which achieves the most accurate results, primarily due to human ability to identify features/phenomena of interest. On the other hand, the process itself is relatively time-consuming because each part of the photograph is analyzed separately. Consequently, depending on the size of the area to be classified, this can significantly increase processing time and costs.

An alternative approach to the application of remote sensing in forestry involves the use of artificial neural networks (Ardö et al. 1997, Skidmore et al.

1997, Wang and Dong 1997, Moisen and Frescino 2002, Ingram et al. 2005, Kuplich 2006, Joshi et al.

2006, Verbeke et al. 2006, Klobu~ar et al. 2008, Klobu~ar and Pernar 2009). In general, the use of artificial neural networks in remote sensing began in the early 1990s (Benediktsson et al. 1990, Hepner et al. 1990, Civco 1993). Among neural networks with supervised learning, the most commonly used model is the multilayer feed – forward network – MLP (Paola and Schowengerdt 1995, Atkinson and Tatnall 1997, Kanellopoulos and Wilkinson 1997), followed to a

157

D. Klobu~ar et al.

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157–163)

much lesser extent by neural networks with radial basic functions (RBF) and a probabilistic neutral network (PNN) (Foody 2001). Neural networks with unsupervised learning, such as, e.g. self-organizing neural networks, are used even less (Beamish 2001).

This research deals with detection of forest damage and determination of its spatial distribution using scene segmentation (CIR) and self-organizing neural networks. The multiple goals are to acquire data with less investments, achieve a high degree of automation which will remove subjectivity contained in classical remote sensing methods, and present some possibilities of applying artificial intelligence to the monitoring of forest ecosystems.

2. Self-Organizing Maps – Samoorganiziraju}a neuronska mre`a Self-organizing neural networks belong to the category of neural networks with unsupervised learning. Unlike supervised learning, there is no information on the expected response for samples from the training set; consequently, no error of the neural network is known that could be used to manage network training. In unsupervised learning, neural networks are trained by sample detection, by correlation or by input data categories. Each neuron in a self-organizing map has a weight vector whose number of elements corresponds to dimensionality of input data. A self-organizing map first finds the winning neuron. This is followed by the adjustment of a weight factor according to Kohonen’s learning rule: iw(q) = iw(q – 1) + a(p(q) – iw(q – 1)), or iw(q) = (1 – a) ´ iw(q – 1) + ap(q)

If the current value of neuron weights is perceived as coordinates in the space of input features, the neurons of a self-organizing map are trained to shift in this space towards the centers of input vector groups. This allows classification of a particular class of input vectors with each neuron of the self-organizing map. A self-organizing map thus simplifies the problem of classification because dimensionality of the problem is implicitly reduced from an arbitrarily high number of input space dimensions to two topological dimensions of the self-organizing map. Topology is selected experimentally since there is no direct and obvious link between the expected network performance and the selected topology for the selected training set.

If the problem is defined so that sample classification from the training set is known, when training is over we determine which neurons correspond to which class of input samples. This serves as the basis for classification in further network operation.

3. Material and Methods – Materijal i metode A CIR aerial photograph of the area was first visually interpreted in the stereomodel. The image (Fig. 1) was divided into four equal parts, since the process of training the network with all the pixels is very slow or almost impossible.

The matrices (R G B components) of the four scenes for input into the neural network were then prepared. The neural network was created using the newsom function, with the number of set classes

(1)

Where: iw(q) neuron weight in a number of q iterations p(q) input vector in a given iteration and a learning constant.

The consequence of this learning rule is that the weight of the winning neuron will tend towards the input vector values. The difference between the classical competitive network and the self-organizing network is that in self-organizing maps (according to Kohonen’s learning rule) not only the weights of the winning neuron are corrected but also the weights of the neurons adjacent to the winning one. Neighborhood is defined by the distance function, which is closely connected with the selected topology of the self-organizing map. The most common topologies of self-organizing networks are two-dimensional, and less so one-dimensional. Examples of the most common two-dimensional topologies are square, triangular or hexagonal topologies (Kohonen 2001).

158

Fig. 1 CIR aerial photograph of the investigation area Slika 1. ICK aerosnimak istra`ivanoga podru~ja Croat. j. for. eng. 31(2010)2

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157–163)

D. Klobu~ar et al.

Table 1 Results of comparison between visual interpretation and segmentation for the scene S2 Tablica 1. Rezultati usporedbe vizualne interpretacije i segmentacije scene S2 Cluster – Klaster 1 2 3 4 5 6

17.29

14.97

Ho:p1=p2 p 0.90

9.84

12.14

0.83

11.92

14.84

0.85

19.05

17.24

0.91

19.61 22.29

17.63 23.18

0.95 0.96

Visual Interpretation, % – Vizualna interpretacija, % Shadows – Sjene Severely diseased trees (considerable loss of color, damage ³ 60%) Jako bolesna stabla (zna~ajan gubitak boje, o{te}enost ³ 60 %) Dead standing trees (snags) – Mrtva stabla (su{ci) Transitional pixels between vegetation and shadows Prijelazni pikseli izme|u vegetacije i sjene Diseased trees (damage 25–60%) – Bolesna stabla (o{te}enost 25 – 60 %) Healthy trees – Zdrava stabla

(clusters) for segmentation as follows: 6, 6 (2 x 3), 7, 8, 8 (2 x 4), 9, 9 (3 x 3). After training, the network segmented a given scene into the set number of classes.

The segmented scenes were then visually interpreted and compared with the results of photointerpretation of the study area, and the number of classes were defined, which provided the most acceptable segmentation. MATLAB 6.5 software was used to construct the architecture of the artificial neural networks and to conduct digital scene processing.

Segmentation, % – Segmentacija, %

A proportion test was used to test the percentage equality of visual interpretation and segmentation.

The probability that the proportions are equal ranges from 0.83 for cluster 2 (severely diseased trees) to

4. Results and Discussion – Rezultati i rasprava Visual interpretation of the study area delineated the following: four degrees of tree damage, shadows, and transition regions of pixels between vegetation and shadow. The most acceptable results or the best match between visual interpretation and segmentation of individual scenes was achieved for 6 clusters in scene S2 (topology 6, Fig. 2 – 3), which was also determined with visual interpretation of the entire scene.

Fig. 2 SOM architecture – with six clusters in the output layer Slika 2. Arhitektura SOM-a sa {est klastera u izlaznom sloju Croat. j. for. eng. 31(2010)2

Fig. 3 Scene S2 segmentation in 6 classes Slika 3. Segmentacija scene S2 u 6 klasa 159

D. Klobu~ar et al.

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157â&#x20AC;&#x201C;163)

Fig. 4 Delineated clusters of the scene S2 Slika 4. Delineirani klasteri scene S2 0.96 for cluster 6 (healthy trees). There is no statistically significant difference between visual interpretation and segmentation of the analyzed scene S2 (Table 1).

Since a digital record of the tree crown contains a larger number of pixels (different digital values), which have been distributed into a certain number of clusters during the segmentation process, it is clear that no conclusions on the damage degree of a particular tree or parts of a standing tree can be made only on the basis of the participation of a particular cluster in relation to the participation of other clusters within the crown. A scene segmented in this manner is a good basis for determining spatial damage distribution by delineating homogeneous clusters (Fig. 4).

Topology 2 x 3 provided poorer results due to the unfavorable neuron distribution in color space. In this case, the results obtained with topology, in which the neurons cover color space in a series, are more acceptable (Fig. 5).

A larger number of clusters obtained by segmentation, i.e. the use of other architectures with a larger number of neurons, did not indicate any new features that could be determined with visual interpretation.

Due to variations of each scene, it is almost impossible to expect that the results of segmentation for each scene will fully correspond to the results of

160

visual interpretation, or that segmentations for all the scenes will be identical.

The structure of the neural network for S2 scene was used in this order. A simulation of this network was made for input matrices of the scenes S, S3 and S4. By simulating the neural network (Scene S2) on

Fig. 5 Neurons in the space of R and G components Slika 5. Neuroni u prostoru sastavnica R i G Croat. j. for. eng. 31(2010)2

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157–163)

the three remaining scenes an improvement was achieved; in other words, the scenes were segmented according to the determined classes in the process of visual interpretation.

The application of simulation is justified by the fact that segmentation results were obtained rapidly; in contrast, neural network training requires a certain period of time, as well as repeated interpretation of the obtained clusters.

Such segmentation makes it possible to monitor (analyze) both the entire area and individual scenes under almost equal conditions. This is enabled by the establishment of a relationship (delineation) between the segmented classes and the corresponding colors in the entire study.

In order to successfully apply remote sensing technologies to forestry, it is necessary to carry out prior terrestrial reconnaissance, as well as check the obtained results afterwards. This relates particularly to the application of unsupervised learning, as is the case in our example.

Accordingly, certain procedures in the use of self-organizing neural network can be outlined for the purpose of detecting forest damage and determining its spatial distribution: Þ Form the input matrix of the scene in the size and shape acceptable for neural network training, Þ set the number of clusters (classes) according to the results of visual interpretation (one or two classes more can be set with the goal of determining whether segmentation provided an additional feature), Þ test the corresponding topology, Þ after determining the scene in which suitable segmentation results have been obtained, use the obtained structure of the neural network for simulation on other scenes.

5. Conclusion – Zaklju~ak Some priorities of sustainable forest management refer to detection of forest damage and its spatial distribution and monitoring the forest condition.

Remote sensing techniques can be successfully used to install these measures over large forest areas in as short a period as possible. These techniques include an integral approach that consists of utilizing good features of color infrared aerial imagery, methods of digital image analysis, artificial intelligence and visual scene interpretation. This research has proved that self-organizing neural networks can be reliably used to detect stand damage in color infrared (CIR) aerial photographs.

Croat. j. for. eng. 31(2010)2

D. Klobu~ar et al.

6. References – Literatura Ärdo, J., Pilesjo, P., Skidmore, A., 1997: Neural networks, multitemporal Landsat Thematic Mapper data and topographic data to classify forest damage in the Czech Republic. Canadian Journal of Remote Sensing 23(3): 217–229.

Atkinson, P. M., Tatnall, A. R. L., 1997: Neural networks in remote sensing. International Journal of Remote Sensing 18: 699–709.

Beamish, D., 2001: A Review of Neural Networks in Remote Sensing, 1–45 p.

Benediktsson, J. A., Swain, P. H., Evsoy, O. K., 1990: Neural network approach versus statistical methods in classification of multi-source remote sensing data. IEEE Transactions on Geoscience and Remote Sensing 28(4): 540–551.

Bütler, R., Schlaepfer, R., 2004: Spruce snag quantification by coupling colour infrared aerial photos and a GIS. Forest Ecology and Management 195(3): 325–339.

Civco, D. L., 1993: Artificial neural networks for land cover classification and mapping. International Journal of Geographical Information Systems 7: 173–186.

Foody, G. M., 2001: Thematic mapping from remotely sensed data with neural networks: MLP, RBF and PNN based appraches, Journal of Geographical Systems 3(2): 217–232.

Haara, A., Nevalainen, S., 2002: Detection of dead or defoliated spruces using digital aerial data. Forest Ecology and Management 160(1–3): 97–107.

Hepner, G. F., Logan, T., Ritter, N., Bryant, N., 1990: Artificial neural network classification using a minimal training set: Comparison to conventional supervised classification.

Photogrammetric Engineering and Remote Sensing 56: 469–473.

Ingram, J. C., Dawson, T. P., Whittaker R. J., 2005: Mapping tropical forest structure in southeastern Madagascar using remote sensing and artificial neural networks. Remote Sensing of Environment 94(4): 491–507.

Joshi, C., De Leeuw Jan., Skidmore, A. K., Van Duren, I. C., Van Oosten, H., 2006: Remotely sensed estimation of forest canopy density: A comparison of the performance of four methods. International Journal of Applied Earth Observation and Geoinformation 8(2): 84–95.

Kalafad`i}, Z., 1987: Primjena ICK aerosnimaka u {umarstvu. [umarski list 111(1–2): 61–66.

Kanellopoulos, I., Wilkinson, G. G., 1997: Strategies and best practice for neural network image classification. International Journal of Remote Sensing 18(4): 711–725.

Klobu~ar, D., Pernar, R., Lon~ari}, S., Suba{i}, M., 2008: Artificial neural networks in the assessment of stand parameters from an IKONOS satellite image. Croatian Journal of Forest Engineering 29(2): 201–211.

Klobu~ar, D., Pernar, R., 2009: Umjetne neuronske mre`e u procjeni sastojinskih obrasta s cikli~kih snimaka. [umarski list (3–4): 145–155.

Kohonen, T., 2001: Self-Organizing Maps. Springer, Berlin.

161

D. Klobu~ar et al.

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157–163)

Kuplich, T. M., 2006: Classifying regenerating forest stages in Amazonia using remotely sensed images and a neural network. Forest Ecology and Management 234(1–3): 1–9.

Pernar, R., Seletkovi}, A., An~i}, M., 2007a: Utvr|ivanje o{te}enosti Spa~vanskog bazena primjenom infracrvenih kolornih aerosnimaka. [umarski list 7–8: 315–332.

Mas, J. F., Ramirez, I., 1996: Comparison of land use classifications obtained by visual interpretation and digital processing. ITC Journal 3/4: 278–283.

Pernar, R., Seletkovi}, A., An~i}, M., 2007b: Primjena ICK aerosnimaka za utvr|uvanje o{te}enosti {uma na podru~ju U[P Gospi}. [umarski list 11–12: 507–522.

Moisen, G. G., Frescino, T. S., 2002: Comparing five modelling techniques for predicting forest characteristics. Ecological Modelling, special issue: 209–225.

Skidmore, A. K., Turner, B. J., Brinkhof, W., Knowles, W., 1997: Performance of a neural network: mapping forests using GIS and remotely sensed data. Photogrammetric Engineering and Remote Sensing 63(5): 501–514.

Paola, J. D., Schowengerdt, R. A., 1995: A review and analysis of backpropagation neural networks for classification of remotely – sensed multi spectral imagery. International Journal of Remote Sensing 16: 3033–3058.

Pernar, R., 1994: Na~in i pouzdanost odre|ivanja o{te}enosti hrasta lu`njaka (Quercus robur L.) na infracrvenim kolornim (ICK) aerosnimkama. Glas. {um. pokuse 31: 1–34, Zagreb.

Pernar, R., Ku{an, V., 2001: Aerosnimanje {uma bukve i jele pomo}u ICK snimaka za pra}enje stanja {uma.

Znanstvena knjiga »Znanost u potrajnom gospodarenju Hrvatskim {umama»: 457–463.

Verbeke, L. P. C., Van Coillie, F. M. B., DeWulf, R. R., 2006: Object-based forest stand density estimation from very high resolution optical imagery using wavelet-based texture measures. In: 1st International Conference on Object-based Image Analysis (OBIA).

Wang, Y., Dong, D., 1997: Retrieving forest stand parameters from SAR backscatter data using a neural network trained by a canopy backscatter model. International Journal of Remote Sensing 18: 981–989.

Sa`etak

Otkrivanje o{te}enosti {uma na ICK aerosnimcima pomo}u neuronske mre`e Pove}anje udjela slu~ajnoga prihoda u propisanim godi{njim etatima negativno utje~e na potrajno gospodarenje {umama. Sve ve}i udio sanitarnih sje~a tra`i da se posebna pa`nja posveti zdravstvenomu stanju {uma i koli~ini slu~ajnoga prihoda od su{aca. Potrebno je u prvom redu otkriti sastojine slabijega zdravstvenoga stanja kako bi se pravodobnim mjerama odr`ala njihova vitalnost i prirodnost na optimalnoj razini (Pernar i dr. 2007b).

Su{ci se uobi~ajeno inventariziraju intenzivnim teresti~kim na~inom, koji zahtijeva zna~ajna materijalna sredstva. Na velikim povr{inama prakti~nija je, materijalno prihvatljivija i pouzdanija metoda daljinskih istra`ivanja.

Infracrveni kolorni (ICK) aerosnimci upotrebljavaju se u {umarstvu u planiranju, opisu i kartiranju stani{ta, vremenskim studijama ili u procjeni bolesti (Bütler i Schlaepfer 2004). Zbog svojih obilje`ja vrlo su pogodni za inventarizaciju o{te}enosti vegetacije, osobito {uma i {umskoga drve}a (Kalafad`i} 1987, Pernar i Ku{an 1994, Haara i Nevalainen 2002, Pernar i dr. 2007a, b).

Kao alternativni pristup u primjeni daljinskih istra`ivanja u {umarstvu se koriste umjetne neuronske mre`e.

Stoga se u ovom istra`ivanju opisuje otkrivanje o{te}enosti {uma i utvr|ivanje njezina prostornoga rasporeda primjenom segmentacije scene i samooorganiziraju}e neuronske mre`e radi pridobivanja podataka, {to tra`i manja materijalna ulaganja. Time se posti`e visok stupanj automatizma, kojim se uklanja subjektivnost klasi~nih metoda daljinskih istra`ivanja te prikazuju neke mogu}nosti primjene umjetne inteligencije u pra}enju {umskih ekosustava.

Prvotno je provedena vizualna interpretacija ICK aerosnimaka istra`ivanoga podru~ja u stereomodelu. Zatim je snimak (slika 1) podijeljen na ~etiri jednaka dijela (S, S2, S3, S4) jer je jako sporo ili gotovo nemogu}e trenirati mre`u pomo}u svih piksela. Nakon provedenoga treniranja mre`a je segmentirala odre|enu scenu u zadani broj razreda (6 – 9). Zatim su vizualno interpretirane segmentirane scene i uspore|ene s rezultatima fotointerpretacije istra`ivanoga podru~ja te definiran broj razreda s kojima je dobivena najprihvatljivija segmentacija.

Provedenom vizualnom interpretacijom podru~ja izdvojena su ~etiri stupnja o{te}enosti stabala, sjene, te regije prijelaznih piksela izme|u vegetacije i sjene. Najprihvatljiviji rezultat, odnosno najbolja podudarnost vizualne interpretacije i segmentacije pojedinih scena postignuta je kod scene S2 za 6 klastera (topologija 6, slike 2, 3), koliko je i bilo utvr|eno vizualnom interpretacijom cijele scene. Naime, ve}i broj klastera, koji je dobiven segmentiranjem, tj. kori{tenjem ostalih arhitektura s ve}im brojem neurona, nije upu}ivao na nova obilje`ja koja bi se utvrdila

162

Croat. j. for. eng. 31(2010)2

Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network (157–163)

D. Klobu~ar et al.

vizualnom interpretacijom. Testom proporcija testirane su postotne jednakosti vizualne interpretacije i segmentacije. Vjerojatnost da su proporcije jednake jest u rasponu od 0,83 za klaster 2 (jako bolesna stabla) do 0,96 za klaster 6 (zdrava stabla). Nema statisti~ki zna~ajne razlike izme|u vizualne interpretacije i segmentacije analizirane scene S2 (tablica 1).

Zbog varijacija svake scene gotovo je nemogu}e o~ekivati da rezultati segmentacije za svaku scenu u potpunosti odgovaraju rezultatima vizualne interpretacije, odnosno da su segmentacije za sve scene identi~ne. Tim je slijedom iskori{tena struktura neuronske mre`e za scenu S2. Naime, provedena je simulacija ove mre`e nad ulaznim matricama scena: S, S3 i S4. Simulacijom neuronske mre`e (scena S2) na trima ostalim scenama postignuto je pobolj{anje, odnosno segmentiranje scena prema utvr|enim razredima u postupku vizualne interpretacije. Postupak (primjena) simulacije ima i opravdanost u brzom dobivanju rezultata segmentacije, dok treniranje neuronske mre`e zahtijeva odre|eno razdoblje te ponovnu interpretaciju dobivenih klastera. Ovako provedena segmentacija omogu}uje promatranje cijeloga podru~ja i pojedinih scena pod gotovo jednakim uvjetima. Naime, uspostavljen je odnos segmentiranih razreda i korespodentnih boja na cijelom promatranom podru~ju.

Utvr|ivanje o{te}enosti {umâ, njezina prostornoga rasporeda te pra}enje njezina stanja va`na je sastavnica potrajnoga gospodarenja {umama. U provo|enju tih mjera u {to kra}em roku na velikim {umskim povr{inama uspje{no se mogu primijeniti tehnike daljinskih istra`ivanja integralnim pristupom, kori{tenjem dobrih svojstava infracrvenih kolornih aerosnimaka, metoda digitalne analize slike, umjetne inteligencije i vizualne interpretacije scene. Ovim je istra`ivanjem potvr|eno da se samoorganiziraju}a neuronska mre`a mo`e pouzdano primijeniti u otkrivanju o{te}enosti sastojina na infracrvenim kolornim (ICK) aerosnimcima.

Klju~ne rije~i: o{te}enost {uma, infracrveni kolorni aerosnimak, segmentacija, neuronske mre`e, Hrvatska

Authors’ address – Adresa autorâ: Damir Klobu~ar, PhD. e-mail:damir.klobucar@hrsume.hr »Hrvatske {ume« d. o. o. Zagreb Headquaters Zagreb Ljudevita F. Vukotinovi}a 2 HR-10000 Zagreb Prof. Renata Pernar, PhD. e-mail: rpernar@sumfak.hr Asst. Prof. Ante Seletkovi}, PhD. e-mail: aseletkovic@sumfak.hr Mario An~i}, Bsc. e-mail: mancic@sumfak.hr Forestry Faculty of Zagreb University Department of Forest Management and Remote Sensing Sveto{imunska 25 HR-10000 Zagreb

Received (Primljeno): September 13, 2010 Accepted (Prihva}eno): November 8, 2010 Croat. j. for. eng. 31(2010)2

Prof. Sven Lon~ari}, PhD. e-mail:sven.loncaric@fer.hr Asst. Prof. Marko Suba{i}. PhD. e-mail:marko.subasic@fer.hr University of Zagreb Faculty of Electrical Engineering and Computing Department of Electronic Systems and Information Processing Unska 3 HR-10000 Zagreb CROATIA

163

Preliminary note – Prethodno priop}enje

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis of Conjoint Analysis Krzysztof Leszczyñski Abstract – Nacrtak The aim of this paper was the conjoint analysis of wood harvesting processes performed using motor-manual methods. Distribution of the characteristics consisted of nine features, based on the results of multidimensional scaling, which were aggregated into two groups: ergonomic (5) and technological (4). The scope of research was limited to four wood harvesting processes. The cuts were carried out in selected 100-year old spruce stands on the steep terrain (13–30°) in the Beskid ¯ywiecki Mts. The value of utility function was defined on the basis of normalized eigenvectors for the comparison matrix. The weight of the features was defined on the basis of the Tytyk (2001) simplified method of rank aggregation by preserving the maximum values (for ergonomic issues), the Satty method of subjective assessment and the partially determined stochastic factors (for technological issues). The results of the calculations indicate the occurrence of dominant preferences within two groups of factors and their mutual polarization. The results of the total evaluation indicate disappearance of the strong dominance of alternatives.

Key words: Conjoint Analysis, ergonomics, forestry, wood utilization

1. Introduction – Uvod The research concept and methodology of Conjoint Analysis (C.A.) method was derived from conjoint measurements executed in psychology and constitutes a mathematical model (Kuhfeld 2005), where statistic methods are used to set estimation error easily. The theoretical background for C.A. method was developed by R.D. Luce (psychologist and mathematician) and J.W. Tukey (statistician). In the seventies it became one of the main tools for measuring customer preferences and simulation of consumer behavior. It uses the utility theory with the functioning notion of a preference as a relation between multidimensional objects. In order to measure the structure of a preference the following formal model is constructed: Ui = f (u1(i),..., um(i))

(1)

Where: Ui defines total utility of i-th profile, f preference function, uj location of i-th profile in regard of j-th variable.

Croat. j. for. eng. 31(2010)2

The main area of application of C.A. method covers preference analysis, market segmentation and simulation analyses. Research procedure consists of a number of stages comprising: Þ Specification of a research problem, Þ Selection of a dependence model for variables and preferences, Þ Generating profile sets, Þ Defining the measurement scale for dependent variables, Þ Selection of a parameter estimation method, Þ Reliability assessment and interpretation of the results.

Basically, the process of analysis consists in searching for partial utility by decomposition of a known final utility. The main objective is a monotonic transformation of an explanatory variable to the attribute sum equation – independent explanatory variables.

The effect of such action is the objectification of decisions that are made by qualitative and quantitative approach to the possessed information about variables and interdependencies (Forman and Selly 2001). The

165

K. Leszczyñski

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

concept presented here is used both in marketing and finance as well as in natural environment management and technology analysis and optimization (Stampfer and Lexer 2001; Kangas et al. 2002; Bodelschwingh et al. 2005; Heinimann 2007).

2. Aim and scope of the paper – Cilj i svrha rada The aim of this paper was the conjoint analysis of work processes based on nine selected ergonomic and technological aspects. The scope of the research concerned four wood harvesting processes performed by petrol chainsaw operators (30–40 years old) who periodically carried out the following tasks: Þ S1 – felling of trees, Þ S2 – debranching, Þ S3 – wood assortments cross-cutting at the temporary depot, Þ S4 – wood stacking and sorting.

The cuts were carried out in selected 100-year old spruce stands on the steep terrain (13–30°) in the Beskid ¯ywiecki Mts.

3. Materials and methods – Materijal i metode For the purpose of realizing the research assumption adopted in this study, a system of two experimental factors was established, which includes work processes (S1–S4) and a group of parameters: Þ Ergonomic: ü noise level of daily exposure – LEX8h, dB(A); ü equivalent value of mechanical vibration – aweq, ms–2, ü daily concentration of carbon monoxide – Cw, mgm–3, ü Lundqvist burden index of musculoskeletal system – Lunq, ü energy expenditure for working shift – WE8h, kJ. Þ technological: ü relative part on effective time in operating time – K02, ü relative part on effective time in overall working shift time – K07, ü efficiency in operating time – W02, m3/h, ü working time efficiency – W07, m3/h.

The experiment covered a balanced system with nine replications. The value of the daily noise exposure level, the equivalent vibration level and concentration of carbon monoxide were determined with the use of direct (dosimetric) methods. Assessment

166

of the burden on the musculoskeletal system was carried out by using the OWAS method (Karhu et al.

1986), separately for different technological operations. A total result was obtained by calculation of Lundqvist burden index, which accounts for the percent share (pi) of distinguished categories of work processes (i=1–4). The first number denotes to which of the four categories of work processes it belongs (Stampfer 1996).

Work energy expenditure was generally calculated using a simplified Lehman method. However, the correctness of the adopted values for unit expenditure was tested by performing a comparative analysis of the results of pulmonary function measurement and the table values found in many studies (Lehmann 1966, Ronay Slama 1989, Löffler 1990, Lipoglavsek 1997, Sowa et al. 2006). The values of technological indicators were the final result of structural analysis of the work-day. For the purpose of enabling comparison of variables, the values were converted to a time-frame of 480 minutes.

The variables obtained underwent a similarity analysis applying Multi Dimensional Scaling (MDS).

The number of dimensions constituted a basis for the identification of a group of descriptive variables.

The variables obtained are characterized by different ranges and measurement scales resulting from specific features examined, therefore the data have been subjected to standardization using quotient transformations. In this way relative values were obtained from a closed interval [0, 1], where 0 denotes the smallest and 1 the biggest preference. Final utility was calculated using additive model: m

Ui = ∑ wj × c ij j =1

(2)

Where: Ui is general alternative utility i, cij alternative value and taking into account j alternative, wj weight of j criterion.

The adopted method of feature aggregation is analogous to the Analytic Hierarchy Process (AHP) method suggested by Satty (2000). Therefore, it was adopted that the weights of individual attributes constitute normalized values, right eigenvector (w) defined for the highest own value (lmax) of comparison matrix (A): n

A × w = l × w; ∑ wj = 1 j =1

(3)

Where: A = (aij) aji = 1/aij Croat. j. for. eng. 31(2010)2

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

K. Leszczyñski

The domination degree of i over j was defined according to ranks suggested by Satty, where: 1 – denotes equal significance, 3 – moderate advantage, 5 – strong domination, 7 – very strong domination, 9 – extremely strong domination.

In case of ergonomic criteria, the weights were defined according to a schema suggested by Tytyk (2001). This method imposes the need to carry out the assessment of five features of ergonomic factors taking into account features such as pathogenicity, cumulativity, inclination towards synergy, aggressiveness and destructiveness towards the environment. Each of them is given a partial rank (slight – 0.1, low – 0.3, moderate – 0.5, high – 0.7, very high – 0.9). However, the final value of the rank is obtained after the aggregation of ordered partial values xi > xi+1 according to the equation: rk = ∑ 10 i − 1 × xi

(4)

Preserving extreme values is a characteristic feature of this method. This attitude is explained by Tytyk who explains this by the fact that it is difficult to expect that strong influence of one feature (e.g. pathogenicity) will weaken by weaker influence of the other feature (e.g. cumulativity).

The correctness of weight estimation was reviewed using the criterion of Satty (2000), which says that the estimation of values is considered stable when the consistency ratio (CR) is smaller than 0.1: Pe ´ kor =

l max − n n−1

CR =

CI RI

(5)

Fig. 1 Map of variable similarity as the result of Multi Dimensional Scaling Slika 1. Prikaz sli~nosti varijabli na osnovi multidimenzionalnoga skaliranja Ja³owska 2009). Similar values of coordinates of work processes S1 and S4 prove their similarity.

The descriptive statistics analysis (Fig. 2) of ergonomic aspects suggests that the largest burden on

(6)

Where: CI consistency index, RI random index for n-dimensional matrix.

4. Results and Discussion – Rezultati s raspravom In order to define the variable structure, a similarity analysis was carried out using Multi Dimensional Scaling (MDS). In the course of the analysis a significantly smaller number of dimensions were identified in comparison to a number of descriptive variables and the results obtained enabled the picture of similarity (Fig. 1).

The map of variable similarities obtained indicates central location of technological indicators surrounded by dispersed ergonomic elements. The emergence of two dimensions that can be used to describe the analyzed work processes proves the correctness of the adopted variable groups (Leszczyñski and Croat. j. for. eng. 31(2010)2

Fig. 2 Mean and standard error of scaled value of ergonomics attributes Slika 2. Srednja vrijednost i standardna pogre{ka skaliranih vrijednosti ergonomskih pokazatelja 167

K. Leszczyñski

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

Table 1 Normalized eigenvector for paired comparison Tablica 1. Normalizirani svojstveni vektori za usporedbu u parovima Factor – ^imbenik

Ergonomics – Ergonomski

Technological – Tehnolo{ki

Attribute Pokazatelj LEX8h, dB(A) Aweq, ms–2 Cw, mgm–3 Lunq WE8h, kJ K02 K07 W02, m3/h W07, m3/h

S1 0.2333 0.0217 0.0386 0.2254 0.1964 0.1571 0.2084 0.3915 0.4595

musculoskeletal system of the worker occurs in the work process S2.

The physical factors (mechanical vibration and noise) in the work processes S1, S2 and S3 stay at a similar unfavorable level. Technological factor values (Fig. 3) are characterized by larger degree of diversity.

The highest level of efficiency during the working time (W07) was reported on the work process S1 together with the lowest indicator of operating timeuse (K02).

In order to set the preference values for individual factors, mutual comparison matrix A was devel-

Fig. 3 Mean and standard error of scaled value of technological attributes Slika 3. Srednja vrijednost i standardna pogre{ka skaliranih vrijednosti tehnolo{kih pokazatelja 168

Workplace – Radno mjesto S2 S3 0.2340 0.2363 0.0160 0.0258 0.0519 0.0512 0.2069 0.2846 0.2101 0.3035 0.2821 0.2884 0.3629 0.2383 0.2233 0.1805 0.2757 0.1278

S4 0.2964 0.9365 0.8584 0.2831 0.2900 0.2725 0.1905 0.2047 0.1370

CR Konzistentnost 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

oped, characterized by paired consistency aij = aji–1 and global consistency aik ´ akj = aij, for i,j,k ∈ {1…n}.

Using the Schur’s theorem, which says that the sum of module squares of eigenvalues is limited from the top by the Euclidean norm square, a matrix spectrum was set Sp(A) = {l1, l2, l3, …}. Solving the matrix eigenequation, a normalized eigenvector was set for the highest eigenvalues of j-th priority (aspect), which as suggested by Satty (2000) defines the value of preference.

The results obtained (Table 1) suggest similar preference values for LEX8h (0.23–0.29) factor and the domination of the work process S4 because of high value of aweq factor.

The analysis of technological aspects, however, indicates preference for the work process S1 (W07=0.46). The calculated value of CR (Consistency Ratio) confirms the stability of estimated parameters.

The next step of the analysis was to define weights for individual ergonomic and technological aspects. Partial ranks for ergonomic criteria were calculated using the described Tytyk’s method (2001), and individual stages were presented in Table 2.

Satty’s method of ranks was used (2000) for the assessment of technological aspects. The value of dominance of individual features was established in a group of five experts. The assessment values obtained enabled the definition of values of partial weights: K02=0.1250, K07=0.2083, W02=0.2917, W07=0.3750. Taking into consideration the results obtained, the value of partial utilities was calculated (Table 3).

The analysis of Table 3 shows strong polarity of partial utility for the distinguished ergonomic and technological aspects. The highest value due to ergonomic factors can be found in the work process S4 Croat. j. for. eng. 31(2010)2

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

K. Leszczyñski

Table 2 Calculation of partial range for ergonomic issues Tablica 2. Izra~un djelomi~noga raspona ergonomskih pokazatelja Pathogenity – Patogenost Cumulativity – Kumulativnost Tendency to synergy – Te`nja sinergiji Aggressiveness to Environment – Agresivnost prema okoli{u Destructiveness to man – Destruktivnost za ~ovjeka Rang – Rang Normalized rang – Normalizirani rang

LEX8h, dB(A) 0.9 0.7 0.7 0.3 0.3 0.97733 0.21811

Table 3 Partial utility Tablica 3. Djelomi~na korisnost Workplace Radno mjesto S1 S2 S3 S4

Ergonomics Ergonomska 0.1352 0.1362 0.1676 0.5610

Technological Tehnolo{ka 0.3495 0.2794 0.1862 0.1849

(0.561) with the lowest value of technological indicators (0.184). For the work process S1 the highest utilities were determined due to technological aspects (0.349) with the lowest value of ergonomic indicators (0.135).

The next step of the analysis was to define weights for the group of ergonomic and technological criteria. The process of their determination consisted in defining the value of dominance by a group of experts. It was made of five people with higher education diploma who were professionally connected with designing and analysis of wood harvesting technology. The values of assessments were obtained by providing independent answers, i.e. excluding the knowledge of the opinion of the others. The values obtained are characterized by divergence of work process.

In order to handle an unknown random error, the weight estimator of the ergonomic criteria was determined by means of simulation tests that were carried out. The experiment covered 15 tests which covered the process of drawing 1000 elements. The values obtained in this way enabled setting of the target estimators: Mean = 0.2495, Standard Deviation = 0.1135.

Since the density function of ergonomic factor weights was unknown, there was an attempt to determine it using non-parametric methods. The aim of such action was to extend the values of the obtained discrete variables to the whole area of variCroat. j. for. eng. 31(2010)2

aweq, ms–2 0.9 0.9 0.7 0.5 0.3 0.99753 0.22262

Cw, mgm–3 0.9 0.9 0.9 0.7 0.7 0.99977 0.22312

Lunq 0.7 0.5 0.3 0.1 0.1 0.75311 0.16807

WE8h, kJ 0.7 0.5 0.3 0.1 0.1 0.75310 0.16807

able arguments and to smooth the histogram facilitating the interpretation of the results. Therefore, at this analysis stage, estimation of the probability density was carried out setting the kernel estimator Parzen (1961): fn(x) =

1 n × hn

n

 x − Xi   hn 

∑ K  i=1

(7)

Where: K estimation kernel, hn smoothing coefficient, x grid points, Xi value of variable realization.

Probability distribution was established using Gauss’s kernel for the range width adopted according to Silverman’s criterion. The scope of estimation was limited to the interval [0,1]. In order to verify the correctness of calculations, a relative Mean Squared Error of function estimator was developed (relative Mean Squared Error, MSE%) from the equation: MSE% =

E∫ ( f n ( x) − ( x)) 2 dx

∫f

2

( x) dx

´ 100 %

(8)

The value of MSE% amounted to 13.6%, which indicates the permissible estimation error, and hence the correctness of the achieved density. On the basis of the calculations, the estimated density was drawn as well (Fig. 4).

The density function presented in Fig. 4 implies the occurrence of double modal distribution and at the same time the isolation of two groups of experts who represent different work processes. Cutting the density function at local minimum x=0.4, the following distribution was obtained with the following parameters: Mean1 = 0.1929, StandardDeviation1 = 0.0945, Mean2 = 0.5839, StandardDeviation2 = 0.1650.

169

K. Leszczyñski

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

Fig. 4 Results of kernel smoothing estimations of density function Slika 4. Rezultati izjedna~ivanja procjena funkcije gusto}e vjerojatnosti

Fig. 5 Overall utility of workspaces Slika 5. Ukupna korisnost radnih mjesta

Table 4 Relative weight for groups of attributes Tablica 4. Relativne te`ine za grupe pokazatelja

The results of the calculations presented in Fig. 5 show that in each variant the lowest values were established for the work process S3. However, the isolation of the work process S1 – felling of trees (variant A and C), which took place twice, enables its description as dominant with the highest final utility.

It has been observed that the dual problem of defining the weight factors is specific. It results from the occurrence of two independent elements of opposed character of the analyzed object such as human being – machine. The subsystem such as human being is characterized by invariability of nature and the subsystem – a technical object – by many possibilities of adaptation (Tytyk 2001). Therefore, for many years different aspects of work systems have been considered looping for such criteria that would enable the definition of optimal solutions. (Luczak 1993, Sowa 1995, Stampfer 2001).

Variant Ina~ica A B C

Ergonomic, w1 Ergonomski, w1 0.1929 0.5839 0.2495

Technological, w2 Tehnolo{ki, w2 0.8071 0.4161 0.7505

The calculations of ergonomic factor weights imply the occurrence of several correct values which are difficult to get rid of in this stage of work. Therefore, the overall utility value of the analyzed work process was calculated for three variants, concurrently ordering the obtained weight values in Table 4.

The calculated values of final utility were presented in Fig. 5.

The data analysis shows that in the variant A – characterized by over four times higher value of technological criterion over the ergonomic one, the highest value was stated for the work process S1 – felling of trees. In variant B – with a slight domination of ergonomic criterion, the highest value was defined for the work process S4 (wood stacking and sorting), for which the calculated value was 2.3 times higher than for S3 (wood assortments cross-cutting at the temporary depot). The variant C, whose indicators were defined based on stochastic sample realization, is characterized by the smallest gap of utility values with the highest value for S1 and the lowest value for S3.

170

5. Conclusions –Zaklju~ci Þ The results of multidimensional scaling indicate the classification correctness of indicator groups and work process similarity S1 – felling of trees and S4 – wood stacking and sorting with the estimated final utility of 0.29 and 0.28, respectively, in the general variant (C). Þ The analysis of ergonomic aspects indicated that in the work process S2 – debranching, the worker musculoskeletal system was exposed to the highest load.

Croat. j. for. eng. 31(2010)2

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

Þ The greatest efficiency in working time was observed in the work process S1 – felling of trees, with the lowest operating time-use indicator. Þ The kernel estimation results (Parzen) of ergonomic aspect weight density indicate the mixture of two distributions, the first of which describes strong advantage of technological aspects and the second – weak advantage of ergonomic criteria. Þ The analysis of three variants of final utility (A, B, C) enables the classification of the work process S3 – wood assortments cross-cutting at the temporary depot as the least preferable, and the work process S1 – feeling of trees as the most useful.

6. References – Literatura Bodelschwingh, E., Bauer, J., Warkotsch, W., 2005: Potenziale einer effizienten Logistik nutzen. Using potentials of efficient logistics structures. Holz–Zentralblatt 86: p. 1163.

Forman, E., Selly, M. A., 2001: Decision By Objectives.

How To Convince Others That You Are Right. World Scientific Publishing Co. Pte. Ltd., Singapore.

Heinimann, R. H., 2007: Präzisions–Forstwirtschaft – was ist das? Schweizerische Zeitschrift für Forstwesen 158(8): 235–242.

Kangas, J., Kangas, A., Leskinen, P., Pykäläinen, J., 2002: MCDM methods in strategic planning of forestry on state-owned lands in Finland: applications and experiences 10(5): 257–271.

Karhu, O., Kansi, P., Kourinka, I., 1977: OWAS – Ovako Working Posture Analysis System. Autorzy: Correcting working postures in industry. A practical method for analysis. Applied Ergonomics 8(4): 199–201.

Kuhlfeld, W. F., 2005: Marketing Research Methods in SAS. Experimental Design,Choice, Conjoint, and Graphical Techniques. SAS 9.1 Edition TS–722.

Lehmann, G., 1966: Praktyczna fizjologia pracy. 1. Ed. Oryg: Praktische Arbeitsphysiologie. PZWL, Warszawa.

Leszczyñski, K., Ja³owska, M., 2009: Analiza podobieñstwa stanowisk pracy w ciêciach sanitarnych. In: Miêdzynarodowa Konferencja Naukowa: Leœnictwo w górach i regio-

K. Leszczyñski

nach przemys³owych. Kraków – Krynica Zdrój, 21–22.

09.2009 z okazji 60-lecia powo³ania Wydzia³u Leœnego UR w Krakowie. Streszczenia referatów, p.123–125.

Lipoglavsek, M., 1997: Logger's Loads at Work with Power-Saw. In: IUFRO/FAO Seminar on Forest Operations in Himalayan Forests with Special Consideration of Ergonomic and Socio-Economic Problems, 20–23 October editors Heinimann H. R., Sessions J., p. 105–113, Thimphu, Bhutan.

Löffler, H., 1990: Arbeitswissenschaft für studierende der Forstwissenschaft. Second ed., Technische Universität München.

Luczak, H., 1993: Arbeitswissenschaft. Springer–Verlag, Berlin–Heidelberg.

Parzen, E., 1961: On estimation of a probability density function and mode. The Annals of Mathematical Statistics.

Stanford University.

Ronay, E., Slama, O., 1989: Ergonomia a bezpecnost pri praci v lesnom hospodarstve (Ergonomics and safety at forestry work), Priroda, Bratislava.

Saaty, T. L., 2000: The seven pillars of the Analytic Hierarchy Process. In: Multiple Criteria Decision Making in the New Millenium, Lecture Notes in Economics and Mathematical Systems, p. 15–37.

Sowa, J. M., Leszczyñski, K., Szewczyk, G., 2006: Human energy expenditure in late thinning performed in mountain spruce stands. Acta Scientarium Polonorum series Silvarum Colendarum Ratio et Industria Lignaria 5(1): 73–80.

Sowa, J. M., 1995: Badania nad okreœleniem modelu funkcji stanu zagro¿eñ od drgañ pilarek spalinowych w procesie pozyskiwania drewna. Zeszyty Naukowe, Akademia Rolnicza w Krakowie, Rozprawy nr 205.

Stampfer, K., 1996: Determining work load and demand of mechanized forestry work systems. University of Natural Resources and Applied Life Sciences, Vienna.

Stampfer, K., Lexer, M. J., 2001: Multicriteria Evaluation of Thinning Operations on Steep Terrain. In: New trends in wood harvesting with cable systems for sustainable forest management in the mountains, Ossiach, Austria, 1–24 June 2001.

Tytyk, E., 2001: Projektowanie ergonomiczne. WN PWN, Warszawa.

Sa`etak

Ocjena ru~no-strojnih postupaka pridobivanja drva na osnovi objedinjene analize Koncept i metodologija objedinjene (conjoint) analize razvijeni su na osnovi mjerenja i istra`ivanja provedenih u psihologiji i predstavljaju matemati~ki model u kojem se statisti~ke metode primjenjuju za neizravno otkrivanje preferencija. Metoda je na{la {iroku primjenu, ponajprije u marketin{kim istra`ivanjima potro{a~kih sklonosti, a Croat. j. for. eng. 31(2010)2

171

K. Leszczyñski

Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis ... (165–172)

danas nalazi svoje mjesto i u drugim podru~jima. Postupak se objedinjene analize oslanja na teoriju korisnosti i u osnovi se sastoji u tra`enju djelomi~ne korisnosti na temelju rastavljanja poznate kona~ne funkcije korisnosti.

Svrha je ovoga rada bila da se primijeni objedinjena analiza u ocjeni ru~no-strojnih postupaka pridobivanja drva s obzirom na devet odabranih ergonomskih i tehnolo{kih ~imbenika. Predmet su istra`ivanja ~etiri razli~ita procesa koja su na izvr{avanje radnih zadataka u pridobivanju drva povremeno obavljali rukovatelji motornom pilom u dobi 30 – 40 godina. To su: Þ S1 – sje~a, obaranje stabala Þ S2 – ~i{}enje, kresanje grana Þ S3 – izrada sortimenata, prepiljivanje na stovari{tu Þ S4 – slaganje i sortiranje drva.

Sje~a je obavljena u odabranim 100-godi{njim smrekovim sastojinama na strmom terenu (13–30°) u predjelu Beskid ¯ywiecki (gorske, srednje visoke {ume).

S namjerom ostvarivanja pretpostavki i ciljeva istra`ivanja u radu je postavljen sustav s dva eksperimentalna ~imbenika koje sa~injavaju radna mjesta (S1 – S4) i grupa pokazatelja: Þ ergonomski pokazatelji: ü dnevna razina buke (LEX8h, dB(A)) ü ekvivalentna vrijednost mehani~kih vibracija (aweq, ms-2) ü dnevna koncentracija ugljikova monoksida (Cw, mgm-3) ü Lundqvistov indeks optere}enja mi{i}no-ko{tanoga sustava (Lunq) ü energetska potro{nja u radnoj smjeni (WE8h, kJ) Þ tehnolo{ki pokazatelji: ü relativni udio efektivnoga vremena u vremenu rada (KO2) ü relativni udio efektivnoga vremena u trajanju radne smjene (KO7) ü u~inkovitost pri radu (W02, m3/h) ü u~inkovitost u radnom vremenu (W07, m3/h).

Rezultati izjedna~ivanja Parzenove funkcije gusto}e vjerojatnosti upu}uju na pomije{anost dviju distribucija.

Prva od njih opisuje sna`nu prednost tehnolo{kih aspekata (ina~ice A – vi{e od ~etiri puta ve}a vrijednost tehnolo{kih kriterija u odnosu na ergonomske i C – dva puta ve}a vrijednost tehnolo{kih kriterija) i slabu prednost ergonomskih kriterija (u varijanti B). Analiza ukupne korisnosti u trima ina~icama (A, B, C) omogu}uje klasifikaciju radnoga procesa S3 kao najmanje po`eljnoga i radnoga procesa S1 kao najkorisnijega. Rezultati vi{edimenzionalnoga skaliranja upu}uju na ispravnost klasifikacije grupa pokazatelja i sli~nost izme|u radnih procesa S1 i S4, s procijenjenim iznosom kona~ne korisnosti od 0,29, odnosno 0,28 za op}u varijantu (C). Analiza ergonomskih kriterija pokazuje da je u radnom procesu S2 prisutno najve}e optere}enje mi{i}no-ko{tanoga sustava radnika.

Problem odre|ivanja te`ina uspore|ivanih pokazatelja koji se obra|uje u radu je vrlo specifi~an. On proizlazi iz pojave dvaju neovisnih elemenata koji su suprotnoga karaktera s obzirom na analizirane objekte, kao {to su ljudsko bi}e – stroj. Podsustav koji se mo`e promatrati u ljudskom bi}u obilje`ava nepromjenjivost prirode, a podsustav tehni~ki objekt opisuje mnoge mogu}nosti prilagodbe. Zbog toga se razli~iti aspekti radnih sustava razmatraju ve} godinama u potrazi za takvim kriterijima koji bi omogu}ili definiranje optimalnih rje{enja.

Klju~ne rije~i: objedinjena (conjoint) analiza, {umarstvo, ergonomija, pridobivanje drva, u~inkovitost

Author's address – Autorova adresa:

Received (Primljeno): August 06, 2010 Accepted (Prihva}eno): November 26, 2010

172

Leszczyñski Krzysztof, PhD. e-mail: rlleszcz@cyf-kr.edu.pl University of Agriculture in Krakow Faculty of Forestry Department of Forest and Wood Utilization Al. 29 Listopada 46 31–425 Kraków POLAND Croat. j. for. eng. 31(2010)2

Orginal scientific papers – Izvorni znanstveni radovi MARTIN KÜHMAIER, KARL STAMPFER Development of a Multi-Attribute Spatial Decision Support System in Selecting Timber Harvesting Systems .......................................................... 75 Razvoj vi{eatributnoga prostornoga sustava za pomo} pri odlu~ivanju kod odabira sustava pridobivanja drva FRANK THOMAS PURFÜRST Learning Curves of Harvester Operators ............................................ 89 Krivulje usavr{avanja voza~a harvestera ESMAEL GHAJAR, AKBAR NAJAFI, SATTAR EZZATI Skidding Machines Allocation (SMA) Using Fuzzy Set Theory ............................. 99 Planiranje rada skidera (SMA) pomo}u neizrazite teorije YURI GERASIMOV, ALEXANDER SELIVERSTOV Industrial Round-wood Losses Associated with the Harvesting Systems in Russia .............. 111 Gubitci obloga drva pri sustavima pridobivanja drva u Rusiji PETR HRUZA, ILJA VYSKOT Social–Recreation Evaluation of Forest Roads and their Suitability for Trails: Towards a Complex Approach ....................................................... 127 Odre|ivanje dru{tveno-rekreativne kakvo}e {umskih prometnica IGOR POTO^NIK, ANTON POJE Noise Pollution in Forest Environment due to Forest Operations.......................... 137 Zaga|enje {umskog okoli{a bukom pri izvo|enju {umskih radova ALOIS SKOUPY , RADOMIR KLVAC, SEYEDMOHAMMAD HOSSEINI Changes in the External Speed Characteristics of Chainsaw Engines with the Use of Mineral and Vegetable Oils ........................................... 149 Promjene brzinskih zna~ajki motora motornih pila pri uporabi mineralnih ulja i ulja biljnoga podrijetla DAMIR KLOBU^AR, RENATA PERNAR, SVEN LON^ARI], MARKO SUBA[I], ANTE SELETKOVI], MARIO AN^I] Detecting Forest Damage in Cir Aerial Photographs Using a Neural Network ................ 157 Otkrivanje o{te}enosti {uma na ICK aerosnimcima pomo}u neuronske mre`e

Preliminary note – Prethodno priop}enje ~

KRZYSZTOF LESZCZYNSKI Different Evaluations of Motor-Manual Wood Harvesting Processes on the Basis of Conjoint Analysis ............................................. .. 165 Ocjena ru~no-strojnih postupaka pridobivanja drva na osnovi objedinjene analize

ISSN 1845-5719

9 771845 571000