Crojfe 1 2017 kb

Page 1

8

1

2017



Original scientific paper

Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying Models for Oak (Quercus L.) Logs Stored in Piles for Energy Purposes Gernot Erber, Franz Holzleitner, Maximilian Kastner, Karl Stampfer Abstract Natural drying of fuel wood is a feasible option to increase resource efficiency in biomass based energy supply. Meteorological data based drying models are the state-of-the-art to monitor the drying progress. The continuous weighing approach is used to gain data for developing these models. The aim of this study was to investigate the drying performance of oak (Quercus L.) logs stored in piles for energy purposes and assess the effect of model time interval base on the accuracy of meteorological data based drying models. The log pile’s moisture content dropped from initial 38.9% on February 1, 2013 to 24.8% on October 21, 2013, resulting in a total reduction of 14.1%. At the end, moisture content was distributed evenly within the logs and total dry matter losses were low (2.4%). From load and meteorological data, models were developed including 10-minute, hourly, daily and monthly time interval bases. Model performance was validated by comparing the model estimates to the basic observation. Models proved to be very accurate in estimating moisture content change. Compared to the observation, the hourly time interval based model was the most accurate option (mean deviation of 0.10 ±0.13%), while the least accurate option (10-min interval; 1.49 ±1.29%) was still reasonably accurate. Daily and monthly time interval based models are most suitable for use in the forest industry, as they are accurate, while requiring less extensive and detailed input data than models based on hourly or 10-minute time interval. Keywords: meteorological models, drying modeling, fuel wood, natural drying, log wood, woody biomass

1. Introduction Wood is a major source of renewable energy in the European Union. According to Verkerk et al. (2011), about 380 million m3 of round wood and 100 million m3 of fuel wood are harvested per year and the demand is expected to increase between 10% and 35% by 2030. Under current conditions, 744 million m3 per year could be harvested economically, which is about 58% of the total potential. To realize more of the potential, efficiency in fuel wood supply has to increase (Kamimura et al. 2012). Moisture content is a key parameter in fuel wood supply, as it strongly influences the calorific value and transportation economics. Decrease Croat. j. for. eng. 38(2017)1

of moisture content increases the calorific value of the material. In addition, it decreases the amount of water transported per truckload of fuel wood, which is beneficial in terms of payload and shipping volume optimization (Stokes et al. 1987 and 1993, Kaltschmitt et al. 2001, Kofman and Kent 2009, Erber et al. 2016). Natural drying is an efficient and low-cost method to reduce the moisture content of fuel wood (Erber et al. 2012 and 2016). Moisture content of whole trees and logwood is likely to decrease by 20% to 30% during one drying season (Nurmi 1995, Suadicani and Gamborg 1999, Gigler et al. 2000, Nurmi and Hillebrand 2007, Röser et al. 2010, Erber et al. 2012, 2014 and 2016, Raitila et al. 2015). Spring through autumn is the ideal

1


G. Erber et al.

Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

period for drying fuel wood in a stack (Nurmi 1995 and 1999, Höldrich et al. 2006, Nurmi and Hillebrand 2007, Petterson and Nordfjell 2007, Erber et al. 2012 and 2016) and the sooner wood is stored in the year, the faster it will dry to a specific point (Kofman and Kent 2009). Logwood is more advantageous than logging residues, as latter are subject to remarkable dry matter losses caused by biological decomposition processes, which leads to a decrease in net calorific value (Routa et al. 2015a). According to Golser et al. (2005), dry matter losses of logwood usually do not exceed 2% within the first year of storage, while Erber et al. (2012, 2014 and 2016) reported dry matter losses of up to 3.7% when drying beech logs. Contrary, dry matter losses of up to 30% per year (Golser et al. 2005, Routa et al. 2015 a) are reported for logging residues. A further advantage of logs is that they are easier to manipulate during stacking and during chipping. The chip size distribution of logwood chips is more uniform (Nati et al. 2014), which is beneficial in handling at the plant. Neither large shares of oversize particles, which are prone to arching or bridging (Jensen et al. 2004), nor fines, which represent a health issue during storage, are produced. Oak fuelwood is especially valued in many regions in Europe. Throughout history, oaks have been a source of building material and feedstock for manifold uses (Hogan 2012). In Europe, pedunculate oak (Quercus robur L.) and sessile oak (Quercus petraea (Matt.) Liebl.) are the two most common and valuable species, covering about 49,000 and 38,000 km2, respectively, in the European Union. In addition, downy oak (Quercus pubescens Willd.) covers 25,000 km2 (Hemery 2008). In Austria, oak species are common only in the eastern part, where a dry climate is predominant. They represent 2.4% (27.4 million m3) of the stocking volume. The largest shares of oak can be found in the eastern provinces of Lower Austria (4.3% of the province’s stocking volume), Burgenland (19.4% of the province’s stocking volume) and Vienna (37.7% of the province’s stocking volume) (BFW 2016). Drying of fuel wood has been an intensively studied topic during the last five years. A relatively new approach, continuous weighing, has been employed in a series of drying experiments. The basic principle of this approach is to stack biomass e.g. logs, logging residues, etc. into a metal frame, similar to structures used to hold the load on timber trucks. These frames are placed on load cells. Consequently, the drying progress of the biomass can be followed continuously, assuming that any change in the load weight is due to changes in moisture content. This approach was employed in smaller scale by Gigler et al. (2000) and in

2

larger scale by Kofman and Kent (2009), Erber et al. (2012, 2014 and 2016), Raitila et al. (2015) and Routa et al. (2015a, 2015b). Studied tree species and material types covered spruce (logs and logging residues), pine (logs), birch (logs) and beech (logs). Natural drying is a process governed by meteorological parameters (Kröll 1978). Accordingly, these are commonly used for modeling the drying progress of fuel wood. Explaining variables usually include precipitation, air temperature, relative air humidity, solar radiation, wind speed or evapotranspiration potential, as well as storage design related parameters like storage duration, use of cover, tree species, mean diameter and volume or diffusion coefficients (Stokes et al. 1987, Gigler et al. 2000, Erber et al. 2012, 2014 and 2016, Raitila et al. 2015, Routa et al. 2015a and 2015b). In case of the continuous weighing, weight change and thus moisture content change is related to one or more of these variables. Continuous weighing allows gaining high resolution data during the drying progress. As a consequence, models based on raw data require high resolution input meteorological data if they are to be used for estimating a certain pile’s moisture content change over a given period of time. Data availability in general and its resolution in particular can be a problem. If an entrepreneur needs to track the moisture content of a large number of piles, he will not be prepared to install a meteorological station near each pile. Therefore, data from state-run stations or interpolated grid data have to be used. Data on daily means or even on monthly level is much easier to access than data on hourly level or, as in the case of the studies by Erber et al. (2012, 2014 and 2016) on a basic 10-minute interval. For this reason, models covering longer time intervals (daily, monthly) are more attractive and convenient for use in the forest industry. Consequently, it needs to be determined if these models are accurate enough. Data from continuous weighing experiments offers the opportunity to answer this question. Accordingly, the objectives of the present study were, firstly, to develop meteorological data based drying models for estimating the moisture content change of oak logwood piles for different time interval bases and, secondly, to determine the impact of different time interval bases on the accuracy of the result.

2. Materials and methods The storage site was located at Pilgersdorf (province of Burgenland in eastern Austria, 47.4° latitude and 16.3° longitude, 371 m above the Adriatic). The grassy storage site was slightly sloping to the east (~5%), and Croat. j. for. eng. 38(2017)1


Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

skirted by a tree line (about 20 m from the pile) in the south, while open in all other directions. In this region, the mean annual precipitation is 749 mm, while the mean annual temperature is 8.4 °C (ZAMG 2014). A metal frame of 600 kg, similar to those used on timber trucks, was used to stack the logs for the study. Each frame was 2.5 m wide and high, 2.6 to 2.7 m long and built on four load cells (Type HBM 150 kN, Hottinger Baldwin, Germany). Under each of the four frame edges, a load cell was based and the load pressure was distributed evenly by metal plates (30x30 cm) put under the load cells and on top of square edged wooden beams. Ten meters to the west of the frame, a mobile, solar powered meteorological station was set up. All measuring components were installed at a height of 1.5 m to 3.0 m above the ground. Averages of relative air humidity (RH, %), air temperature (TC, ±0.4 °C), wind speed (WS, ±0.3 m s-1), wind direction (WD, ±3°), precipitation (PRain, mm, resolution 0.1 mm h-1), solar radiation (RA, W m-2; light spectrum waveband 300 nm to 1100 nm, ±5%) were recorded by a data logger (CR 3000, Campbell Scientific, Great Britain) at a ten minute interval. Total pile mass (±0.05%) was recorded at the same interval from the load cell data. After an hour, data was transferred to a server via GSM network. Temperature data was converted to Kelvin (TK) as otherwise contradictory effects could have occurred around 0 °C when determining modeling coefficients. Oak was harvested in January 2013 and the frame was stacked with oak logs on January 17, 2013. All of the logs (n=129) were measured in length (average of 3.94 ±0.23 m) and diameter (15.8 ±3.8 cm) Average log volume was 0.04 ±0.03 m3. Total pile volume was 10.5 m3, while total pile mass was 9772.6 kg. For laboratory analysis, 35 randomly selected logs were sampled by cutting four cm thick sample segments 20 cm from the thicker end of the log. Sample segments were weighed immediately at the storage site and sealed into paper bags for transport. Analysis of initial (39.5 ±3.9%) and final moisture content was carried out in accordance with the European standard EN 14774-2. Additionally, initial bone dry density (658.9 ±50.1 kg m-3) was assessed. All moisture content values were reported on wet weight basis. In October 2013, the frame was unloaded and logs were sampled again. This time two four cm thick sample segments were taken per log (n=24). Sample segments were located 50 cm from the upper and lower end of the log, respectively. The pile was chipped immediately and ten samples of about one kg chips each were collected for laboratory analysis. Again, both sample segments and chip samples were immediately Croat. j. for. eng. 38(2017)1

G. Erber et al.

weighed at the storage site and sealed into paper bags for transport. Laboratory analysis was carried out similarly to the procedure at the beginning. Afterwards, initial and final bone dry densities were compared to estimate dry matter loss. Additionally, in-log moisture content distribution was sampled in detail at the end of the experiment. From eight logs, four cm thick sample segments were cut at 25%, 50% and 75% of the total log length. These logs were selected randomly and originated from all sections of the pile, except the undermost and topmost section. Due to technical issues related to the meteorological station, the study period started 13 days after stacking. Nevertheless, pile weight and moisture content were monitored from the stacking on. Therefore, starting weight and moisture content could be determined on February 1, 2013. From this point on, the pile was monitored for 262 days till October 21, 2013. During this period, the drying progress could be tracked via the pile weight change. However, the reason for this change cannot be determined that easily during winter season, as it can either result from drying, rewetting or snow cover buildup. Snowfall amount was estimated under assumption that a rise in pile weight at an air temperature of around 0 °C (273.1 K) is the result of snow deposition on the pile. Accordingly, the rise in weight was divided by the pile surface area. This weight was then converted to rainfall equivalents (PSnow; one mm precipitation per one kg m-2). Nevertheless, both due to snow drift by wind and the inability to accurately capture snow melting in spring, snow remains a source of inaccuracy in load monitoring. Yet it is not unreasonable to assume that the effect of snow related load weight change is also present in spring. Analysis followed an approach successfully employed in former studies (Erber et al. 2012, 2014 and 2016, Routa et al. 2015a and 2015b). Initially, weight change between measurement points was calculated. Next, weight change was converted to moisture content change (MCA), starting from the baseline moisture content assessed in laboratory. After calculation on the basic 10-minute interval level, data was aggregated on hourly, daily and monthly level. Means (MC, WS, TK, RA and RH) and sums (MCA, sum of PRain and PSnow (P)) were calculated. On 10-minute, hourly and daily level, the season factor (SE; defined by winter and summer solstice, spring and fall equinox) was included into the explaining variables dataset. Lastly, the days since storage start (DY) were included as possible explaining variables in the monthly level dataset. After that, multiple linear regression models, which estimate moisture content change (%) on the respective time interval level, were developed. Their

3


G. Erber et al.

Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

common principle is that moisture content change is a function (1) of meteorological parameters and season (1) (10-minute, hourly, daily time interval level) or storage duration (2) (monthly time interval level). Variance analysis and variable correlation checks were performed. Accuracy of the models was assessed by their mean deviation and respective standard deviance from the observed basic 10-minute time interval level curve. For this purpose, estimates were calculated on the respective time interval basis. To enable comparison of the different time interval base estimates to the 10 minute observation level, for hourly, daily and monthly level, missing values between estimates were approximated linearly.

MCA = Intercept + WS + TK + RH + RA + P + SE (1)

MCA = Intercept + WS + TK + RH + RA + P + DY (2)

Where: MCA Moisture content alteration per time interval, % WS Mean wind speed per time interval, m s–1 TK Mean air temperature per time interval, K RH Mean relative air humidity per time interval, % RA Mean solar radiation per time interval, W m–2 P Mean precipitation per time interval, mm SE Season DY Days since storage start, n To prevent any unfavorable results due to use of the models in an improper environment, the valid range of the models was limited by the 5% and 95% quantile of the explaining variables. Respective limits are presented in Table 1. Generally, the models shall only be used for oak logs with bark with a length between 3.50 m and 4.50 m and diameters of 10 cm to 30 cm. Table 1 Valid range of the models is defined by the 5% and 95% quantile of the explaining meteorological parameters Model basis WS, m s–1

TC, °C

RH, %

RA, W m–2

P, mm

10 minutes

0.0–3.7 –2.6–26.4 42.7–99.8

0.0–0.8

0.0–0.1

Hourly

0.0–3.5 –2.6–26.4 42.8–99.8

0.0–0.7

0.0–0.5

Daily

0.3–2.9 –2.1–23.6 60.2–94.2 36.5–319.6 0.0–13.3

Monthly

0.7–1.5

The oak log pile’s moisture content decreased from 38.9% on February 1, 2013 to 24.8% on October 21, 2013, resulting in a total reduction of 14.1%. From March to August the moisture content dropped steadily. Drying then slowed down considerably, and finally remained relatively stable until the end of the observation period. In March, June and July, the highest monthly drying rates (4.1–4.8%) were observed, while the pile’s moisture content increased during rainy May (1.2%) and humid September (1.8%). The lowest moisture content (24.1%) on daily basis was observed on September 8, while the highest moisture content (40.0%) on daily basis was recorded on February 25, 2013. At the end, the mean moisture content of the sample segments (n=24) was 24.1 ± 4.1%, while the mean moisture content of the chip samples (n=10) was 25.2 ± 1.1%. Final pile weight was 7932.1 kg. No dry matter losses were observed, if only the mean of all logs sampled in the beginning was compared to the mean of all logs sampled in the end. If comparison was only limited to logs that were sampled both in the beginning and at the end, a mean dry matter loss of 2.4% was observed. The initial intention had been to sample the same logs in the beginning and at the end, but even though logs had been provided with an ID, not all could be recognized again, as a share of the ID tags came off from the logs. For this reason, a number of substitute logs were selected. Reasonably fitting models could be developed for all time intervals (Tables 3–6). Model statistics are displayed in Table 2 and models are used together with Eq. 1 and 2. SE was only significant on 10 minute and hourly level, while DY was not significant at all. RA and P were significant in all models. It showed that Table 2 Overview of developed oak log drying models

0.4–19.9 68.5–84.9 78.0–257.5 7.0–145.3

WS = wind speed; TC = air temperature; RH = relative air humidity RA = solar radiation; P = precipitation

3. Results Daily mean wind speed was 1.0 ± 0.9 m s–1, while daily mean temperature and relative air humidity

4

were 11.6 °C (284.7 ± 7.8 K) and 77.7 ± 10.9%, respectively. Solar radiation was 168.0 ± 92.0 W m–2 on average. A total sum of 530.9 mm of rainfall was recorded. During the observation period, the largest shares of rainfall were observed in May (135.0 mm), September (86.3 mm) and July (79.3 mm). The estimated amount of snowfall totaled 180.0 mm in rainfall equivalents.

Time interval Std. error 10-minute

0.01

R2 adj.

p-value

Mean deviation from the observation

0.54

<2.2x10–16

1.49±1.29%

–16

Hourly

0.05

0.58

<2.2x10

0.10±0.13%

Daily

0.25

0.71

<2.2x10–16

0.66±0.51%

Monthly

1.17

0.76

–2

1.6x10

0.62±0.49%

Croat. j. for. eng. 38(2017)1


Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

G. Erber et al.

Fig. 1 Observed drying performance (continuous line) and estimated drying performance (dashed line) of the oak log pile during the drying period February to October 2013 for different time interval bases (estimated performance on hourly time interval base was almost perfectly similar to the observation) deviation of the hourly level estimation from the observation was the lowest (0.10 Âą 0.13% in moisture content). Estimates generally did not differ from the observed drying performance to a large degree, except for the 10-minute time interval based estimation. This level also showed the largest single deviation (5.11%) (Fig. 1). Croat. j. for. eng. 38(2017)1

Detailed analysis of in-log moisture content distribution showed that moisture content did not differ significantly between the bottom, middle and top section of the logs. Moisture content did differ significantly (p-value=1.8x10-5) between the logs, but no significant correlation between moisture content and log volume could be observed.

5


G. Erber et al.

Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

Table 3 Model statistics, parameters estimate, Student’s t-test and summarized test statistics of 10-minute time interval oak log drying model Parameter

Estimate

Std. error

t-value

p-value

Intercept

–4.43x10–2

4.55x10–3

–9.74

<2.0x10–16

WS, m s–1

–5.30x10–4

6.24x10–5

–8.50

TK, K

2.04x10–4

1.52x10–5

RH, %

–1.39x10–4

RA, W m–2

Table 6 Model statistics, parameters estimate, Student’s t-test and summarized test statistics of monthly time interval oak log drying model Estimate

Std. error

t-value

p-value

Intercept

–134.80

37.84

–3.56

1.6x10–2

<2.0x10–16

TK, K

4.87x10–1

1.39x10–1

3.51

1.7x10–2

13.46

<2.0x10–16

RA, W m–2

–4.95x10–2

1.34x10–2

–3.71

1.4x10–2

5.97x10–6

–23.31

<2.0x10–16

P, mm

3.62x10–2

9.58x10–3

3.78

1.3x10–2

–2.48x10–2

4.12x10–4

–60.34

<2.0x10–16

P, mm

9.69x10–2

4.92x10–4

197.01

<2.0x10–16

SESpring

–4.66x10–4

2.21x10–4

–2.11

3.4x10–2

–4

–4

–2.60

–2

SESummer SEWinter

–6.26x10

–3

–1.01x10

2.41x10

–4

2.87x10

–3.50

9.5x10

–2

4.7x10

WS = wind speed; TK = air temperature; RH = relative air humidity RA = solar radiation; P = precipitation; SE = Season

Table 4 Model statistics, parameters estimate, Student’s t-test and summarized test statistics of hourly time interval oak log drying model Parameter

Estimate

Std. error

t-value

p-value

Intercept

–3.83x10–1

3.92x10–2

–9.75

<2.0x10–16

TK, K

1.58x10–3

1.33x10–4

11.90

<2.0x10–16

RH, %

–6.65x10–4

5.21x10–5

–12.77

<2.0x10–16

RA, W m–2

–1.46x10–1

3.98x10–3

–36.63

<2.0x10–16

P, mm

-8.68x10–2

1.04x10–3

83.49

<2.0x10–16

SESummer

–5.70x10–3

2.21x10–3

–2.58

1.0x10–2

TK = air temperature; RH = relative air humidity; RA = solar radiation; P = precipitation; SE = Season

Table 5 Model statistics, parameters estimate, Student’s t-test and summarized test statistics of daily time interval oak log drying model Parameter

Estimate

Std. error

t-value

p-value

WS, m s–1

–8.21x10–2

1.77x10–2

–4.64

5.6x10–6

RA, W m–2

–1.15x10–3

2.69x10–4

–4.26

2.9x10–5

P, mm

7.00x10–2

3.24x10–3

21.63

<2.0x10–16

WS = wind speed; RA = solar radiation; P = precipitation

6

Parameter

TK = air temperature; RA = solar radiation; P = precipitation

4. Discussion Natural oak log drying has never before been modeled depending on either meteorological or any other data. The developed models offer the opportunity to estimate the logs drying performance on different time interval bases, ranging from a 10-minute or hourly time interval suitable for scientific purposes to daily and monthly time intervals fit to e.g. entrepreneur’s requirements. The experimental setup did not differ from the setup and procedure in Erber et al. (2012, 2014 and 2016) to a large degree, which is of great benefit in comparing the results. Unique to this experiment were the tree species and partly the modeling time intervals (10 minutes, hourly and monthly interval). Other studies conducted with the same experimental setup confirm the usefulness of the continuous weighing approach (Raitila et al. 2015, Routa et al. 2015b). Moisture contents established from the samples in the end proved to be similar to the moisture content estimated from the continuous weighing. It showed that dry matter loss sampling is a tricky procedure, not only for logging residues (Routa et al. 2015a), but also for logs. If one is not able to compare the same logs, differences between logs and in-log variation can affect the sampling procedure to a certain degree. Thus the established dry matter loss of 2.4% can be considered a reasonable, but nevertheless only rough estimation. The result at least matches dry matter loss rates for log wood found in literature (2% per year; Golser et al. 2005). The total moisture content drop of 14.1% during one drying season indicates the need for a second drying season if one wants to reduce the moisture content to 20% or below. Further, if the starting moisture content was higher than the observed 39.5%, a second drying cycle would be inevitable to reach moisture contents below 30%. Golser et al. (2005) conclude that tree Croat. j. for. eng. 38(2017)1


Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

species with thick bark (like oak) dry slower than those with easily water permeable, thin bark. Oak dried slower than coniferous and other deciduous species. Another explanation could be an effect observed by McMinn (1986). While diffuse-porous and coniferous species moisture content could be substantially reduced by leaf-seasoning (drying trees in foliage), this was not the case with ring-porous species like oak. The author attributed this effect to the fact that ring-porous species have a proportionally smaller area of active conductive tissue. This means that the proportion of sapwood to heartwood is smaller than in diffuse-porous and coniferous species and thus less moisture can be removed within the same period. Other authors came to the same conclusion (Johnson and Zingg 1969, Clark and Phillips 1972, Simpson 1991). The longer the time interval length, the fewer variables constituted the explaining variable dataset. While during short time interval modeling (10 minutes, hourly) almost all variables in the dataset were of significant impact, the number of explaining variables reduced to three for the long term interval models (daily, monthly). It can be assumed that during longer intervals the more slowly altering variables gain importance, while the impact of those responsible for short time change declines. Contrary to other studies (Filbakk et al. 2011), no impact of the storage duration (DY) could be observed. This may indicate that oak has not yet reached the low moisture content level, where drying slows down due to stronger bonds between wood and water. This effect was observed in Erber et al. (2016), where beech dried the slower, the lower the moisture content was on the day before. Among others, Klepac et al. (2008) report significant effects of season on the drying of loblolly pine (Pinus taeda L.). In summer, trees dried about 50% faster than in autumn and winter. In the present study, seasonal effects were discovered for 10-minute and hourly time intervals but not for daily interval models. Here the seasonal effects may cover some of the otherwise unexplained variation in 10-minute and hourly interval data. In daily data, the variation is probably lost due to averaging effects. Datasets were checked for correlations between variables and, especially in short intervals datasets, some were discovered. As their consideration did not improve the models performance and would unnecessarily inflate the models beyond usability, they were excluded from the models. In general, the employed datasets are in line with other studies (Stokes et al. 1987, Filbakk et al. 2011, Erber et al. 2016). Estimation proved to be the most accurate on hourly level, but an extensive and detailed set of basic data is still required for its use. However, daily and monthly time interval based models were almost as accurate Croat. j. for. eng. 38(2017)1

G. Erber et al.

as the hourly time interval based model, but they require a smaller number of input parameters and less detailed data. Accordingly, it can be concluded that these models are suitable for use in daily business in the forest industry, while the use of hourly time interval base models is limited to scientific purposes. In-log moisture content distribution at the end of the drying period matches research results of NeuĂ&#x;er et al. (1977). A slight decline in mean moisture content from bottom to top was observed. This effect is likely to correspond to the respective decrease in log diameter. Naturally, logs at the top of a pile are more exposed to wind, rain and solar radiation. To avoid edge effects, sample piles were selected from neither the undermost nor the topmost row. As discovered in previous studies (Erber et al. 2012, 2014 and 2016), the major weakness of the actual setup is the inability to assess the behavior of snow cover appropriately. Since the drying period of this experiment touched the snow season to a lesser degree than experiments with a run time of more than one year, this probably did have less effect on the results. To better fit to real life conditions, future studies will also have to deal with multi-species piles and different climate conditions.

5. Conclusions In the present study, meteorological data based drying models could be developed for oak logs stored in piles. Model time interval bases included 10-minute hourly, daily and monthly intervals. Models proved to be very accurate in estimating moisture content change and in comparison with the observed drying performance, regardless of the time interval base. Daily and monthly time interval level based models are most suitable for use in the forest industry, as they are accurate, while requiring less extensive and detailed input data than models on hourly or 10-minute time interval base. The latter are more suitable for scientific purposes. The present study confirmed that dry matter losses of fuel wood stored as logs may be much lower than those recorded for logging residues and whole trees and that natural drying is an effective method for reducing fuel wood moisture content. However, oak logs shall generally be subject to a drying period of two successive summer seasons to achieve moisture contents of around 20%.

Acknowledgements This work was funded by the INFRES-project (European Union Seventh Framework Programme (FP7/2012-2015) under grant agreement n°311881).

7


G. Erber et al.

Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

6. References BFW – Austrian Federal Forest Office, 2016: Austrian Forest Inventory Database [online]. Retrieved on 2nd February 2016 from http://bfw.ac.at/rz/wi.home. Clark, A., Phillips, D.R., 1972: Slash pine logs lose weight in storage. Southern Lumberman 225(2795): 15–17. Erber, G., Kanzian, C., Stampfer, K., 2012: Predicting moisture content in a pine logwood pile for energy purposes. Silva Fennica 46(4): 555–567. Erber, G., Kanzian, C., Stampfer, K., 2016: Modelling natural drying of European beech (Fagus sylvatica L.) logs for energy based on meteorological data. Scandinavian Journal of Forest Research 31(3): 294–301.

Kofman, P.D., Kent, T., 2009: Long term storage and seasoning of conifer energy wood. Coford Connects Harvesting/ Transportation 20, 4 p. Klepac, J., Rummer, B., Seixas, F., 2008: Seasonal effect on moisture loss of loblolly pine. In: Proceedings of the 31st Council on Forest Engineering Annual Meeting at Charleston, South Carolina. 9 p. Kröll, K., 1978: Trocknungstechnik: Trockner und Trocknungsverfahren [Drying technology: dryers and drying procedures]. Springer Verlag, Berlin-Heidelberg-New York, 654 p. McMinn, J.W., 1986: Transpirational drying of red oaks, sweetgum, and yellow-poplar in the Upper Piedmont of Georgia. Forest Products Journal 36(3): 25–27.

Erber, G., Routa, J., Kolström, M., Kanzian, C., Sikanen, L., Stampfer, K., 2014: Comparing two different approaches in modeling small diameter energy wood drying in logwood piles. Croatian Journal of Forest Engineering 35(1): 15–22.

Nati, C., Eliasson, L., Spinelli, R., 2014: Effect of chipper type, biomass type and blade wear on productivity, fuel consumption and product quality. Croatian Journal of Forest Engineering 35(1): 1–7.

Filbakk, T., Hoibo, O., Nurmi, J., 2011: Modelling natural drying efficiency in covered and uncovered piles of broadleaf trees for energy use. Biomass and Bioenergy 35(8): 454–463.

Neußer, H., Strobach, D., Krames, U., Silbernagl, H., Schedl, C., 1977: Lagerung von Holz in Rinde, Teil 1–3 [Storing of wood in bark, part 1–3]. Österreichisches Holzforschungsinstitut. 40 p.

Gigler, J., Van Loon, W.K.P., Van Der Berg, J.V., Sonneveld, C., Meerdink, G., 2000: Natural wind drying of willow stems. Biomass and Bioenergy 19(3): 153–163. Golser, M., Pichler, W., Hader, F., 2005: Energieholztrocknung [Fuel wood drying]. Vienna (Austria): Holzforschung Austria, 139 p. Hemery, G.E., 2008: Forest management and silvicultural responses to projected climate change impacts on European broadleaved trees and forests. International Forestry Review 10(4): 591–607. Hogan, C., 2012: Oak [online]. Retrieved on 2nd February 2016 from http://www.eoearth.org/view/article/161730. Höldrich, A., Hartmann, H., Decker, T., Reisinger, K., Sommer, W., Schardt, M., Wittkopf, S., Ohrner, G., 2006: Rationelle Scheitholzbereitstellungsverfahren [Efficient fire wood supply methods]. Berichte aus dem TFZ 11. Technologieund Förderzentrum im Kompetenzzentrum für Nachwachsende Rohstoffe, 279 p. Jensen, P.D., Mattsson, J.E., Kofman, P.D., Klausner, A., 2004: Tendency of wood fuels from whole trees, logging residues and roundwood to bridge over openings. Biomass and Bioenergy 26(2): 107–113. Johnson, N.E., Zingg J.G., 1969: Transpirational drying of Douglas-Fir: Effect on log moisture content and insect attack. Journal of Forestry 67(11): 816–819. Kaltschmitt, M., Hartmann, H., Hofbauer, H., 2001: Energie aus Biomasse: Grundlagen, Techniken und Verfahren [Energy from biomass: basics, techniques and procedures]. Berlin, Springer, 770 p. Kamimura, K., Kuboyama, H., Yamamoto, K., 2012: Wood biomass supply costs and potential for biomass energy plants in Japan. Biomass and Bioenergy 36: 107–115.

8

Nurmi, J., Hillebrand, K., 2007: The characteristics of wholetree fuel stocks from silvicultural cleanings and thinnings. Biomass and Bioenergy 31(6): 381–392. Nurmi, J., 1995: The effect of whole-tree-storage on the fuelwood properties of short-rotation Salix crops. Biomass and Bioenergy 8(4): 245–249. Nurmi, J., 1999: The storing of logging residue for fuel. Biomass and Bioenergy 17(1): 41–47. Petterson, M., Nordfjell, T., 2007: Fuel quality changes during seasonal storage of compacted logging residues and young trees. Biomass and Bioenergy 31(11–12): 782–792. Raitila, J., Heiskanen, V-P., Routa, J., Kolström, M., Sikanen, L., 2015: Comparison of moisture prediction models for stacked fuelwood. Bioenergy Research 8(4): 1896–1905. Röser, D., Erkkilä, A., Mola-Yudego, B., Sikanen, L., Prinz, R., Heikkinen, A., Kaipainen, H., Oravainen, H., Hillebrand, K., Emer, B., Väätäinen, K., 2010: Natural drying methods to promote fuel quality enhancement of small energywood stems. Working papers of the Finnish Forest Research Institute 186, 60 p. Routa, J., Kolström, M., Ruotsalainen, J., Sikanen, L., 2015a: Precision measurement of forest harvesting residue moisture change and dry matter losses by constant weight monitoring. International Journal of Forest Engineering 26(1): 71–83. Routa, J., Kolström, M., Ruotsalainen, J., Sikanen, L., 2015b: Validation of prediction models for estimating the moisture content of small diameter stem wood. Croatian Journal of Forest Engineering 36(2): 283–291. Simpson, W.T., 1991: Dry kiln operator’s manual. Agriculture handbook (United States. Dept. of Agriculture) 188. Croat. j. for. eng. 38(2017)1


Impact of Different Time Interval Bases on the Accuracy of Meteorological Data Based Drying ... (1–9)

G. Erber et al.

Madison, Wisconsin, U.S. Dept. of Agriculture, Forest Service, Forest Products Laboratory, 274 p.

on a poor sandy soil and a rich loamy soil. Biomass and Bioenergy 17(3): 199–208.

Stokes, B.J., McDonald, T.P., Kelley, T., 1993: Transpirational drying and costs for transporting woody biomass – a prelimary review. In: Hudson, J.B ., Gingras J.F., Twaddle A., editors. IEA/BA task IX, activity 6: transport and handling. New Brunswick: IEA. 76–91.

Verkerk, P., Anttila, P., Eggers, J., Lindner, M., Asikainen, A., 2011: The realisable potential supply of woody biomass from forests in the European Union. Forest Ecology and Management 261(11): 2007–2015.

Stokes, B.J., Watson, W.F., Miller, D.E., 1987: Transpirational drying of energy wood. ASAE paper No. 87-1530. St. Joseph, MI: American Society of Agricultural Engineers, 13 p. Suadicani, K., Gamborg, C., 1999: Fuel quality of whole-tree chips from freshly felled and summer dried Norway spruce

ZAMG – Austrian Central Institute for Meteorology and Geodynamics, 2002: Klimadaten von Österreich 1971–2000 [Austrian climate data 1971-2000] [online]. Retrieved on 2nd February 2016 from http://www.zamg.ac.at/fix/klima/oe7100/klima2000/klimadaten_oesterreich_1971_frame1. htm.

Authors’ address:

Received: February 9, 2016 Accepted: June 5, 2016 Croat. j. for. eng. 38(2017)1

Gernot Erber, PhD.* e-mail: gernot.erber@boku.ac.at Franz Holzleitner, PhD. e-mail: franz.holzleitner@boku.ac.at Maximilian Kastner, MSc. e-mail: maximilian.kastner@boku.ac.at Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Life Sciences, Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Strasse 82 A-1190 Vienna AUSTRIA * Corresponding author

9



Original scientific paper

Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area Pavel Anisimov, Evgenij Onuchin, Marija Vishnevskaja Abstract Natural drying is practiced to improve the quality of fuel wood. The purpose of this study is to obtain the mathematical model of natural drying of whole trees stacked in bunches in the cutting area. To do this, an experimental study of natural drying of whole trees with different diameter was carried out from March to October 2015. Whole pine and birch trees were sorted according to their diameter and placed in bunches in the cutting area. During the next 7 months, the moisture content of wood and ambient parameters were constantly measured. As a result, the mathematical model describing the change in moisture content of the trees in bunches in the cutting area in the process of natural drying was developed. The resulting mathematical model makes it possible to determine the change in the average moisture content of wood depending on the following factors: diameter of a tree trunk, cumulative precipitation, relative humidity and ambient temperature, average wind speed and duration of natural drying. The developed mathematical model allows to predict changes in moisture content of wood in the process of natural drying. During the experiment, in the process of natural drying from March to October, it decreased from 52% to 27% on average. The results can be used to improve the efficiency of fuel wood production. The proposed mathematical model can be used in practice for predicting the outdoor wood moisture content change of whole trees packed in stacks for natural drying, and therefore, to determine the optimum drying time. Keywords: natural drying, drying modeling, moisture content, fuel wood drying, energy forest

1. Introduction It is a known fact that high moisture content of fuel reduces its calorific value (Erber et al. 2014) and the efficiency of power equipment (Anisimov and Onuchin 2013). This also applies to the direct burning of fuel wood, combustion technology of wood from the intermediate gasification (Rajvanshi 1986), and to the production of wood technology for liquid fuel (Fagernäs et al. 2010). In order to improve the quality of fuel wood, that is to improve the calorific value and decrease the moisture content, natural atmospheric drying is practiced. Logging residues (Pettersson and Nordfjell 2007), as well as fuel wood in the form of whole trees (Kundas 2008) are placed in bunches or stacks for summer drying and shedding greenery. An important issue is to determine the optimal duration of natural drying of fuel wood (Kim and Murphy 2013). To this end, mathematical drying models are developed to predict changes in the moisture content Croat. j. for. eng. 38(2017)1

of wood (Erber et al. 2014). In studies (Routa et al. 2014, Kim and Murphy 2013, RĂśser et al. 2010, Pettersson and Nordfjell 2007), it is noted that the dynamics of round wood moisture content change under natural atmospheric drying in the stack depends on the following factors: average diameter of the tree trunk, ambient temperature and relative humidity, cumulative precipitation, presence of bark, geometrical dimensions of the pile and the density of packing in the stack, wind velocity, as well as the availability of the stack protection from precipitation. Numerous studies indicate increased moisture content (rewetting) when timber is stored in the open air in autumn and spring, during rain and during snowmelt (Routa et al. 2014, Gigler et al. 2000, RĂśser et al. 2010). While some studies (Raitila et al. 2015, Kim and Murphy 2013) report no rewetting in summer, other studies indicate that there was a slight rewetting of wood in covered and non-covered stacks during heavy rains in summer (Filbakk et al. 2011, Erber et al. 2014, Gigler et al. 2000). It

11


P. Anisimov et al.

Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17)

should be mentioned that all papers report about the reduction of moisture content of wood after summer natural drying. By the end of experiments, the moisture of wood was 1.5–2 times lower than the moisture of freshly felled trees despite the periodic rewetting of wood due to rains (Filbakk et al. 2011, Erber et al. 2014, Gigler et al. 2000). In order to determine the timing and duration of the natural drying of fuel wood, its logistics should be taken into account. The consumption of fuel wood chips for heat and electric power has a distinct seasonal pattern with peak consumption in winter months. At the same time, the grinding of frozen wood chips requires higher energy consumption, as it has high hardness (Golovkov et al. 1987, Shelgunov 1982). Therefore, the bulk of production of fuel wood chips is economically feasible to implement in the warmer months before the start of the heating season and (or) before the onset of freezing temperatures of the outside air. Forestry also often has seasonal character, as weak and waterlogged soils (Poršinsky et al. 2006), for example, in Russia, occupy a large part of the forest area. Due to the weak soil on the territory of Russia, the bulk of wood is harvested in the winter months (Suhanov 2008). Thus, the laws governing natural drying of fuel wood that was cut down in winter during the spring – summer – early autumn are of greatest practical interest. Mathematical models of natural drying of round wood are based on empirical data. Raitila et al. (2015) note that it is essential to complete the database models for specific drying conditions as soon as possible. The majority of papers are devoted to the natural drying of logs and logging waste in stockpiles. However, it is necessary to conduct further research of natural drying of whole trees of various diameters in bunches in the cutting area. Therefore, the aim of this study is to get a mathematical model based on the experimental data that describes the change in the moisture content of whole pine and birch trees during their natural drying in the stack in the cutting area. The resulting regression equation should determine the dependence of the average moisture content of wood in a bunch from the diameter of wood, monthly sum of precipitation and evaporation.

2. Materials and methods 2.1 Mathematical modeling In order to predict the change in the average moisture content of the wood exposed to the natural atmospheric drying, we propose eq. (1), which is a modification of the equations proposed in other works (Raitila et al. 2015, Kim and Murphy 2013). Eq. (1)

12

takes into account the effects of temperature, relative humidity, and rainfall on the natural drying of wood:

wi+1 = wi + k1 ×

∑ P + k2 × ∑ E

(1)

Where: wi+1 target relative humidity during natural wood drying at time ti+1, % i unit of time, for example, one week, two weeks or one month wi average relative humidity of timber at time ti, % ΣP the total amount of liquid precipitation for the period (ti+1 – ti), mm ΣE the total direct evaporation of water from the surface for the period (ti+1 – ti), mm k1, k2 coefficients obtained experimentally The total direct evaporation of water from the surface ΣE for the period (ti+1 – ti), mm, is given by (2), which is based on the equation for determining the amount of evaporated water per hour from an open water surface of the proposed drying laboratory − AllRussian Thermal Engineering Institute (Moscow) and is given in reference (Bogoslovskij et al. 1992): 3

∑ E = h × ( 7.4 × ( 0.022 + 0.017u) × (ps − pn )) × 101 Pbar

(2)

Where: h the number of hours in the period under review (ti+1 − ti) u average wind speed measured at stack midheight level ps pressure of saturated water vapor in air at a temperature equal to that of water (in this case, the water temperature is equal to ambient temperature), kPa pn partial pressure of water vapor in the air, kPa pbar barometric pressure, kPa Saturated water vapor pressure over pure water surface (in the temperature range 0–83 °C) can be found in the tables or by the equation proposed in (Fil'nej 1966):

ps = 0.001 × 10

658+10.2×t 236+t

, psi

(3)

Where: t ambient temperature for a period of time (ti+1 − ti), °C The partial pressure of water vapor in the air is found by equation:

pn = ps ×

j ,kPa 100

(4)

Where: j average relative humidity of ambient air for a period of time (ti+1 − ti), % Croat. j. for. eng. 38(2017)1


Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17)

The coefficients k1, k2 have been obtained by experiment. They take into account other parameters influencing the process of natural drying of wood, such as the average diameter of a tree trunk in a pack of trees, wood species, type of fuel wood (whole trees, forest residues, non-debarked, partially debarked, etc.). Additionally, the coefficients k1, k2 consider storage conditions (dimensions of the pack or stack, the density of stacking packs or stacks, cover of a stack from above against rain, etc.).

2.2 Experimental layout In order to estimate values of k1, k2 coefficients of eq. (1) and to determine the regression equations defining the dependence of the coefficients k1, k2 from the average diameter of the tree trunk in a stack, an experiment was conducted with natural drying of whole trees of pine and birch (Fig. 1) arranged in a specific way in the cutting area. The study was conducted in the territory of Training and Experimental Forestry »Volga State University of Technology« in Medvedev’s district of the Republic of Mari El (56.477°N; 47.861°E). From whole trees felled at the logging site, 4 bunches (3 bunches of pine trees, sorted by diameter, and 1 bunch of birch trees, not sorted by diameter) were formed. The volume of a bunch and method of stacking bunches of trees is determined by the employed timber harvesting technology for energy use, which consists of two stages. In the first stage, the trees are cut and stacked in bunches using a feller-buncher. The number of trees in the stack is determined by the reach of the boom. In the second stage, after summer natural drying, stacks of whole trees are ground into chips in the cutting area by using a mobile wood chipper with a storage container.

Fig. 1 Way of laying the stacks of whole trees (fuel wood) for natural drying in a cutting area Croat. j. for. eng. 38(2017)1

P. Anisimov et al.

Table 1 Parameters of stacks of the whole trees of pine laid in a cutting area for natural drying Parameter\No bunch of tree

1 (pine) 2 (pine) 3 (pine) 4 (birch)

Average diameter of tree trunks in a bunch, cm

7

11

15

14

Average height of a tree, m

8

10.5

14.5

13

Number of trees in a bunch, pcs.

21

18

15

25

0.65

1.9

3.7

5

1

1.1

1.1

1.3

Wood volume in a bunch, density, m3 Average height of a bunch, m

Trees of approximately the same diameter were placed in one pack. Parameters of tree packs are shown in Table 1. The diameter mentioned is the diameter at half tree height. The trees were entirely placed in stacks. Stacks were not covered. Each stack of trees was placed on a pad on the ground directly in the cutting area, as shown in Fig. 1. Pine trees and birch trees were felled with fellerbuncher and stacked in bunches at the logging site in early March 2015 in the open area after cutting. The bunches were left in the cutting area for natural drying of wood and shedding of pine needles and part of the bark till the end of September. During the experiment, on the 1st or on the 2nd day of each month from April to October 2015, moisture content of the wood was measured in accordance with GOST 17231-78 and ISO 4470-81. The moisture content of green wood was determined by weight method in early March and at the end of the experiment on October 1st. The minimum amount of samples for determining wood moisture is chosen in accordance with GOST 16483.0-89. The minimum amount of samples providing 5% relative accuracy of the sample mean with confidence coefficient 0.95 is 6 pieces. Consequently, it was necessary to take 6 trees from each bunch for measuring the moisture. In accordance with GOST Р 54217-2010, each bunch of trees had 3 levels of height, and 2 trees were taken from each level for the experiment. The samples were round cross sections of 10–15 mm along fiber. Four cross sections were made from each tree. A total of 24 samples were made from each bunch. Samples were repeatedly weighed at an accuracy of less than 0.1 grams before and after drying in an electric muffle at 103 °C. The drying was practiced in electric muffle until the constant mass of samples was reached. Thus the mass of moisture in samples was determined by measuring it before and after drying.

13


P. Anisimov et al.

Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17)

3. Results

Moisture content of wood in the process of natural drying in the experiment on the 1st or the 2nd day of each month from April to September was measured using an electrical moisture meter »GANN Hydromette НТ 85 Т«. The following electrodes were used: Drive-in Electrode M 20 with measuring pins 23 mm long, Ram-in Electrode M 18 with pins 60 mm long and Deep Electrod M 21–100 with pins 100 mm long. There were 6 samples. There were 4 sections for the determination on each tree in accordance with ISO 4470-81. The sections were determined in a random manner at the distance no less than 0.5 m from the butt end. The measuring at each section was conducted at different depth. In each stack, there were 24 sections. The first measurement was taken at the surface of the body. Bark and cambium were taken away and electrodes were placed into the timber. For the second measurement, the blind hole for electrodes was bored out at the depth of 1/4 of body diameter. And for the third measurement, the blind hole was 2/5 of body diameter. After measuring, the blind holes were filled to prevent water penetration. The arithmetical mean of 3 measurements was accepted as moisture content of the section at the local depth. As a result, 216 measurements were taken. In the course of the experiment, the environmental conditions, affecting the drying process at the experimental site, were recorded. These included the amount of precipitation in the form of rain, relative humidity, air temperature and wind speed. Rainfall was measured using a M-99 field rain gauge. The average relative air humidity was recorded using the thermal hygrometer. Wind speed was measured using an anemometer placed at mid stack height level.

The recorded environmental parameters were reduced to monthly averages, while all the other parameters were calculated according to the eq. (2–4). All environmental parameters are presented in Table 2. The cumulative precipitation ΣP, in mm, for the period of time (ti+1 – ti) in the experiment is the total amount of rain fallen for ith month of trees drying. The total direct evaporation of water from the surface for the ith month was taken as ΣE both in the experiment and simulation. In the experiment and in the mathematical model, t was the average ambient temperature measured near the stacks of trees. Wind speed was measured at mid-height of the stack, at 0.55 meters height from the ground. The results of measuring of moisture content of wood wi in the process of air seasoning are shown in charts 1–4 of Fig. 2. Individual points on the graphs, indicating experimental values of moisture content, were calculated using eq. (1). Deviation of the theoretical values calculated by the eq. (1) from the experimental values was less than 5%. Fig. 4 includes an additional dotted line. It represents the estimated birch moisture content calculated by eq. (1) with coefficients k1, k2 (eq. 5, 6) for pine. Substitution of experimental coefficients k1, k2 by those obtained from pine experiments leads to an increased deviation of up to 10%, which confirms the influence of tree species on the drying process. Regression coefficients k1, k2 in eq. (1) were determined by the method of least squares, i.e. by minimizing the sum of squared timber moisture deviations from the measured values calculated using the regres-

Table 2 Environmental parameters in the process of natural drying Partial pressure Average speed of Total direct evaporation Average relative Total amount of Saturated of water vapor movement of air, blowing of water from the humidity of liquid precipitation pressure in the air a stack of trees surface ambient air SP, mm ps, kPa un, kPa J, mps SE, mm j, %

i

Month

Average air temperature t, °C

0

March

–2.1

1

April

3.95

74

66

0.81

0.6

1.2

47.9

2

May

16

57

22.1

1.82

1.04

1

169.3

3

June

18.75

66.5

54.2

2.17

1.44

0.8

139

4

July

17.1

90

138.7

1.95

1.76

1.3

48

5

August

15.35

75.4

64.1

1.74

1.32

1

9.3

6

September

14.5

65

30.8

1.65

1.07

0.9

115.9

Total

375.8

613.5

14

Croat. j. for. eng. 38(2017)1


Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17)

P. Anisimov et al.

Fig. 2 Experimental and theoretical (eq. 1) values of moisture content of wood sion model (1). The experimental values of the coefficients k1, k2 of the mathematical model (1) are shown in Table 3. Table 3 Experimental coefficients k1, k2 of mathematical model (1) No bunches of trees

k1

k2

1

0.1226

–0.1219

2

0.113

–0.1092

3

0.1012

–0.0966

4 (birch)

0.076

–0.0802

Croat. j. for. eng. 38(2017)1

As it can be seen from the graphs in Fig. 2, natural drying process is greatly influenced by the average diameter of the tree trunk. Approximation of the values of the coefficients k1, k2 for drying pine obtained experimentally were obtained by eq. (5, 6) and determine the dependence of the coefficients k1, k2 from wood diameter:

k1 = – 0.00275 ´ D + 0.1426 R2 = (0.98)

(5)

k2 = – 0.003125 ´ D + 0.1437 R2 = (0.98)

(6)

Where: D average diameter of tree trunks in a bunch, cm reliability of approximation R2

15


P. Anisimov et al.

Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17)

4. Discussion The maximum decrease in wood moisture for all stored timber was observed in the first stack with medium-diameter timber of 7 cm. From March to September, the relative humidity decreased from 53.5% to 23.7%. The larger the diameter of trees, the less the moisture content decreased. The third stack of wood with a diameter of 0.15 m dried to a moisture content of 27.8% from the initial 50.1%. Rewetting of all stacks was observed in July, when precipitation ΣR mm is significantly (3 times) higher than the theoretical total evaporation from the surface of water ΣE, mm. In July 2015, precipitation was up to 190% of the monthly norm. In other months, there was a decrease of moisture content. Thus, for freely stacked stacks in the environment, where ΣP does not exceed ΣE, it is not necessary to cover the trees. After 7 months from the start of the experiment, the bark loss was insignificant and it was about 6%. Pine trees lost 95% of needles and the bark became crisp and started to peel. Since tree stacks were placed directly on the ground, the two trees at the bottom are a kind of basis. Thus their moisture content was not taken into account when determining the average moisture content of stacks. The proposed mathematical model can be used in practice to predict the changes in the moisture content of whole trees, stacked in bunches for natural drying in the open air. Predicting the moisture content, the optimal duration of natural drying can be determined. For predicting the moisture content of round logs and whole trees in natural drying process on the basis of the proposed mathematical model, statistical data on the climate of a particular region can be used, for example from the Scientific and Applied Reference Book on the USSR Climate, long-term data, 1992. Naturally, there are limitations and conditions for the correct use of the research results. The proposed mathematical model to determine the moisture content of wood during drying whole trees stacked in a stack outdoors is only valid for the warm season with positive average daily temperature of the ambient air, when evapotranspiration ΣE for the entire drying period exceeds the total precipitation in the form of rain ΣR. Regression dependence (5, 6) for the determination of the coefficients k1, k2 is applicable to pine trees, whole or sawn in half, with a height of 7–16 meters stacked in the open air in a stack as shown in Fig. 1, with no protection from the weather. For the estimation of k1 and k2 coefficients, the average diameter at half tree height should be used. The mathematical

16

model can be used for natural drying of trees with an average diameter between 5 and 18 cm, an average monthly ambient temperature between +3 and +35 °C and an average wind speed of 0.1 to 3.5 mps. Concerning the optimum drying period, for the climatic conditions of this experiment, it can be concluded that it ends either at the end of June or in late September (Fig. 2).

5. Conclusions A mathematical model was developed describing the moisture content change of the pine and birch trees during natural drying in a stack in the cutting area. The alteration of the average moisture content of trees stacked in a bunch depends on the diameter of the trunk, the amount of liquid precipitation, relative humidity and ambient temperature, average wind speed and length of the drying period. Deviation of the theoretical values, calculated using the proposed mathematical model, from the experimental data was less than 5%, confirming the adequacy of the proposed regression model (1) and of the experimental coefficients k1, k2 (Table 3, and Eq. 5, 6).

Acknowledgments The authors would like to thank the staff of the Training and Experimental Forestry »Volga State University of Technology« for their help with the experiment.

6. References Anisimov, P.N., Onuchin, E.M., 2013: Modelling of the energy supply system of mobile technological lines for the production of dry fuel wood chips with the partial usage of the producible biogenic fuel. Polythematic online scientific journal of Kuban State Agrarian University 89: 518–530. Bogoslovskij, V.N., Pirumov, A.I., Posohin, V.N., 1992: Ventilation and air conditioning. Strojizdat, 319 p. CEN/TS 14778-1:2005, 2005: Solid biofuels. Sampling. Part 1. Methods for sampling, 16 p. Erber, G., Routa, J., Kolström, M., Kanzian, C., Sikanen, L., Stampfer K., 2014: Comparing two different approaches in modeling small diameter energy wood drying in logwood piles. Croatian Journal of Forest Engineering 35(1): 15–22. Fagernäs, L., Brammer, J., Wilen, C., Lauer, M., Verhoeff, F., 2010: Drying of biomass for second generation synfuel production. Biomass and Bioenergy 34(9): 1267–1277. Filbakk, T., Høibø, O., Nurmi, J., 2011: Modelling natural drying efficiency in covered and uncovered piles of whole broadleaf trees for energy use. Biomass and Bioenergy 35(1): 454–463. Croat. j. for. eng. 38(2017)1


Modeling Pine and Birch Whole Tree Drying in Bunches in the Cutting Area (11–17) Fil’nej, M.I., 1966: New formulas for determination of thermodynamic properties of the water vapor containing in atmospheric air. News of higher education institutions. Construction and architecture 9: 1–90. Gigler, J.K., van Loon, W.K.P., van den Berg, J.V., Sonneveld, C., Meerdink, G., 2000: Natural wind drying of willow stems. Biomass and Bioenergy 19(3): 153–163. Golovkov, S.I., Koperin, I.F., Najdenov, V.I., 1987: Wood waste utilization as an energy source. Timber industry Publ. Moscow, 224 p. GOST 16483.0-89, 1989: Wood. General requirements for physical and mechanical tests, 15 p. GOST 17231-78, 1992: Round timber and splitted timber. Methods for determination of moisture content, 8 p. ISO 4470-81, 2009: Sawn products and wooden details. Methods for determining moisture content, 6 p. Kim, D.W., Murphy, G., 2013: Forecasting air-drying rates of small Douglas-fir and hybrid poplar stacked logs in Oregon, USA. International Journal of Forest Engineering 24(2): 137– 141. Kundas, S.P., 2008: Wood-biomass utilization as an energy source: scientific review. ISEU. Minsk. 85 p. Pettersson, M., Nordfjell, T., 2007: Fuel quality changes during seasonal storage of compacted logging residues and young trees. Biomass and Bioenergy 31(11–12): 782–792.

P. Anisimov et al.

Poršinsky, T., Sraka, M., Stankic, I., 2006: Comparison of two approaches to soil strength classifications. Croatian Journal of Forest Engineering 27(1): 17–26. Raitila, J., Heiskanen V.P., Routa, J., Kolström, M., Sikanen, L., 2015: Comparison of moisture prediction models for stacked fuelwood. Bioenergy Research 8(4): 1896–1905. Rajvanshi, A., 1986: Biomass gasification. Alternative Energy in Agriculture 2: 83–102. Routa, J., Kolström, M., Ruotsalainen J., Sikanen, L., 2014: Validation of prediction models for estimating the moisture content of small diameter stem wood. Croatian Journal of Forest Engineering 36(2): 283–291. Röser, D., Erkkilä, A., Mola-Yudego, B., Sikanen, L., Prinz, R., Heikkinen, A., Kaipainen, H., Oravainen, H., Hillebrand, K., Emer, B., Väätäinen, K., 2010: Natural drying methods to promote fuel quality enhancement of small energywood stems. Working Papers of the Finnish Forest Research Institute 186: 1–60. Scientific and Applied Reference Book on the USSR Climate. Long-term data, 1992. Gidrometeoizdat. St. Petersburg, 582 p. Shelgunov, Ju.V., 1982: Machines and equipment of logging, timber rafting and forestry: studies for higher education institutions. Timber industry Publ. Moscow, 520 p. Suhanov, V., 2008: Seasonal nature of logging. WOOD.RU 6: 42–45.

Authors’ addresses:

Received: March 12, 2016 Accepted: October 27, 2016 Croat. j. for. eng. 38(2017)1

Pavel Anisimov, MSc. * e-mail: anisimovpn@list.ru Assoc. prof. Evgenij Onuchin, PhD. e-mail: onuchinem@volgatech.net Marija Vishnevskaja e-mail: vishnevskajamm@volgatech.net Volga State University of Technology Yoshkar-Ola Republic of Mari El RUSSIA * Corresponding author

17



Original scientific paper

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy: Case Study of New Brunswick, Canada Lyes Bennamoun, Muhammad T. Afzal, Satyaveer Chauhan Abstract In order to investigate the effect of variation of the moisture content of forest biomass residues on a supply chain strategy, a simulation was performed using integrated biomass supply analysis and logistics modeling. A simple supply chain strategy was chosen and applied for Miramichi and Plaster Rock, two different regions in New Brunswick, Canada. These regions are selected based on three criteria: annual potential harvest of forest biomass residues, annual production potential of electric and thermal energy and distribution of transportation zones. The moisture content of forest biomass residues was dependant on the weather conditions of the selected regions. The results show that the moisture content of the biomass in Plaster Rock was more stable but higher than the biomass in the Miramichi region. In simulating the supply chain strategy, particular attention is given to harvest, baling, storage and transportation of the biomass. The simulation results show that, during harvest and baling of the biomass, the moisture content affects the dry matter loss and, as a consequence, the customer and ownership costs of the operations. It also affects the energy input and the quantity of carbon dioxide released in the atmosphere. However, dry matter loss and accordingly the cost of the operations are the main parameters affecting the storage and transportation of forest biomass residues. Keywords: supply chain, moisture content, forest biomass, transportation, storage, IBSAL model, dry matter losses

1. Introduction In Canada, the forest sector is one of the most important industries that contribute significantly to the economy of the country. About 80% of the forestry sector’s contribution to the Canadian economy is from solid wood product manufacturing for domestic consumption and export and pulp and paper product manufacturing (Government of Canada 2013). Logging is one the main activities of the forestry sector. It includes harvesting, storage and transportation of the wood to the mill. This sub-sector contributes about 20% of the forest sector’s contribution to the Canadian economy. The forestry and logging operation engender a huge amount of residues estimated at about 20¹0.6 million dry metric tonnes per year after the harvesting process (Cambero et al. 2015). Following the Croat. j. for. eng. 38(2017)1

Bio-pathways Project developed by the Forest Products Association of Canada (2010) the interest in the use of the biomass forest residues for the production of bioenergy, such as pellets, briquettes or biofuel, is continuously increasing in Canada. Depending on the produced quantity and its cost, the bioenergy, in particular the biofuel, can be produced to meet the local energy need of the mills or sawmills existing in the region and the production surplus can be considered for export. The biomass residues pass through several processes, mainly harvesting, storage in a pile or other forms, storage and transportation before delivery to destination (i.e. mills, sawmills or a biorefinery). Storage and transportation costs make the exploitation of biomass residues an expensive process that needs to

19


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

be optimized in order to make it suitable, efficient and cost effective. To address this, different mathematical models have been developed, to support and optimize the planning and management of the biomass residues supply chain. Cambero and Sawlati (2014) and Hughes et al. (2014) presented different supply chain models designed for forest biomass with respect to the economic, social and environmental aspects. The mathematical models were classified in seven main categories according to the desired criteria (Cambero and Sawlati 2014) or to the type of modeling approach (Hughes et al. 2014). Cambero et al. (2015) used a multi-period mixed integer linear program (MILP) to optimize the design of the supply chain strategy for forest residues for bioenergy and biofuel. Their study was applied to a region of Williams Lake in British Columbia, Canada. They determined, in their developed model, different types of the collected biomass and their sourcing points, the needed amount of biomass to transport from source to facilities and from facilities to market, the location, type and size of the conversion technology to be installed and the amount of bioenergy and biofuel to be generated. Cambero et al. (2015) investigated the production of heat, electricity, pellets as well as the option of producing bio-oil using pyrolysis process. Frombo et al. 2009a and 2009b focused in their study on the introduction of the geographical position and distances between the biomass sources and facilities by using GIS (Geographic Information System) based EDSS (Environmental Decision Support System). The optimized supply chain strategy considered different types of biomass (untreated woody and agro-forest biomass) with the particular objective of optimizing the transportation cost, considered as the one of the most expensive elements in the supply chain strategy. Freppaz et al. (2004) and Zhang et al. (2016), using the same GIS system, developed a more complete model by introducing the economic, regulatory and social criteria. This approach allowed to make a more appropriate and suitable supply chain strategy for the exploitation of forest biomass for energy supply. However, the application of the developed model was limited to a small mountain region in Italy that has its specific characteristics. The environmental criterion is crucial for the development of an appropriate supply chain strategy (Rafael et al. 2015a, 2015b, Jäppinen et al. 2014, Palak et al. 2014). Accordingly, including this criterion, the developed mathematical models will play an important role for its credibility. The multi-criteria supply chain optimal strategy developed by Vasković et al. (2015) was based on three criteria: energy efficiency of the production, economy of the production and the environmental issues represented mainly by the greenhouse gaseous (GHG)

20

emissions. The authors (Vasković et al. 2015) investigated the production of solid biofuel including different options, such as production of briquettes and pellets using wood residues from mills or sawmills. In addition to the different criteria cited above, variation of the moisture content of the biomass and all the parameters depending on the moisture content, such as the density and the heating value of materials and the influence of this change on different stages of the supply chain, have been introduced in the studies published recently (Jäppinen et al. 2014, Windisch et al. 2015, Daystar et al. 2014). Sokhansanj et al. (2006) developed and implemented a supply chain model called integrated biomass supply analysis and logistics model (IBSAL). The authors (Sokhansanj et al. 2006) studied the outdoor storage of agricultural biomass and subsequently simulated the effect of weather conditions on the variation of the moisture content of agricultural biomass and all moisture content related parameters (i.e. mass, density, equilibrium moisture content and heating value). The results of the study were presented in terms of energy input, carbon emissions, but did not specify which element of the supply chain model was affected by the variation of the moisture content. Ebadian et al. (2013) focused on the effect of different storage methods using (IBSAL) model. The authors (Ebadian et al. 2013) compared three different storage systems. It was possible, using IBSAL model, to determine the most suitable storage method by determining the cost of storage and transportation and by calculating the amount of the dry matter loss. This parameter is considered as an important element for analyzing the suitability of the supply chain strategy. The aim of this study is to answer the question: does the variation of the moisture content of forest biomass residues have an effect on supply chain strategy? A simple supply chain strategy is chosen and simulated, using IBSAL model, in two different regions in New Brunswick, Canada. A particular attention is given to the effect of the variation of the moisture content during storage and the consequences during transportation.

2. Methodology 2.1 Selection of the potential regions in New Brunswick As reported by Bouchard et al. (2012, 2013) and Wilson et al. (2010), around 80 to 85% of the area of the province is covered with forests. It represents approximately 60,000 km2 of productive forests and from 3.3 to 2.8 million ha of forest is managed by the New Croat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

Table 1 Total annual potential harvest of forest biomass residues, in green metric tonnes (Bouchard et al. 2012) Region

Total annual harvest, GMT

L. Bennamoun et al.

Table 2 Total annual production potential of electric and thermal energy presented by regions in New Brunswick (Bouchard et al. 2013) Region

Energy, PJ

Power, MW

Heat, MW

Dalhousie

143,634

Dalhousie

1.42

11.3

27.0

Bathurst

249,335

Bathurst

2.55

20.2

48.5

Tracadie-Sheila

68,017

Tracadie-Sheila

0.61

4.9

11.6

Miramichi

337,942

Miramichi

3.47

27.5

66.0

Richibucto

138,368

Richibucto

1.43

11.3

27.1

Doaktown

260,541

Doaktown

2.65

21.0

50.4

Moncton

322,973

Moncton

3.26

25.8

62.0

Chipman

177,056

Chipman

1.84

14.6

34.9

Sussex

261,992

Sussex

2.65

21.0

50.4

St-John

172,239

St-John

1.73

13.7

33.0

St Stephen

189,989

St Stephen

1.93

15.3

36.8

Fredericton

260,816

Fredericton

2.64

21.0

50.3

Nackawic

241,868

Nackawic

2.44

19.3

46.4

Juniper

257,893

Juniper

2.54

20.2

48.4

Plaster Rock

482,217

Plaster Rock

4.98

39.5

94.8

Edmundston

321,596

Edmundston

3.25

25.8

61.8

Kedgwick

365,451

Kedgwick

3.76

29.8

71.4

Brunswick Department of Natural Resources (NBDNR) (Wilson et al. 2010, Martin 2003). The focus of this part of the work is to determine the most suitable regions of the New Brunswick that have a potential of production and exploitation of forest biomass residues. The selection was made based on the potential harvest of forest biomass, generation of electric and thermal power and finally transportation distances from the source to the plant. Five potential regions in New Brunswick were selected in terms of production of forest biomass residues, based on the study by Bouchard et al. (2012, 2013). These regions are: Plaster Rock, Miramichi, Kedgwick, Edmundston and Moncton. Table 1 summarizes the total annual potential harvest of forest biomass residues expressed in green metric tonnes (GMT) calculated on wet basis and assuming an average value of the moisture content of forest biomass residues to be about 50% (Bouchard et al. 2012). The production of this main group varies from 480,000 to 320,000 GMT. Fredericton, Doaktown and Sussex can also be selected in a secondary group. Croat. j. for. eng. 38(2017)1

These regions produce around 260,000 GMT. Table 2 represents the total annual potential of producing electric and thermal energy using forest biomass residuals. It shows that the selected regions are now restricted to three regions; Plaster Rock with a total potential of energy equal to 4.98 PJ, then Kedgwick and Miramichi with 3.76 and 3.47 PJ, respectively. The transportation zones were divided into 5 ranges, 1–25 km, 26–50 km, 51–75 km, 76–100 km and 101–125 km. It was assumed that going above 50 km as transportation zone is not efficient and, therefore, this study is limited to the first two laps. Table 3 shows the percentage of distribution of forest biomass across the transportation zone. Table 3 confirms the selection of the three above said regions, with a total of 63% of forest biomass being within the transportation zone not exceeding 50 km. However, the regions of Miramichi and Plaster Rock show a higher percentage in the distribution of forest biomass in the transportation zone from 1 to 25 km. The selection of potential regions for this study is limited to Miramichi and Plaster Rock.

21


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

Table 3 Distribution of forest biomass (in percentage %) across the transportation zones in New Brunswick (Bouchard et al. 2013) Transportation zone Region 1–25 km

26–50 km

Dalhousie

25

47

Bathurst

19

48

Tracadie-Sheila

23

77

Miramichi

20

43

Richibucto

29

65

Doaktown

26

43

Moncton

21

58

Chipman

26

53

Sussex

27

65

St-John

26

46

St Stephen

22

46

Fredericton

26

54

Nackawic

22

51

Juniper

19

59

Plaster Rock

24

39

Edmundston

21

52

Kedgwick

15

48

2.2 Presentation of the simulation supply chain model and scenario The main objective of this study is to investigate the effect of the moisture content of biomass on the supply chain strategy. Accordingly, the Integrated Biomass Supply Analysis and Logistics (IBSAL) model was selected. As described by Sokhansanj et al. (2006), IBSAL model is a simulation of a biomass supply chain using different linked modules, and each module represents an element (process or event) of the supply chain (Sokhansanj and Ebadian 2007, Sokhansanj et al. 2008). As input data, logistic features, such as the number of agricultural farms, the average yield and the progress of the harvest schedule, are primarily introduced into the model. The variation of the moisture content of biomass is an important part of the model. Accordingly, this variation is defined by the introduction of the daily weather conditions (i.e. temperature, relative humidity, wind speed, rainfall and snow fall). The model will then deliver as output data, the cost and energy needed to complete different processes of the supply chain, carbon emission, final quantity of biomass delivered and dry matter loss of biomass. The simulation is performed using the simulation language EXTEND. The input and output data are inserted and obtained in Excel sheets. Both IBSAL and EXTEND are available in public domain. As the main focus of this work is to study the effect of the moisture content on different steps of supply chain, a simple scenario is adopted. The scenario is shown in Fig. 1. It starts with harvesting forest biomass residues (after wood harvesting). Usually, the

Fig. 1 Adopted supply chain scenario

22

Croat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

harvested biomass comes in the form of rectangular or round bales. In our case, the bale of the biomass is supposed to have a rectangular shape with the following dimensions 1.22×1.22×2.44 m3. The next procedure of the supply chain scenario is transportation of the biomass bales for a short distance to the storage points. In order to obtain variation in the moisture content of biomass, we simulated outdoor storage. The weather condition changes will then have a direct impact on the variation of the moisture content of the biomass bales. Afterward, the biomass bales are transported to the biorefinery for utilization as a source of bioenergy. As, in general, the biomass bales come in a wet form with moisture content around 50%, performing drying process is inevitable to make the biomass bales exploitable. 2.2.1 Modeling forest biomass residue moisture content The moisture content of forest biomass residues was supposed to be the sum of the internal and external moisture (Sokhansanj et al. 2006). This approach was successfully used for different types of forest and agriculture biomass (Sokhansanj et al. 2006, He et al. 2015, Nilsson and Karlsson 2005). The internal moisture content was represented using Lewis equation, written in the following form:

dMi = −aEp (Mi − M eq ) dt

(1)

Where: Mi internal moisture content of forest biomass residues expressed in dry basis (kg of water/kg of dry matter) Ep pan evaporation (mm/day). This coefficient depends on weather conditions (Appendix A) a a coefficient that depends on the studied material Meq equilibrium moisture content of biomass. It is represented using the following equation:

M eq = D +

E.T  1   rh − 1

1 F

(2)

Where: D, E and F c oefficients that can be determined from the sorption isotherms of the studied material. He et al. 2015 find that D, E and F have the values of: 0.1211, –0.00074 and 2.41 for the adsorption process of aspen biomass and equal to 0.1248, –0.000011 and 2.059 for desorption process of the same material. T and rh temperature and relative humidity of the air. Croat. j. for. eng. 38(2017)1

L. Bennamoun et al.

The external part of the moisture content of forest biomass residues varies with the weather conditions. It was proposed as a function of the pan evaporation Ep and the precipitation rate noted Pe, as defined below (Sokhansanj et al. 2006):

dM ex = bPe − cEp dt

(3)

b, c coefficients depending on the studied material. The pan evaporation is presented as a function of the weather conditions and the saturation vapor pressure. The formulas used for determination of the pan evaporation (from Holman 2002 and Eluripati 2007) and the saturation vapor pressure (from ASABE 1994) are presented in Appendix A. The total moisture content is simply the sum of the internal and external moisture content:

M tot = Mi + M ex

(4)

Furthermore and as proposed by Sokhansanj et al. (2006), the initial moisture content of the internal and external parts are proposed to be equal:

Mi = 0.8 M0

(5)

Mex = 0.2 M0

(6)

and

Where: M0 initial moisture content of biomass, usually determined by introducing samples of the biomass in an oven at 105 °C until no variation in the moisture content of the biomass can be recorded (Bennamoun et al. 2015). It is assumed for the simulation performed in this study that the initial moisture content of forest biomass residues, before storage process, is equal to 50% on wet basis. He et al. (2015) determined experimentally the value of (a) and (b) defined in equations (1) and (3) for aspen biomass. They find (a) equal to (0.203±0.023 mm–1), with a mean value of 0.206 mm–1 and the range of (b) was between 0.11 and 0.15 mm–1 with a mean value of 0.129 mm–1.

3. Results and discussion The focus is, as discussed previously, on the two selected regions in New Brunswick; Miramichi and Plaster Rock. The weather conditions of the selected regions, represented by air temperature, air humidity, wind speed, precipitation and snow on the ground, with daily frequency are first elements to be introduced to simulate the supply chain scenario using

23


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

IBSAL model. Fig. 2a and 2b represent the weather conditions for the region of Miramichi and Fig. 3a and 3b are data for Plaster Rock region. The presented data are for the year 2014 and are obtained from the government of Canada (http://climate.weather.gc.ca/). The observation and the comparison of the figures (2a, 2b and 3a,3b) show that the variation of the weather temperature, for both regions, is quite similar with low

temperatures in winter. The lowest temperature was registered in January and was around –26.5 °C in Miramichi and –31.5 °C in Plaster Rock. In this study, the focus will be on the after-snow period, which means from April to October. In this period of the year, in two regions the weather temperature changed from around 5 °C to 25 °C. The highest temperatures were registered in the months of June and July. The observa-

Fig. 2 Weather conditions in Miramichi in 2014: a) Temperature and relative humidity of the air and wind speed; b) Precipitation and snow in the ground

24

Croat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

tion of the variation of the relative humidity (rh) of the air, from Figures 2a and 3a, shows that rh in Miramichi changed from 40 to 95% against 50 to 90% for Plaster Rock. Regarding the studied period (April to October), the rh in Plaster Rock was more stable and more humid with variation between 70 to 90%, in particular during the month of June and July, comparing to Miramichi. In this latter region (Miramichi), in June and

L. Bennamoun et al.

July the rh changed from 50 to 90%. This means that in the period of April to October, the weather was more humid in Plaster Rock than in Miramichi, which could have an influence on the moisture content of forest biomass residues. The registered wind speed in Miramichi was higher than in Plaster Rock; the wind speed for Miramichi region was not less than 10 km/h with a mean value of about 15 km/h. The values of the

Fig. 3 Weather conditions in Plaster Rock in 2014: a) Temperature and relative humidity of the air and wind speed; b) Precipitation and snow in the ground Croat. j. for. eng. 38(2017)1

25


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

wind speed in this region ranged between 5 km/h and 35 km/h. However, in Plaster Rock region (Fig. 3a), the range of the wind speed was between 0 km/h and 20 km/h with several days without wind, which is not the case for Miramichi (Fig. 2a). As, for this study, outdoor storage of biomass residues is simulated, having a windy region can be an advantage, as moisture transfer exchange with the air can be increased with the wind speed increase. On the other hand, the data presented in Fig. 2b and 3b shows that Miramichi region is more exposed to rain and that it had received more precipitation with a total of around 1266 mm in 2014 compared to Plaster Rock region with 935 mm. The maximum amount of precipitation received per day in Miramichi was 120 mm, which is two times more than the maximum quantity in Plaster Rock. However, precipitation in Plaster Rock was more frequent than in Miramichi with a range between 10 to 20 mm per day. In Miramichi, several days without precipitation can be observed, in particular during the study period (April to October). As shown in the moisture content modeling (section 2.2.1.), the precipitation has a direct influence on the external part of the moisture content of biomass residues, which will increase with the increase of the received quantities of rain. Furthermore, Fig. 2b and 3b show that the quantity of snow registered in the ground increases continuously until the month of March. Afterwards, within a month, the quantity of snow dramatically decreases. The registered quantity of snow was much higher in Plaster Rock than in Miramichi with around 1000 mm for the first region (Plaster Rock) and 700 mm for the second region (Miramichi).

3.1 Effect of weather conditions on moisture content of forest biomass residues during outdoor storage During outdoor storage, the physical properties of biomass residues, in particular its moisture content, are directly linked to and influenced by the weather conditions (i.e. temperature, wind speed and relative humidity), as confirmed by the studies published by Mohanraj (2014), Visser et al. (2014) and He et al. (2015). Accordingly, the proposed model for simulating the variation of the moisture content represented by equations 1 to 6 clearly shows the effect of the weather conditions, introduced in equations 1 to 3. Fig. 4 shows the variation of the moisture content of forest biomass residues, during outdoor storage from April to October. In fact, Fig. 4 confirms the direct effect of the weather conditions on the variation of the moisture content of biomass residues. A particular attention is given to the temperature of the weather and the rain precipita-

26

tion. It is shown that the moisture content decreases during periods with high or moderate temperature (above 15°C) and low precipitation, as shown in the first 20 days. In other terms, there is a significant evaporation of the moisture content of the biomass. Moreover, Fig. 4 shows clearly that during periods with high precipitation, such as around day 190 with 120 mm for Miramichi and about 20 mm for Plaster Rock, the moisture content of the biomass in Miramichi increased from around 0.35 to 0.5 (wet basis) and from 0.3 to around 0.35 for the region of Plater Rock. Similarly, at the end of the studied period, the precipitation in Plaster Rock was much higher (40 mm) than in Miramichi (less than 20 mm), which was replicated by a higher moisture content of Plaster Rock’s biomass. Due to the large and frequent variation of the wind speed and relative humidity, it was not convenient to follow the effect of each parameter. Nevertheless, it is common to see in drying, wetting and heat and mass transfer field that increasing the velocity of the air or the wind speed increase the heat transfer exchanges with the surrounding air, which leads to a faster drying or evaporation of the moisture content of the studied material (Bennamoun and Belhamri 2006). Similarly, it was found that the rh of the air has a negative effect on the evaporation of the moisture content and accordingly on the drying time; the increase of rh results in an increase of the moisture content, which is reflected in a longer drying time (Bennamoun and Belhamri 2008a, Bennamoun and Belhamri 2008b). Mostly due to the frequent precipitation of rain in the region of Plaster Rock comparing to Miramichi, the moisture content of biomass residues in Plaster Rock had no low values as in Miramichi. Hence, it can be speculated that the moisture content of biomass in Plaster Rock was higher than the one in Miramichi.

3.2 Effect of moisture content on simulated supply chain strategy Dry matter loss (DML) and cost of the operations are the most important elements that give a picture of the effectiveness of the studied supply chain strategy. Accordingly, the effect of the moisture content on the supply chain strategy is then represented in terms of variation of the dry matter loss and the cost of different operations. A particular attention is given to storage and transportation of biomass residues. Furthermore, the environmental aspect defined by means of studying the energy input consumed by different equipment and the gaseous emissions are presented in this study. Fig. 5 shows the simulation results of harvest and storage operations of biomass residues for Miramichi (Fig. 5a) and Plater Rock (Fig. 5b) regions. The comCroat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

L. Bennamoun et al.

Fig. 4 Effect of weather conditions on moisture content of forest biomass residues during outdoor storage parison between Fig. 5a and Fig. 5b shows that the dry matter loss (DML), after shredding, forming the collected biomass residues in rectangular bales, transporting to the storage and then storing the bales in outdoor gravel pad storage spaces, attained a maximum of 31.88% of the initial total amount of the biomass in the region of Plaster Rock. This amount increased to 31.93% for the region of Miramichi. This difference is probably due to the variation of the

weather conditions, which has a direct effect on the moisture content of the collected biomass, as discussed previously (Fig. 4). Accordingly, increasing the moisture content of the biomass leads to a decrease of the dry matter loss. A similar result was obtained by Sokhansanj et al. (2008). As an example, they showed that increasing the moisture content of biomass from 30% to 40% (wet basis), decreased the dry matter loss from 25% to around 15%. The simulation using IBSAL

Fig. 5 Simulation results of harvest and storage operations using IBSAL model; a) Miramichi, b) Plaster Rock Croat. j. for. eng. 38(2017)1

27


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

model showed almost similar total custom and ownership costs for both studied regions. However, the costs were higher by 0.3% per ton for Miramichi than in Plaster Rock, probably as a direct consequence of the dry matter loss. Besides, the total energy input consumed by the equipment and the total quantity of carbon dioxide released in the atmosphere were quite similar for both regions. The quantities in Miramichi were higher by 0.4% per ton for the total consumed energy and 0.3% per ton for the carbon dioxide released. Transportation is considered as one of the most expensive operations, amounting to approximately 50% of the total cost of the supply chain (Hamedani 2015). Fig. 6 shows the quantity of dry matter loss during different transportation steps. The simulation gives two types of transportation: »bale«, which can be defined as transportation for short distance, such as to the storage. The second type is defined by »bulk«, which represents transportation for a long distance from storage to the biorefinery. The simulation results, using IBSAL model, show that the DML during transportation of the bales in Plaster Rock was higher than in Miramichi with 28.59% and 22.65%, respectively. This means that, during transportation, increasing the moisture content of the biomass increases the quantity of the DML. Accordingly and as discussed previously, the bales created after harvesting forest biomass residues in the region of Miramichi had a lower moisture content, which implies that the bales were more compact in Miramichi than in Plaster Rock. In other words,

more DML is obtained with the bales in Plaster Rock. A consequence of compact bales can also be seen in the bulk transportation, where the DML in Miramichi was much lower than in the region of Plaster Rock. The effect of the moisture content on the custom cost of the »bale« transportation was not significant, as the total amount was $27.27 for both regions. However, the total ownership cost (calculated per ton) was 0.9% higher in Plaster Rock probably because the DML was higher in this region. Likewise, no effect of the moisture content of the biomass was observed on the results related to the energy input and the total amount of carbon dioxide released in the atmosphere with 556.98 MJ per ton and 103.16 kg of CO2 per ton of biomass for both regions.

4. Conclusions and recommendation This study presents the variation of the moisture content of forest biomass residues defined by studying the weather conditions of the regions of Miramichi and Plaster Rock in New Brunswick, Canada and its effects on a simple supply chain scenario. The study shows that the dry matter loss DML is directly linked to the moisture content of biomass residues. From harvest to storage of the biomass, the increase of the moisture content leads to the decrease in the DML, and consequently also to the decrease of the cost of operations, energy input and gaseous emissions of the harvest and storage operations. However, during transportation, the decrease of the moisture content helps to have biomass bales more compact, which reduces the DML. Accordingly, the study shows that the moisture content can have a direct or an indirect effect on a supply chain strategy. Consequently, introducing the variation of the moisture content of the studied material can be an important element for the development of a realistic model and simulation of a supply chain strategy. During the harvesting period, from April to October, the solar radiation in Canada and in New Brunswick in particular, is significant. Taking advantage of this source of energy to reduce the moisture content of the biomass during outdoor storage and before long distance transportation can be a real advantage that can additionally reduce the total transportation cost of forest biomass residues to the biorefinery by reducing the moisture content of forest biomass residues.

Acknowledgements Fig. 6 Dry matter loss (DML) simulation results during transportation of biomass bales using IBSAL model

28

The authors acknowledge Value Chain Optimization (VCO), NSERC strategic network on value chain optimization, for its financial support of this project. Croat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

5. References ASABE, 1944: Psychrometric data (D271.2). American Society of Agricultural and Biological Engineers, MI, US. Bennamoun, L., Belhamri, A., 2006: Numerical simulation of drying under variable external conditions: Application to solar drying of seedless grapes. Journal of Food Engineering 76(2): 179–187. Bennamoun, L., Belhamri, A., 2008: Mathematical description of heat and mass transfer during deep bed drying: Effect of product shrinkage on bed porosity. Applied Thermal Engineering 28(17–18): 2236–2244. Bennamoun, L., Belhamri, A., 2008: Study of heat and mass transfer in porous media: Application to packed bed drying. Fluid Dynamics & Materials Processing 4(4): 221–230. Bennamoun, L., Chen, Z., Salema, A.A., Afzal, M.T., 2015: Moisture diffusivity during microwave drying of wastewater sewage sludge. Trans ASABE 58: 501–508.

L. Bennamoun et al.

ca/en/bio-pathways-network/details/phase-1-transformingcanadas-forest-products-industry-summary-of-findings. Freppaz, D., Minciardi, R., Robba, M., Rovatti, M., Sacile, R., Taramasso, A., 2004: Optimizing forest biomass exploitation for energy supply at a region level. Biomass and Bioenergy 26(1): 15–25. Frombo, F., Minciardi, R., Robba, M., Rosso, F., Sacile, R., 2009a: Planning woody biomass logistics for energy production: A strategic decision model. Biomass and Bioenergy 33(3): 372–383. Frombo, F., Minciardi, R., Robba, M., Sacile, R., 2009b: A decision support system for planning biomass-based energy production. Energy 34(3): 362–369. Government of Canada, 2013: Natural Resources Canada. http://www.nrcan.gc.ca/forests/industry/overview/13311; last update 2015-06-19.

Bouchard, S., Landry, M., Gagnon, Y., 2012: Forest biomass to energy atlas of New Brunswick. Report 2012, University of Moncton, New Brunswick, Canada, 30 p.

Hamedani, H.K., 2015: Logistics modeling of biomass supply chain in Ontario. Master thesis, University of British Columbia; https://circle.ubc.ca/bitstream/handle/2429/52710/ubc_2015_may_khaleghihamedani_hamid. pdf?sequence=7

Bouchard, S., Landry, M., Gagnon, Y., 2013: Methodology for the large scale assessment of the technical power of forest biomass: Application to the province of New Brunswick, Canada. Biomass and Bioenergy 54: 1–17.

He, X., Lau, A., Sokhansanj, S., Lim, J., Bi, X., 2015: Application of a model to simulate the wetting and drying processes of woody biomass in the field. Drying Technology 33(4): 434–442.

Cambero, C., Sowlati, T., 2014: Assessment and optimization of forest biomass supply chains from economic, social and environmental perspectives –A review of literature. Renewable and Sustainable Energy Reviews 36: 62–73.

Holman, J.P., 2002: Heat transfer. 9th ed. New York, McGrawHill; 665 p.

Cambero, C., Sowlati, T., Marinescu, M., Röser, D., 2015: Strategic optimization of forest residues to bioenergy and biofuel supply chain. International Journal of Energy Research 39(4): 439–452. Climate, 2015: Government of Canada. http://climate.weather.gc.ca/, last update: 2015-02-11.Daystar, J., Gonzalez, R., Reeb, C., Venditti, R., Treasure, T., Abt, R., Kelley, S., 2014: Economics, environmental impacts, and supply chain analysis of cellulosic biomass for biofuels in the Southern US: Pine, eucalyptus, unmanaged hardwoods, forest residues, switchgrass, and sweet sorghum. BioResources 9(1): 393–444. Ebadian, M., Sowlati, T., Sokhansanj, S., Townley-Smith, L., 2013: Modeling and analysing storage systems in agricultural biomass supply chain for cellulosic ethanol production. Applied Energy 102: 840–849. Eluripati, R.P., 2007: An improved model for estimating evaporation over lakes and ponds. University of New Orleans theses and Dessertations. Paper 567; Open access: http://scholarworks.uno.edu/cgi/viewcontent.cgi?article= 1567&context=td EXTEND. http://www.extendsim.com/support_downloads. html. Imagine That Inc. California, US. Forest Products Association of Canada, 2010: APAC. Transforming Canada’s forest products industry, summary of findings from the future Bio-pathways Project. http://www.fpac. Croat. j. for. eng. 38(2017)1

Hughes, N.M., Shahi, C., Pulkki, R., 2014: A review of the wood pellet value chain, modern value/supply chain management approaches, and value/supply chain models. Journal of Renewable Energy; Art. ID 654158: 1–14. IBSAL. http://biomass.ubc.ca/software/. Biomass and Bioenergy Research Group, University of British Columbia. Jäppinen, E., Korpinen, O.J., Ranta, T., 2014: GHG emissions of forest-biomass supply chains to commercial-scale liquidbiofuel production plant in Finland. GCB Bioenergy 6(3): 290–299. Martin, G., 2003: Management of New Brunswick’s crown forest. Department of Natural Resources, New Brunswick, Canada, 6 p. Mohanraj, M., 2014: Performance of a solar-ambient hybrid source heat pump drier for copra drying under hot-humid weather conditions. Energy for Sustainable Development 23: 165–169. Nilsson, D., Karlsson, S., 2005: A model for the field drying and wetting processes of cut flax straw. Biosystems Engineering 92(1): 25–35. Palak, G., Ekşiolu, S.D., Geunes, J., 2014: Analysing the impacts of carbon regulatory mechanisms on supplier and mode selection decisions: An application to a biofuel supply chain. International Journal of Production Economics 154: 198-216. Rafael, S., Tarelho, L., Monteiro, A., Monteiro, T., Gonçalves, C., Freitas, S., Lopes, M., 2015b: Atmospheric emissions from

29


L. Bennamoun et al.

Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

forest biomass residues to energy supply chain: A case study in Portugal. Environmental Engineering Science 32(6): 505– 515. Rafael, S., Tarelho, L., Monteiro, A., Sá, E., Miranda, A.I., Borrego, C., Lopes, M., 2015a: Impact of biomass residues to the energy supply chain on regional air quality. Science of The Total Environment 505: 640–648. Sokhansanj, S., Ebadian, M., 2007: Integrated biomass supply analysis and logistics (IBSAL) User’s guide. University of British Columbia. Sokhansanj, S., Kumar, A., Turhollow, A.F., 2006: Development and implementation of integrated biomass supply analysis and logistics model (IBSAL). Biomass and Bioenergy 30(10): 838–847. Sokhansanj, S., Turhollow, A., Wilkerson, E., 2008: Development of the integrated biomass supply analysis and logistics model (IBSAL). Oak Ridge National Laboratory, US, 49 p. Vasković, S., Halilović, V., Gvero, P., Medaković, V., Musić, J., 2015: Multi-criteria optimization concept for the selection

of optimal solid fuels supply chain from wooden biomass. Croatian Journal of Forest Engineering 36(1): 109–124. Visser, R., Berkett, H., Spinelli, R., 2014: Determining the effect of storage conditions on the natural drying of radiata pine logs for energy use. New Zealand Journal of Forestry Science 44(3): doi:10.1186/1179-5395-44-3 Wilson, J.J., Lantz, V.A., MacLean, D.A., 2010: A benefit-cost analysis of establishing protected natural areas in New Brunswick, Canada. Forest Policy and Economics 12(2): 94–103. Windisch, J., Väätäinen, K., Anttila, P., Nivala, M., Laitila, J., Asikainen, A., Sikanen, L., 2015: Discrete-event simulation of an information-based raw material allocation process for increasing the efficiency of an energy wood supply chain. Applied Energy 149: 315–325. Zhang, F., Johnson, D., Johnson, M., Watkins, D., Froese, R., Wang, J., 2016: Decision support system integrating GIS with simulation and optimisation for a biofuel supply chain. Renewable Energy 85: 740–748.

Appendix A The pan evaporation is determined using Holman’s evaporation formula (Holman 2002, Eluripati 2007):

Ep = A (0.37 + 0.0041 u) (Ps – Pw)0 · 88

(A–1)

The saturation vapor pressure is determined using the following equations (ASABE 1994): For 255.38≤T≤273.16 (in K)

Where: A a coefficient that takes the value of 0.8 for the floating pan and 0.7 for a land pan Ep, u, Ps and Pw expressed in in/day, mi/day and in of Hg respectively

Using the International system of units, equation (A–1) takes the following form:

Where: R = 2210564925 A = –27405.526 B = 97.5413 C = –0.146244 D = 0.12558 x 10–3 E = –0.48502 x 10–7 F = 4.34903 G = 0.39381 x 10–2

(

)

0   0.0041V ⋅ 0186 2.953 ⋅10–4 Ps (1 − rh ) ⋅ 88  0.37 +  0 Ep = A (A–2) 3401575

Where: Ep the pan evaporation expressed in mm/day V the speed of the wind, m/s Ps the saturation vapor pressure, given in Pa rh the relative humidity of the air, in decimal

30

ln ( Ps ) = 31.9602 − 6270 ⋅

3605 − 0.46057ln(T) (A–3) T

And: 273.16≤T≤533.16 (in K)

 P  A + BT + CT 2 + DT 3 + ET 4 ln  s  =  R FT − GT 2

(A–4)

The temperature T of the air is expressed in K.

Croat. j. for. eng. 38(2017)1


Assessment of Moisture Effect in Simulating Forestry Biomass Supply Chain Strategy ... (19–31)

L. Bennamoun et al.

Authors’ address: Lyes Bennamoun, PhD.* e-mail: lyes.bennamoun@unb.ca Prof. Muhammad T. Afzal, PhD. e-mail: mafzal@unb.ca University of New Brunswick – Fredericton Department of Mechanical Engineering P.O. Box 4400 E3B 5A3 Fredericton, New Brunswick CANADA

Received: January 7, 2016 Accepted: April 11, 2016 Croat. j. for. eng. 38(2017)1

Assoc. Prof. Satyaveer Chauhan, PhD. e-mail: satyaveer.chauhan@concordia.ca Concordia University John Molson School of Business 1450 Guy St. H3H 0A1 Montreal, Quebec CANADA * Corresponding author

31



Original scientific paper

Quantitative Estimation of Logging Residues by Line-Intersect Method Sergey P. Karpachev, Vjacheslav I. Zaprudnov, Maksim A. Bykovskiy, Evgeny N. Scherbakov Abstract Line intersect sampling (LIS) is a method used for quantifying forest residues after logging operations. In conventional LIS theory, forest residues are considered as separate pieces of cylindrical shape, they occur horizontally, and are randomly orientated and randomly distributed. In the case of cut-to-length (CTL) logging operation, forest residues represent separate clusters, consisting of pieces of branches, twigs, tips, etc. So the application of the conventional LIS theory for quantifying forest residues after CTL logging is difficult. The purpose of the article was to assess the accuracy of the modified LIS method for quantifying forest residues after CTL logging. The studies were conducted by computer simulations. In the models, the forest residues are represented as clusters in the form of circles. The laws of distribution of the radius of the clusters and their position in the plot were determined by field measurements. In the simulations, 4 types of clusters were considered: Þ type 1 – clusters uniformly distributed within the entire cutting area (Fig. 7) Þ type 2 – clusters uniformly distributed along the X-axis and five stripes on the Y-axis (Fig. 8) Þ type 3 – clusters uniformly distributed along the X-axis and three stripes on the Y-axis (Fig. 9) Þ type 4 – clusters uniformly distributed along the X-axis and one stripes on the Y-axis (Fig. 10) It was determined through simulation that the formula of the modified LIS method estimated appropriately forest residues after CTL logging. According to the results of simulation experiments, it was found that when the location of the lines of sample are across the area of Fig. 7, 8 (across the stripes with clusters), the results are in good agreement with the theoretical formulas. Differences are within error of 20%. Key words: Line intersect sampling (LIS), cut-to-length logging (CTL), simulation model, logging residues, clusters of logging residues, sample line

1. Introduction In recent years, forest residues are increasingly used in various industries, particularly in bioenergy industry. The choice of efficient technologies for collecting and processing of logging residuals are based on information about their quantity and quality. For quantifying forest residues, indirect assessment methods can be used as well as direct measurement in the cutting area. Indirect methods (regulatory, balance sheet and regulatory balance sheet) are used for quantitative esCroat. j. for. eng. 38(2017)1

timation of logging residues based on the number of branches and tops of the plants. According to the Russian Guidelines, the volume of logging residues can be calculated by the formula (Guidelines 1985):

Q=

V ×N 100

(1)

Where: Q volume of logging residues, m3 V volume of raw materials, m3 N norm of wood waste generation, %

33


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

The plot sampling is mostly used for estimating plants in the forest, animal nests, soil fauna, etc. According to this method, every item is counted within each plot (plants, nests, etc). This method may also be used for sampling logging residuals (Ghaffariyan et al. 2012, Ghaffariyan 2013). In this case, the estimated volume of wood W (m3/m2) in a cutting area:

Fig. 1 Methods of estimation of logging residues The norm of wood waste generation N depends on the technology of logging operations and growing conditions of the forest. So the indirect methods are not accurate and require field inspection of the cutting area. Methods of quantitative estimation of logging residues by direct measurements on the cutting area can be divided into the following groups (Fig. 1): Þ method of expert evaluations Þ method of complete counting Þ statistical methods. All the above estimation methods were tested by the authors in practice, for assessing logging residues, windfalls, sunken logs, etc. The method of expert evaluations, the Delphi Method (Dalkey and Olaf 1963), was applied by the authors of the article to assess the sunken logs in rivers. The method included groups of highly skilled professionals with experience in logging and timber rafting. For each river, where there were sunken logs, a group of experts of 7 to 9 people was selected. The comparison of the experimental data on the number of sunken logs with the data obtained by the method of expert estimates showed difference of about 100%. For this reason, this method cannot be applied in practice. The method of complete counting assumes a continuous recalculation of the whole wood. Complete counting consumes more time, but can be effective in the use of modern technologies, for example, using aerial and satellite images. Statistical methods are best suited to assess the quantity and quality of logging residues. The accuracy and complexity of statistical methods depends on the volume of the sample. The authors have used the following statistical methods: Þ plot sampling method Þ line-intersect sampling (LIS).

34

W=

Q S

(2)

Where: S total area of the sample plot, m2 Q volume of wood on the sample plot, m3 The line intersect sample (LIS) was originally described by Warren and Olsen (Warren and Olsen 1964). Since then, the LIS method has been extensively used for the quantity estimation of logging residues (Van Wagner 1968, Van Wagner and Wilson 1976, Van Wagner 1982, Bailey 1969, Bailey 1970, Bailey and Lefebvre 1971, Howard and Ward 1972, De Vries 1973, De Vries 1974, Brown and Roussopoulos 1974, Brown 1974, Van Wagner 1976, Pickford and Hazard 1978, Pickford and Hazard 1986, Harmon et al. 1986, Karpachev 2008a, Karpachev 2008b, Karpachev and Scherbakov 1990, Karpachev and Scherbakov 2009, Karpachev and Scherbakov 2013, Karpachev et al. 2010, Linnel Nemec and Davis 2002, O’Hehir and Leech 1997, Woldendorp et al. 2004, Bate et al. 2009, Bell et al. 1996). The LIS method allows obtaining sufficiently accurate results, and the time for sampling is reduced, as compared with the plot sampling method, by about 60–70%. However, it was noted that the accuracy of the estimations is significantly affected by the concentration of the logging residuals per unit area. In recent years, the LIS method has probably been one of the most common techniques for the assessment of forest residuals. According to the LIS method, logs and their pieces that are crossed by a sample line are selected into a sample (Fig. 2). The volume of all wood on the cutting area is estimated by the sample logs and their pieces.

Fig. 2 A graphical explanation of the LIS method (1 – sample line, 2 – intersected log) Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

The theory of the LIS method was the subject of numerous scientific papers (De Vries 1973, Marshall et al. 2000, De Vries, 1979, De Vries 1986, Hazard and Pickford 1984, De Vries 1974, Van Wagner 1976, Karpachev and Scherbakov 2013). It should be noted that much of the published literature on LIS theory and its application in practice falls under the topic of estimation of logging residues in the form of separate logs and their pieces distributed across the area (Fig. 2). Such characteristics of forest residuals are typical for whole-tree logging (Feller buncher + skidder technology). In the case of cut-to-length (CTL) logging operation, forest residues look like separate heaps (Fig. 3), consisting of pieces of branches, twigs, tips, etc. So the application of the conventional LIS theory for quantifying forest residues after CTL logging is difficult. In recent years, theoretical and field studies on the application of the LIS method for estimating logging residues after CTL logging have been carried out. (Karpachev et al. 2010, Karpachev and Scherbakov 2013). The purpose of this paper is to provide information on LIS method for estimating logging residues in the form of separate heaps (it will be called clusters of logging residues or simply – clusters) after CTL logging. This paper: Þ explains the theory underlying LIS for estimating clusters Þ provides basic formulas for estimating the number of clusters Þ provides simulation models of the heaps with different statistical characteristics and LIS field procedures for estimating the number of clusters Þ provides basic results of computer simulation of LIS method for estimating the number of clusters Þ briefly describes the field-sampling requirements for LIS.

2. Theoretical Approach As defined before, this paper deals with forest residues after CTL logging. These forest residuals are clusters consisting of branches, twigs, tips, etc. (Fig. 3). When considering a plane rectangular cutting area of size LхH, the area contains n clusters (Fig. 4) and all the clusters have the shape of a circle of radius R. The sample line of length l passes through the cluster area. The sample line is parallel to the ordinate axis. The sample line will be equal to the width of the area: l = H. The coordinates of the beginning of the line will Croat. j. for. eng. 38(2017)1

S. P. Karpachev et al.

Fig. 3 The clusters after CTL logging (author’s photo)

Fig. 4 Scheme of clusters on the cutting area (1 – sample line, 2 – clusters of logging residues) be M1(X1, Y1= 0). Coordinates of the end of the sample line will be equal to M2 (X2 = X1, Y2 = H). The number of clusters on the cutting area of size L × N can be defined by the formula: Where:

N=

M ,  m  p

(3)

M ,  m  mathematical expectation of the number of intersections between the sample line and clusters

P

robability that the sample line will intersect p the cluster.

35


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

In practice, the value M, [m] can be estimated by the average number of clusters that intersects with n lines. If it is defined that all sample lines have the same length l, which is convenient in practice, then the mean will be equal to: n

m=

∑m j=1

n

j

(4)

Estimation of the number of clusters on the cutting area can be defined by the formula: m N ≈ Ñ = p

(5)

The probability that the sample line will intersect the cluster of radius R, will be equal to:

p=

Where:

W+ W

(6)

Ω a ll position of clusters on the cutting area, i.e. a complete system of events. The coordinates of the centers of the clusters xi, yi are defined as a uniform distribution on the interval [0, L] [0, H]. Then the probability that the sample line with the length l > 2R on an area of the size L × H will intersect the cluster with the radius R, will be equal to (Karpachev and Scherbakov 2013). Assuming the condition: l >> 2R, then (Eq. 7) is transformed as follows: 2×R×l p= (8) L×H If the sample line is parallel to the ordinate axis and equal to the width of the cutting area l = H, then the formula (Eq. 8) can be represented in the form: p=

2×R L

n 1 L × × ∑ mi (10) n 2 × R i=1 The required number of sample lines can be defined according to the known formula:

Nń=

2

Where:

V ⋅t  N=   P 

(11)

t confidence factor P accuracy rate, % V coefficient of variation, %. In theoretical studies, some assumptions are accepted: Þ the radius of all clusters are the same and equal

to R

Þ the coordinates of the cluster centers xi,yi on the

Ω+ event that will bring the sample line to intersect the cluster

If there are n sample lines, then, in this case, the estimate of the number of clusters is equal to:

(9)

cutting area are defined according to a uniform distribution law

Þ the length of all sample lines are the same l = H

and l >>R.

Because of the accepted assumptions, the theoretical formula may not be accurate enough in practice. In particular, a number of questions arise: Þ What will be the effect of variability of the ra-

dius of clusters on the accuracy of the estimate?

Þ What will be the effect of variability of the ra-

dius of clusters on sampling?

Þ What will be the effect of the law of distribution

of coordinates of the cluster centers xi, yi on the accuracy of the estimate?

3. Methods In theoretical studies, we have accepted a number of assumptions about the characteristics of clusters. It was decided to covnduct simulation experiments with

2

 2  l 2 2   R −   + r   2   2 2   2 l  l 1 R  l  Rc R − arctg arcsin p 2 ⋅ R 2 + 4p  − + 2 4⋅r 2  2  2 R 2 R  8 ⋅ dr  l   4p ∫ r ⋅  l 1+  R*     2R   2 + p= p ⋅ L⋅ H p ⋅ L⋅ H

36

(7)

Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

mathematical models using the actual characteristics of clusters. Simulation studies on LIS were carried out by various authors (Pickford and Hazard 1978, Pickford and Hazard 1986, Karpachev 1990). All these studies had different purposes but all papers considered the logging residues in the form of separate logs and their pieces distributed across the area. In this paper, the main task was to develop stochastic mathematical models that adequately describe the process of quantifying clusters by the LIS method. Also, it was decided to undertake studies of accuracy of LIS method in order to estimate clusters in a wide range of variation of their characteristics. In these models, we used the statistical characteristics of the clusters obtained in field measurements in the central regions of Russia. (Karpachev et al. 2010). Based on the field measurement data, the following assumptions were accepted and used in the model: Þ shape of the clusters is a correct circle with vari-

able radius Rj Þ variation of radius of the clusters is described by the normal distribution law.

In theoretical studies, the distribution of the cluster’s centers coordinates xi, yi on the cutting area was defined as uniform. Field measurements showed that the distance between the clusters was in accordance with uniform distribution law (Karpachev et al. 2010). However, the location of the clusters is determined by the technology of the harvester and may differ from the uniform law. Usually clusters are located in accordance with the technology traffic lane (strip) of the harvester. Because of this, the distribution law of the cluster’s coordinates on the cutting area may be different from the uniform law. In the model, two types of distribution of the cluster’s coordinates xi, yi were considered: Þ uniform distribution within the entire cutting

area (Fig. 4) Þ uniform distribution within the technological strips of width b (Fig. 5). According to (Eq. 10), it would be necessary to know the number of clusters to estimate the number of clusters that intersected with the sample lines. The fact of intersection of the cluster with the sample lines was determined in the model by the following algorithm: Þ each sample line was represented by the equa-

tion of a straight line on the area and was defined by the point of its beginning M1j(X1j, Y1j = 0):

Croat. j. for. eng. 38(2017)1

S. P. Karpachev et al.

X = Xij

(12)

Þ each cluster was defined as a circle with center

Ci(xi, yi) and variable radius Ri

Þ event of intersection of the jth sample line with

the ith cluster was determined as (Fig. 4):

(xi + Ri) ³ Xij ³ (xi + Ri)

(13)

The model should simulate the LIS process of quantifying clusters. To implement this process, the model was composed of two blocks (Fig. 6): Þ generation block of clusters on the cutting area

with the specified characteristics

Þ generation block of sample lines on the cutting

area with the specified characteristics.

The model contains three procedures: Þ procedure of counting the actual number of

clusters on the cutting area

Þ procedure of LIS method for the clusters and

estimation of their number

Þ procedure of comparison between the estimated

number and the actual number of clusters.

The modeling principles and algorithms of these programs were taken into consideration. Characteristics of clusters in the model were assigned according to the following algorithm: Þ the number of clusters n on the cutting area was

specified in the primary source data

Þ the radii Ri of clusters was generated according

to the normal distribution law.

The coordinates of the cluster’s center xi, yi were generated in the intervals [0, L] [0, H]. In the model, two cases of distribution of the cluster’s coordinates on the cutting area were considered: Þ uniform distribution Þ distribution by stripes.

In the latter case, the coordinates of the center of clusters xi were generated by the uniform distribution on the interval [0, L], and the coordinates yi were generated by the normal distribution law in the interval [0, H] with the mean in the centers of the strips b (Fig. 5). The model generated the sample lines with a set of specified characteristics. The main characteristics of the sample line were: Þ length of the sample line l Þ coordinates of the sample line on the cutting

area Xj, Yj = 0.

37


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

Characteristics of the sample lines in the model were assigned according to the following algorithm: Þ the beginning of coordinate Y1j of each sample line was Y1j = 0 Þ the beginning of coordinate X1j of the sample line was generated by the uniform distribution law in the interval [0, L], so that the line passed across the whole cutting area Þ length of each sample line was equal to the width of cutting area H. Examples of generating the clusters with specified characteristics are presented in Fig. 7 and Fig. 8. Fig. 7 shows only one sample line. The coordinates of the center of clusters xi, yi were generated in accordance with the following laws: Þ Fig. 7 – uniform law Þ Fig. 8 – distribution by strips. The specified characteristics of clusters (xi, yi, Ri) were generated by means of Excel software. 4 types of clusters were considered in the experiments: Þ type 1 – clusters uniformly distributed within the entire cutting area (Fig. 7) Þ type 2 – clusters uniformly distributed along the X-axis and within five stripes on the Yaxis (Fig. 8) Þ type 3 – clusters uniformly distributed along the X-axis and within three stripes on the Yaxis (Fig. 9)

Fig. 6 Structural diagram of a simulation model Þ type 4 – clusters uniformly distributed along the

X-axis and within one stripes on the Y-axis (Fig. 10). The generation of various types of clusters was saved as database for the simulation experiments. The number of clusters in the model varied in each type of clusters from 10 to 170 pieces in increments of 40 pieces. The specified characteristics of the clusters were tested for compliance with the above mentioned laws of distribution and their statistical characteristics were Table 1 Example of statistical characteristics of the clusters (Fig. 7) Statistical characteristic

Fig. 5 Location scheme of cluster stripes on the cutting area (1 – sample line, 2 – clusters, a – harvester movement technology corridor, b – technology stripes of clusters)

38

Value x1

y1

R

Mean

51.45

53.90

3.11

Variance

867.56

740.05

2.96

Standard deviation

29.45

27.20

1.72

Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

S. P. Karpachev et al.

Table 2 Example of statistical characteristics of the clusters (Fig. 8) Value Statistical characteristic

Strip 1

Strip 2

Strip 3

Strip 4

Strip 5

R

x1

y1

x2

y2

x3

y3

x4

y4

x5

y5

Mean

52.27

7.43

49.03

27.46

49.73

47.71

49.07

67.44

51.58

87.47

2.95

Variance

1032.42

2.45

749.38

2.11

814.22

2.13

799.88

2.03

913.34

1.71

3.46

Standard deviation

32.13

1.57

27.37

1.45

28.53

1.46

28.28

1.43

30.22

1.31

1.86

Fig. 7 An example of generation of 90 clusters with the uniform distribution (one sample line and its coordinates are demonstrated)

Fig. 8 An example of generation of 90 clusters within five stripes

calculated. Examples of the estimated statistical characteristics of the clusters are given in Table 1 and Table 2. After generation of the characteristics of clusters, they were saved in Excel tables. The sample lines on the cutting area were set as follows: Þ for 1st type of clusters – along the Y-axis (case 1, Table 3) Þ for 2nd type of clusters – along the Y-axis (case 2, Table 4) and along the X-axis (case 3, Table 4) Þ for 3th type of clusters – along the X-axis (case 4, Table 5) Þ for 4th type of clusters – along the X-axis (case 5, Table 6). The required numbers of sample lines for estimation of the number of clusters were defined according to the formula (Eq. 11). Croat. j. for. eng. 38(2017)1

Fig. 9 An example of generation of 50 clusters within three stripes

39


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

Fig. 10 An example of generation of 10 clusters within one stripe

Fig. 11 Program’s interface for estimation of clusters by LIS method

Estimation of the number of clusters was conducted using the algorithm described above. Estimation procedures were implemented in Delphi 7 program. The program’s interface of the estimation of clusters by LIS method is shown in Fig. 11. In the experiments, the accuracy rate was assumed to be 20%.

4. Results and Discussion Examples of LIS simulation results are displayed in Tables 3 – 7. The results of simulation experiments (Table 3 – case 1, Table 4 – case 2) with the clusters on cutting

Table 3 Example of results of LIS simulation (case 1) True number

Estimation number

Mean number of

Standard

Error

of clusters

Theoretical number Estimation number of lines

of lines

of clusters

intersections per line

deviation

%

10

150

188

8

0.48

0.672

20

50

30

29

43

2.633

1.449

12.22

90

16

16

95

5.75

2.352

–6.48

130

11

7

139

8.363

2.377

–7.22

170

8

13

138

11

4.14

–7.84

Table 4 Example of results of LIS simulation (case 2) True number

40

Estimation number

Mean number of

Standard

Error

of clusters

Theoretical number Estimation number of lines

of lines

of clusters

intersections per line

deviation

%

10

150

126

12

0.733

0.84

–22.22

50

30

24

48

2.9

1.47

3.33

90

16

6

84

5.062

1.34

6.25

130

11

11

137

8.272

2.831

–6.06

170

8

3

193

11.625

2.326

–13.97

Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

S. P. Karpachev et al.

Table 5 Example of results of LIS simulation (case 3) True number of clusters

Theoretical number Estimation number of lines of lines

Estimation number Mean number of Standard deviation of clusters intersections per line

Error %

10

150

121

12

0.733

0.824

–22.22

50

30

124

46

2.8

3.188

6.66

90

16

83

84

5.062

4.711

6.25

130

11

96

148

8.909

8.938

–14.21

170

8

68

241

14.5

12.247

–42.15

Table 6 Example of the results of LIS simulation (case 4) True number of clusters

Theoretical number Estimation number of lines of lines

Estimation number Mean number of Standard deviation of clusters intersections per line

Error %

10

150

586

7

0.433

1.07

27.77

50

30

220

41

2.466

3.739

17.77

90

16

164

105

6.312

8.268

–16.89

130

11

346

80

4.818

4.818

38.22

170

8

250

202

12.125

19.577

–18.87

Table 7 Example of results of LIS simulation (case 5) True number of clusters

Theoretical number Estimation number of lines of lines

Estimation number Mean number of Standard deviation of clusters intersections per line

Error %

10

150

883

8

0.493

1.496

17.77

50

30

532

81

4.866

11.458

–62.22

90

16

816

101

6.062

17.68

–12.268

130

11

476

287

17.272

38.471

–121.44

170

8

559

60

3.625

8.749

64.46

areas (Fig. 7, Fig. 8) were plotted in Fig. 12 as markers. For comparison, Fig. 12 shows the theoretical curve obtained from the equation (Eq. 11). As follows from the graph in Fig. 12, the discrepancy between the theoretical and experimental results has exceeded the error of 20% in only one case. It should be emphasized that the sample lines in these cases were directed across the cutting area (along Y axis). The X coordinate of the lines were determined according to the uniform distribution law. It is logical to assume that if the sample lines are drawn along the cutting area (Fig. 8, 9, 10) (along the Croat. j. for. eng. 38(2017)1

X-axis and along the stripes of the clusters), then the variance of the average number of intersection of clusters with the lines will be increased with a decrease of the number of stripes. This should increase the required number of sample lines. Simulation experiments (Fig. 8, 9, 10) have confirmed this hypothesis. This is also shown in Table 5, 6, 7 and the graph in Fig. 13. For example, for the cutting area with one stripe with 50 clusters on the area (Fig. 10 and Table 7), the theoretical numbers of sample lines are 30 lines. The number of sample lines, calculated by the results of the experiments, amounted to 532 lines.

41


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

Fig. 12 Dependence of the required number of sample lines on the number of clusters

Fig. 14 Dependence of the required number of sample lines on the mean number of intersections of clusters with sample line

In order to estimate the variance (standard deviation) of the number of clusters intersected with the lines, an additional simulation experiment was carried out with clusters located on one stripe (Fig. 10).

In a simulation model carried out to estimate the number of clusters, 1000 sample lines were generated for each cutting area. The results are displayed in Table 8. The table shows that the sample lines should be located across the area (along Y axis). In this case, the required number of sample lines is close to the theoretical number. Correspondingly, 20 and 16 lines are required. In comparison to the lines drawn along the area (along the X-axis), the required number of sample lines is very different from the theoretical one. Correspondingly, 823 and 16 lines are required. This can be explained by large differences in standard deviation (16.168 and 2453, Table 8). From a practical point of view, the graph in Fig. 14 is very interesting. The graph shows the dependence of the required number of sample lines on the average number of intersections of clusters with the sample line.

5. Conclusions and Practical Recommendations Estimation of the number of clusters on the cutting area of a rectangular shape can be made according to the formula (Eq. 5).

Fig. 13 Dependence of the required number of sample lines on the number of clusters for different types of clusters

42

If the sample lines cross the cutting area (parallel to the Y-axis) and are equal to the width of the area, then the probability that the sample line will intersect the cluster can be determined by the equation (Eq. 9). Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

S. P. Karpachev et al.

Table 8 Results of estimation of clusters by 1000 sample lines (case 5) Direction of line

True number of clusters

Theoretical number of lines

Estimation number of lines

Estimation number of clusters

Mean number of intersections per line

Standard deviation

Error %

OY

90

16

20

88

5.315

2.453

1.574

OX

90

16

823

92

5.522

16.168

–2.259

Estimation of the number of clusters should be done according to the results of the intersections of the clusters with n sample lines according to the formula (Eq. 10). The required number of sample lines can be determined according to the known formula (Eq. 11). Simulation experiments were carried out for 4 types of clusters as shown in graphs in Fig. 7, 8, 9, 10. The results of simulation demonstrated that, when the location of the sample lines are across the area as shown in Fig. 7, 8 (across the stripes with clusters), the results are in good agreement with the theoretical formulas. Differences are within the error of 20%. This is due to the fact that, in this case, the estimation is only affected by the x-coordinate of the clusters. When the location of the sample lines are along the plot (along the strips with clusters, Fig. 8, 9, 10), the results disagreed with the theoretical formulas. The differences can exceed the error of 100%, for example as shown in Table 7. A significant discrepancy is explained by the fact that, in this case, the assessment is only affected by the y-position of the center of cluster, which is distributed within the stripe. Part of the sample lines go through the stripes with a large number of intersections with the clusters, but another part of the lines get between the strips, where there will be no intersections between the lines and clusters. It is clear that this will lead to an increase of the variance (or standard deviation), which will correspondingly increase the required number of sample lines. For clusters with coordinates x, y distributed by the uniform law through the cutting area, it is possible to carry out a sample line across the area and along the area. In this case, in practice, it makes no difference how to draw the sample lines (along or across the cutting area). For clusters, which are distributed in the cutting area inside the stripes, the sample lines should pass across the stripes. The required number of the sample lines can be pre-determined according to the formula (Eq. 11) and can be clarified during the field measurement process using the graph (Fig. 10). Croat. j. for. eng. 38(2017)1

6. References Bailey, G.R., 1969: An evaluation of the line-intersect method of assessing logging residue. Inform. Rep. VP-X-23, Can. Dep. Fisch. Forest., Forest Prod., Lab., Vancouver, B.C., 36 p. Bailey, G.R., 1970: A simplified method of sampling logging residue. Forest. Chron. 46(4): 288–294. Bailey, G.R., Lefebvre, E.L., 1971: Estimating volume distributions of logging residue from intersect-sampled data. BiMon. Res. Notes 27: 4–5. Bate, L., Torgersen, T., Wisdom, M., Garton, E., 2009: Biased estimation of forest log characteristic using intersect diameters. Forest Ecology and Managment 258(5): 635-640. Bell, G., Kerr, A., McNickle, D., Woollons, R., 1996: Accuracy of the line intesect method of post-logging sampling under orientation bias. Forest Ecology and Management 84(1): 23–28. Brown, J.K., 1974: Handbook for inventorying downed woody material. USDA For. Servo Gen. Tech. Rep. INT-16, p. 24. Brown, J.K., Roussopoulos, P.J., 1974: Eliminating biases in the planar intersect method for estimating volumes of small fuels. For. Sci. 20: 350–356. Dalkey, N., Olaf, H., 1963: An Experimental Application of the Delphi Method to the use of experts. Management Science 9(3): 458–467. doi: 10.1287/mnsc.9.3.458. De Vries, P.G., 1973: A general theory on line intersect sampling with application to logging residue inventory. Madelingen Landbouw Hogeschool. No 73-11, Wageningen, Netherlands, p. 23. De Vries, P.G., 1974: Multistage line intersect sampling. Forest Sci. 20(2): 129–134. De Vries, P.G., 1979: Line Intersect Sampling: Statistical Theory, Applications and Suggestions for Extended Use in Ecological Inventory. In Sampling Biological Populations. Editors G.M. Cormack, G.P. Patil, and D.S. Robson. International Coop. Publishing House. Fairland, Maryland, Statistical Ecology Series, Vol. 5: 1–77. De Vries, P.G. 1986: Sampling Theory for Forest Inventory. Springer-Verlag. New York, 399 p. (Chapter 13, pages 242– 279, deals specifically with line intersect sampling.)

43


S. P. Karpachev et al.

Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

Ghaffariyan, M.R., 2013: Remaining slash in different harvesting operation sites in Australian plantations. Silva Balcanica 14(1): 83–93.

and villages in forest regions of Russia. FORMEC. 41. International Symposium in Schmallenberg, Germany 02 – 05 June, ISBN: 978-3-9811335-2-3, Gross-Umstadt, 51–56.

Ghaffariyan, M.R., Brown, M., Acuna, M., 2012: Sampling methodology for left-slash assessment in collecting forest biomass. Forest Energy Blog, October 31, 2012 at http://blog. forestenergy.org.

Karpachev, S.P., 2009: Operative Method for Quantitative and Qualitative Estimation of Forest Residuals After Harvesting Logging Operations as Biomass for Bioenergy, Bioenergy, Sustainable Bioenergy Business, 4th International Bioenergy Conference and Exhibition, 31. August to 4. September. Available at: http://www.hedon.info/docs/Bioenergy2009_progamme.pdf.

Ghaffariyan, M.R., Jenkin, B, Mitchell, R., Brown, M., 2016: Quantitative and Qualitative Assessment of Timber Harvesting Residues: A Case Study of a Balsa Plantation in Papua New Guinea. IUFRO Research Group 3.08 Small-scale Forestry Conference. Small-scale and Community Forestry and the Changing Nature of Forest Landscapes. 11 – 15 October 2015. Sunshine Coast, Australia. Conference Proceedings, 65–77. Guidelines, 1985: Методические указания по определению объемов древесных отходов. М. Внипиэилеспром, 15 p. Harmon, M.E., Franklin, J.F., Swanson, F.J., Sollins, P., Gregory, S.V., Lattin, J.D., Anderson, N.H., Cline, S.P., Aumen, N.G., Sedell, J.R., Lienkaemper, G.W., Cromack Jr., K., Cummins, K.W., 1986: Ecology of coarse woody debris in temperature ecosystems. Adv. Ecol. Res. 15: 133–302. Hazard, J.W., Pickford, S.G., 1984: Cost Functions for the Line Intersect Method of Sampling Forest Residue in the Pacific Northwest. Canadian Journal of Forest Research 14(1): 57–62. Hazard, J.W., Pickford, S.G., 1986: Simulation studies on line intersect sampling of forest residue. Part II. Forest Sci. 32(2): 447–470. Howard, S.O., Ward, F.R., 1972: Measurement of logging residue – alternative applications of the line intersect method. USDA Forest Serv. Res Note PNW-183, Pac Northwest Forest and Range Exp.Stn., Portland, Oregon, 8 p. Karpachev, S.P., 2005: Quantitative estimation of the environmental impact of the traditional use of forest rivers for floating timber transport: estimating of the volume of sunken timber in rivers. Session 013, XXII IUFRO World Congress. Forests in the Balance: Linking Tradition and Technology, 8 – 13 August. Karpachev, S.P., 2006: Quantitative estimation of the environmental impact of the traditional use of forest rivers for floating timber transport: estimating of the volume of sunken timber in rivers. 2nd International Conference on Environmentally-Compatible Forest Products. Fernando Pessoa University. Oporto, Portugal 20 – 22 September. Karpachev, S.P., 2008a: The quantitative estimation of the forest residuals as biomass to produce bioenergy for local industry and villages in forest regions of Russia. Proceedings World Bioenergy, The Swedish Bioenergy Association, Svebio, Torsgatan 12, SE-111 23 Stockholm, Sweden, ISBN: 97891-977624-0-3, 408–410. Karpachev, S.P., 2008b: The quantitative estimation of the forest residuals as biomass for bioenergy for local industry

44

Karpachev S.P., Shcherbakov, E.N., 1990: Modelling on computer resource estimates and parameters of logging residues by the method of line intersections – M. Scientific. Tr., Moscow State Forest University, 226, 4 p. Karpachev S.P., Shcherbakov E.N., 2009: Basic research of the probabilistic features of the estimation of the heaps of the forest residuals by the linear intersection method. Moscow State Forest University Bulletin – Lesnoj Vestnik, 4(67): 97–99. Karpachev S.P., Shcherbakov E.N., 2013: Statistical evaluation of the quantity and quality of accumulations of wood produced in forest areas and water bodies. Monography, Moscow State Forest University, 132 p. Karpachev S.P., Shcherbakov E.N., Slinchencov A.N., 2010: Quantification of logging residues after CTL logging by harvesters. The lesopromyshlennik journal 4(56): 29–31. Linnel Nemec, A.F., Davis, G., 2002: Efficient of six line intersect sampling designs for estimation volume and density of coarse woody debris. Nanaimo: Forest Service British Columbia, Technical Report TR-021/2002. Marshall, P.L., Davis, G., LeMay, V.M., 2000: Using Line Intersect Sampling for Coarse Woody Debris. Technical Report TR-003 March, Research Section, Vancouver Forest Region, BCMOF, 34 p. O’Hehir, J., Leech, J., 1997: Logging residue assessment by line intersect sampling. Australian Forestry 60(3): 196–201. Pickford, S.G., Hazard, S.W., 1978: Simulation studies on line intersect sampling of forest residue. Forest Sci. 24(4): 469– 483. Van Wagner, C.E., 1968: The line-intersect method in forest fuel sampling. Forest Sci. 14(1): 20–26. Van Wagner, C.E. 1976: Diameter measurement in the line intersect method. Forest Sci. 22(2): 230–232. Van Wagner, C.E., 1982: Practical Aspects of the Line Intersect Method. Information Report PI-X-12. Petawawa National Forestry Institute, Canadian Forest Service, 11 p. Warren, W.G., Olsen, P.F., 1964: A line intersect technique for assessing logging waste. Forest Sci. 10(3): 267–276. Woldendorp, G., Keenan, R., Barry, S., Spencer, R., 2004: Analysis of sampling methods for coarse woody debris. Forest Ecology and Management 198(1): 133–148.

Croat. j. for. eng. 38(2017)1


Quantitative Estimation of Logging Residues by Line-Intersect Method (33–45)

S. P. Karpachev et al.

Author’s address:

Received: June 1, 2016. Accepted: October 10, 2016. Croat. j. for. eng. 38(2017)1

Prof. Sergey P. Karpachev, PhD.* e-mail: karpachevs@mail.ru Prof. Vjacheslav I. Zaprudnov, PhD. e-mail: zaprudnov@mgul.ac.ru Assoc. prof. Maksim A. Bykovskiy, PhD. e-mail: bykovskiy@mgul.ac.ru Assoc. prof. Evgeny N. Scherbakov, PhD. e-mail: scherbakov@mgul.ac.ru Moscow State Forest University 1st Institutskaya street 1 141005 Mytischi, Moscow region RUSSIA *Corresponding author

45



Original scientific paper

Root Biomass and Morphological Characterization of Energy Willow Stumps Paweł Tylek, Marcin Pietrzykowski, Józef Walczyk, Tadeusz Juliszewski, Dariusz Kwaśniewski Abstract Energy willow plantations are used in cycles of 20–25 years. After such a period of use, or earlier, plantations should be liquidated. In the case of arable land, liquidation of a plantation also means restoration of the original production properties of the soil. In particular, this means: permanent elimination of the possibility of plant regrowth from the above-ground rootstock and root systems and disintegration and mixing of the above-ground rootstock and root systems. The present authors undertook the task of developing a technology for stump removal on energy willow plantations that would have the advantage of lower energy consumption and execution costs than the technologies used so far. The development of a new machine for the disintegration of the above-ground rootstock and root systems requires recognition of the variability of their morphological parameters and their biomass. For that purpose, a head for planting trees was used to sample rootstocks and extract them, and a number of biometric parameters were determined with the division into thickness fractions. The average biomass of the root system of an energy willow shrub with a butt-end of approx. 10 cm in height was 3.1 kg, of which the butt-end and roots with a diameter greater than 30 mm accounted for more than 73%. The vertical and horizontal range of thick roots, which should be ground during plantation liquidation, is small and amounts to approx. 26 and 29 cm, respectively. This justifies the use of machines that work along strips of land during plantation reclamation. Keywords: biomass production, energy willow, land reclamation, stump removal, root system

1. Introduction For several years the production and harvesting of the biomass of fast-growing species has been advocated as a new direction for agricultural production. However, the development of this new agricultural activity is largely determined by economic aspects (Faber et al. 2009, Günther 2005, Stolarski 2006). One of the barriers to the production of biomass from energy plants on a large scale is the large cost of that production. The processes of biomass production from perennial energy plants can be divided into the following stages: establishment of a plantation, use (running a plantation) and liquidation of a plantation (Grzybek 2010, Tworkowski et al. 2010). The costs of the establishment and liquidation of a plantation when related to its duration of use will enable an estimation of the average annual cost of running the plantation. Depending on the type of land on which it is planned to Croat. j. for. eng. 38(2017)1

establish a long-term plantation, several different types of technological activities must be performed. All operations and associated activities are carriers of costs. The costs of production of biomass for energy purposes depend on: the yield size and its price, the area of a field, harvesting technology and organization of work, especially using multi-operational forestry machinery (Di Fulvio 2012, Spinelli et al. 2008, 2009). So far, various authors have described in detail the many aspects of willow plantation establishment costs which, according to various sources, amount to € 1224–2164 per ha (Kwaśniewski 2010, Matyka 2008, McKendry 2002, Szczukowski et al. 2004, 2012). Thus, the costs calculated per year, assuming that a plantation will be used for 25 years, will amount to € 49–87. This accounts for approximately 50% of the costs incurred in Nordic countries (Ericsson et al. 2006). Nearly 30% of the costs are accounted for by the prepara-

47


P. Tylek et al.

Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

tion of seedlings and establishment of the crop. An interesting way to reduce energy and hence the cost of the above mentioned treatments, nearly by half, is an innovative method based on horizontal distribution of stems of a length of 1.2 m at a depth of 5–10 cm (Bergante et al. 2016). The use of modern, high performance machinery, working in fields of higher acreage, can significantly reduce costs (Trzepieciński et al. 2016), as can the use of a single-stage harvest via balloting (Stolarski et al. 2015). The estimated total area of plantations of energy willow in Poland is currently approx. 10–12 thousand hectares (Dubas 2010). These are plantations of varied areas: from small area ones (less than 1 ha) to large scale ones (several dozen hectares). Any long term crop cultivation must be liquidated and renewed or converted to another crop. The decision on liquidation depends on many factors. The most important of these are: the demand for a given raw material, the selling price, the yield per unit area and consequently, above all, the profits. In addition, plantation liquidation is also carried out due to the natural aging of the plants. In the literature, there is little information on the results of studies related to the liquidation of long term energy crop plantations. The information published on this subject in Poland consists of theoretical analyses unsupported by fieldwork. This is due to the fact that the vast majority of established plantations are relatively young (present only for a few years) and do not yet require liquidation (Kwaśniewski et al. 2010). Plantations of energy willow have been in use for 20–25 years (Dubas et al. 2004, Larsson 2006, Lisowski 2010, Szczukowski et al. 2004). The optimum period of cultivation on a given site seems to be 22 years: the first year being an initial one, followed by seven threeyear rotations (Szulc and Dach 2014). After that period, or earlier, a plantation should be liquidated. In the case of agricultural land, stump removal also means restoration of the original properties of the soil. In particular, this is done by: Þ permanent removal of the possibility of plant regrowth from the above-ground parts of the rootstocks and the root systems Þ fragmentation of the above-ground parts of the rootstocks and the roots. One of the stump removal methods consists of fullsurface grinding of rootstocks. Generally, this involves the use of soil cutters, with a working width of 2–2.5 m and a working depth of up to 0.3–0.45 m, mounted on and powered by a tractor with the power of 200–300 kW. The volume of soil which is ground and mixed with

48

the cut portions of the root system amounts to 3–4.5 thousand m3 per 1 hectare. The share of the mass of the root system is very small relative to the mass of the broken up soil and amounts to slightly more than 11%. The low working speed of the tractor machine set and the large energy inputs for the drive of working assemblies make fuel consumption and operation costs very large. Objective problems of field reclamation after the cultivation of energy willow provide a basis on which to formulate the conditions of the possibility to develop a plantation liquidation technology with lower energy consumption than the methods used so far. These conditions are as follows: Þ Limitation of the regrowth of willow stems is

generally associated with the destruction of above-ground parts of rootstocks and the part that is directly below it – therefore, mechanical grinding of the root system throughout the surface of the field (aimed at reducing the possibility of shoot regrowth) is not necessary

Þ Part of the field remaining after the plantation

(where grinding the rootstock and the root system located underneath was performed only in strips) may be used for growing and mechanical harvesting of some plants (e.g. maize)

Þ Leaving the unground root system (outside the

strip of the ground rootstocks) for a few years is enough to weaken it (due to biodegradation) so as to allow for further liquidation with the use of less energy-intensive technologies, e.g. by using disk harrows.

Thus, it appears possible to reduce the energy inputs for the reclamation of plantation fields after cultivation of willow through the use of machines with a smaller working width, which will grind and mix with the soil only such strips of the field whose width corresponds to the main part of the willow root system. In contrast, those parts of the field where the root system is too underdeveloped to produce plant stems may be left intact. The aim of the study was to identify the morphology of the willow root system, which would be the basis for a determination of which part should be ground and which can remain intact. Allocation of the biomass of energy willow roots from the viewpoint of the future development of a machine for plantation restoration will constitute key information for design engineers.

2. Materials and methods The biological material was obtained from a plantation area of approx. 3 ha, established in 2003, situCroat. j. for. eng. 38(2017)1


Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

P. Tylek et al.

Fig. 1 Root ball of an energy willow extracted using a head for transplanting large trees

Fig. 3 Washing the soil off the root system using a pressure washer

Fig. 2 A wrapped root ball – ready for transport

Fig. 4 Cleaned root systems of energy willow

ated along the Vistula river in Kaniów (Silesian Voivodeship). Cuttings had been planted at every 0.5 m, with a spacing width of 0.75 m, with no separate tramline system; the planting amounted to approx. 26,600 plants/ha. The plantation was cultivated extensively: in subsequent years no treatments related to management and fertilization were performed, except for mechanical harvesting every 3 years. The current density is approx. 20,000 plants/ha.

after wrapping in burlap and a steel mesh (Fig. 2), the root balls were transported to a storage area where the root systems were cleaned using a pressure washer (Figs. 3 and 4). This allowed for determination of their mass in fresh state and their geometric parameters characterizing the distribution of willow roots in the soil.

After cutting the willow shoots using a saw, each item was labeled separately for later identification when measuring the amount of shoots and specifying their biomass. The willow rootstocks were collected together with lumps of soil by means of the hydraulic head of an OPTIMAL 1100, aggregated with a front loader, used for transplanting large trees (Fig. 1). The device has 4 hydraulic shovels moving along a curved trajectory, which enable complete separation of the root system from undisturbed soil (Tylek 2008). The applied head allowed for the extraction of root balls with repeated geometrical features, i.e. a volume of 0.34 m3, a width of the upper part of 1.14 m and a depth of the blade range of 0.75 m. Then, Croat. j. for. eng. 38(2017)1

Root biomass and allocation were measured by dividing the following fractions into diameters: fine roots bellow 2.0 mm, medium at 2–8 mm and coarse roots at 8–30 mm. Coarse roots were dug up and extracted by machine, washed, measured and weighed. In order to identify the biomass of fine roots, thirty soil monoliths distributed regularly across the experimental plot were collected using steel cylinders to the depth of 40 cm (Böhm 1985, Pietrzykowski et al. 2010, Pietrzykowski and Woś 2010). The collected monoliths were washed in the laboratory, and roots were extracted. Appropriate root fractions were selected by measuring with calipers to obtain fine root fractions, and then roots were weighed in fresh state, and reweighed after drying at 65°C to determine biomass.

49


P. Tylek et al.

Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

3. Results and discussion Fig. 5 presents characteristic parameters of the root system of energy willow, measured on the prepared roots (Juliszewski et al. 2015). Moreover, the biomass (in dry state) of the rootstocks and the roots themselves was determined after distinguishing the individual thickness of fractions (Fig. 6). Measurement results and characteristics of the above-ground parts of willow shrubs are summarized in Table 1. The height of shoots on the experimental plantation was very balanced, but the average height of 3 year old shrubs was relatively small: below 4.3 meters, which was caused by the very high surface density of the shrubs. With more than 20 thousand shrubs planted on one hectare, the dry biomass harvested had a weight of approx. 49 Mg ha−1 (16.3 Mg ha−1y−1). The obtained crops are slightly more abundant in relation to experimental results obtained in Denmark, Finland, Sweden, Lithuania, Latvia and Estonia, considering the number of shrubs per unit area of the plantation, and are comparable with experiments conducted in the USA (Wang and MacFarlane 2012, Mola-Yudego 2010, Sevel et al. 2012). In turn, the crops were slightly less abundant in Hungary when subjected to intensive mineral and organic fertilization (Mikó et al. 2014). In the above mentioned experiments, a close association was demonstrated between yields and soil properties,

Fig. 5 Morphological parameters of the root system of energy willow, identified during the research; D0 –rootstock diameter at the root collar at ground level, Dg – main root diameter at its thickest point, ZGkrg – vertical range of thick roots, ZGkr – vertical range of roots, ZPkrg – horizontal range of thick roots, ZPkr – horizontal range of roots

50

Table 1 Characteristics of the morphology and biomass of energy willow shrubs Parameter

Mean Min.

Max.

CV, %

Shrub height, m

4.28

3.36

5.01

12.1

Number of offshoots, items

11.3

4

28

47.5

Rootstock diameter at the root collar at ground level, cm

18.00 6.45 30.25

Main root diameter at the thickest point, cm 6.68

33.0

3.45 11.50

22.2

Vertical range of thick roots, cm

26.32 16.50 61.50

30.7

Vertical range of roots, cm

48.59 22.50 83.00

24.6

Horizontal range of thick roots, cm

27.82 5.25 53.50

42.4

Horizontal range of roots, cm

71.24 48.75 90.25

14.5

Biomass of the above-ground part (shoots), kg

2.51

0.99

8.69

64,4

Biomass of the basal part (after cutting the shoots), kg

0.98

0.18

2.84

60.6

Biomass of very thick roots (diameter above 30 mm), kg

1.29

0.41

4.22

61.3

Biomass of thick roots (diameter of 8–30 mm), kg

0.36

0.10

1.35

62.4

Biomass of medium-sized roots (diameter above 2–8 mm), kg

0.36

0.13

0.66

39.1

Biomass of fine roots (diameter below 2 mm), kg

0.11

0.02

0.24

43.8

access to water and the type of maple willow. Differences may exceed 30% in extreme cases. In turn, appropriate fertilization may increase the abundance of crops even more than twofold. It should be emphasized that this plantation was carried out extensively and the location of the plantation along a large river ensured proper moisture conditions. The lack of treatments and chemical protection could have a significant impact on the height of shrubs and the number of shoots, which is usually directly translated into the amount of the yield (Schulz et al. 2016). Small variation also characterized the rootstock diameter at the root collar, measured at the ground level, and the main root diameter measured at its thickest point. The coefficients of variation of these morphological parameters amounted to 33% and 22%, respectively. This is important information with respect to the choice or design of machines for plantation reclamation, as it is these fragments of the rootstock that should be ground the most precisely, while the demand for power to drive the working tools must be Croat. j. for. eng. 38(2017)1


Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

P. Tylek et al.

Fig. 6 Illustration of the root division into thickness fractions; a) up to 2 mm, b) 2–8 mm, c) 8–30 mm adapted so as to destroy the structure of these very fragments. It is true that the recorded maximum diameters of rootstocks and main roots were approx. 70% greater than the average, but this applied only to individual items, growing only on the edge of the plantation or in gaps. The vertical range of thick roots (with a diameter of 30 mm) averaged approx. 26 cm, and in an extreme case maximally 62 cm. Analogous values for the horizontal range are 28 cm and 54 cm, respectively. The results regarding the horizontal range of thick roots were similar to the results obtained from the experiment performed using the same method for age-equal plantations, but established on sandy soil. In turn, the vertical range of thick roots was almost 2-fold lower (Juliszewski et al. 2015). This allows for the assumption that the idea presented in the project concerning the applicability of reclamation of strips of land after cultivation of energy willow may be correct. Although the average range of roots with a diameter of less than 8 mm is approx. 2–2.5 times greater, leaving them unground deep in the soil should not result in reduced efficiency of land reclamation. Although the height of the plantation crops was regular, the number of offshoots of individual shrubs varied considerably and ranged from 4 to 28. Large variability in the number of shoots could result from the lack of treatments and protective actions (Schulz et al. 2016). The coefficient of variation of the shoot biomass was very high: over 64%. Only a slightly lower degree of variation characterized the biomass of the underground parts of the shrubs. Despite the large variation, a high correlation was noted between the biomass of the above-ground and underground parts of the energy willow shrubs (Fig. 7). Croat. j. for. eng. 38(2017)1

Fig. 7 Relationship between biomass of above-ground and underground parts of energy willow shrubs The average biomass (in dry state) of the root system of an energy willow shrub with a butt-end of approx. 10 cm in height was 3.1 kg, of which over 73% was the basal part and roots with a diameter greater than 30 mm. Fig. 8 shows the allocation of biomass in different parts of the plant. In the case of wood harvesting after three growing seasons, shoots accounted for 45% of the biomass. Butt-ends accounted for as much as 18% of the biomass. It should be noted that cutting the plantation lower to the ground, i.e. leaving a butt-end of a height not greater than 5 cm, can increase the yield per hectare by up to approx. 10%. It should also be noted that such a shallow cut of shrubs requires the use of an appropriate tool, which is resistant to blunting in consequence of contact with inor-

51


P. Tylek et al.

Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

al labor and the use of reinforced plows (preferably forest ones). It is, therefore, scarcely possible to regard this as a comprehensive mechanized technology. Currently available economic analyzes of biomass production on willow crop plantations do not include the cost of liquidation of the plantation (Ericsson et al. 2006). Recognizing the morphology of root systems of willow will facilitate the development of a comprehensive, low-input technology for the elimination of the below-ground rootstock, in order to restore agricultural production. Fig. 8 Allocation of biomass (in dry state) on the energy willow plantation ganic contaminants, e.g. a beater system. However, in this case, the danger of damage to a part of the belowground rootstock should be taken into account (Stolarski et al. 2015). The biomass remaining in the soil after harvest is approx. 58 Mg ha−1; and the mass that should be ground during plantation reclamation (basal parts as well as thick and very thick roots) amounts to approx. 49 Mg ha−1. Mechanical liquidation of a plantation may consist of extracting the whole rootstocks from the ground. However, this raises another problem: how to deal with rootstocks largely covered with soil debris, at a rate of 10–15 thousand pieces per 1 ha. Experimental elimination of a 15-year old plantation was carried out based on a field experiment over an area of 0.4 hectares on the Kwidzyńska Plain in the Vistula glacial valley (Stolarski 2006, Stolarski 2009, Stolarski et al. 2008). The experiment had the following stages of work: harvesting energy willow in January (shoots were mowed close to the ground surface); spraying young plants with Roundup in the amount of 7 dm3 ha−1 in the third decade of May; rootstock extraction using a plow in the third decade of July; double harrowing of the plowed field and manual removal of rootstock pieces from the field and their transportation. Harrowing was performed twice, in order to facilitate manual collection of rootstocks and their more precise removal. A similar process has been designed for technology based on spraying with Randap in the amount of 5 dm3 ha−1 and the use of a heavy forestry mulcher, followed by a return to the production of grass plants after waiting at least one growing season (Caslin et al. 2015). However, the chemical-mechanical technology requires multiple agrotechnical treatments, hard manu-

52

4. Conclusions The use of a head for transplanting large trees for the purpose of collecting energy willow rootstocks enabled standardized samples to be obtained containing complete butt-ends and thick roots, appropriate for the selection of the technology and equipment for plantation reclamation; Due to the different number of offshoots (4–28), shrub biomass is characterized by high variability. The biomass of offshoots is strongly correlated with the biomass of the root system; The mean rootstock diameter, when measured on the surface of the soil, amounts to 18 cm and this is three times greater than the diameter of the thickest root; The vertical and horizontal ranges of thick roots (ones with a diameter of more than 8 mm), which should be ground during plantation liquidation, are small and amount to approx. 26 and 29 cm, respectively. This justifies the use of machines that work along strips of land during plantation reclamation; The use of machines that leave lower basal parts (5 instead of 10 cm in height) for the harvesting of energy willow results in an approx. 10% increase in yield per hectare of plantation.

Acknowledgments The present study was conducted within the framework of research project No. PBS2/A8/26/2014 entitled »Developing a new technology and a functional model of a machine for reclamation of land after cultivation of energy willow«, funded by the National Research and Development Center (Poland) within the area of the Applied Research Program. We kindly thanks to Bartłomiej Woś, PhD. from Department of Forest Ecology and Reclamation for field investigation and Justyna Likus-Cieślik, MSc. for laboratory work and graphical working on the figures. Croat. j. for. eng. 38(2017)1


Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

5. References Bergante, S., Manzone, M., Facciotto, G., 2016: Alternative planting method for short rotation coppice with poplar and willow. Biomass and Bioenergy 87: 39–45. Böhm, W., 1985: Metody badania systemów korzeniowych. PWRiL Warszawa, 248 p. Caslin, B., Finnan, J., Johnston, C., McCracken, A., Walsh, L., 2015: Short rotation coppice willow – Best practice guidelines. Agriculture and Food Development Authority, 126 p. Di Fulvio, F., Bergstrom, D., Kons, K., Nordfjell, T., 2012: Productivity and profitability of forest machines in the harvesting of normal and overgrown willow plantations. Croatian Journal of Forest Engineering 33(1): 25–37. Dubas, J.W., 2010: Stan i kierunki rozwoju biomasy dla potrzeb elektroenergetyki polskiej. [W:] Odnawialne źródła energii w świetle globalnego kryzysu energetycznego. Wybrane problemy. Wyd. Difin S.A. Warszawa. Dubas, J.W., Grzybek, A., Kotowski, W., Tomczyk, A., 2004: Wierzba energetyczna – uprawa i technologie przetwarzania. Wyższa Szkoła Ekonomii i Administracji w Bytomiu, 35 p. Ericsson, K., Rosenqvist, H., Ganko, E., Pisarek, M., Nilsson, L., 2006: An agro-economic analysis of willow cultivation in Poland. Biomass and Bioenergy 30(5): 16–27. Faber, A., Kuś, J., Matyka, M., 2009: Uprawa roślin na cele energetyczne. Poradnik. Wyd. W i B Wiesław Drzewiecki. Warszawa. Grzybek, A., 2010: Modelowanie energetycznego wykorzystania biomasy. Wyd. Instytut TechnologicznoPrzyrodniczy. Falenty–Warszawa, 230 p. Günther, F., Sylvia, P., van Velthuizen, H., 2005: Biomass potentials of miscanthus, willow and poplar: results and policy implications for Eastern Europe, Northern and Central Asia. Biomass and Bioenergy 28(2): 119–132. Juliszewski, T., Kwaśniewski, D., Pietrzykowski, M., Tylek, P., Walczyk, J., Woś, B., Likus, J., 2015: Root biomass distribution in an energy willow plantation. Agricultural Engineering 4(156): 43–49.

P. Tylek et al.

McKendry, P., 2002: Energy production from biomass (part 1): overview of biomass. Bioresource Technology 83(1): 37–46. Mikó, P., Kovács, G.P., Alexa, L., Balla, I., Póti, P., Gyuricza, C.S., 2014: Biomass production of energy willow under unfavorable field conditions. Applied Ecology and Environmental Research 12(1): 1–11. Mola-Yudego, B., 2010: Regional potential yields of short rotation willow plantations on agricultural land in Northern Europe. Silva Fennica 44(1): 63–76. Pietrzykowski, M., Socha, J., Woś, B., 2010: Biomasa i przekształcenia systemów korzeniowych sosny zwyczajnej (Pinus sylvestris L.) w warunkach siedliskowych zrekultywowanego wyrobiska i zwałowiska górnictwa odkrywkowego. Sylwan 154 (2): 107–116. Pietrzykowski, M., Woś, B., 2010: The biomass and distribution of pine forest phytocenosis fine roots in the sandy soil and sandy clay loam soil on the reclaimed spoil heap of the Piaseczno Sulphur Mine. Teka Kom. Ochr. Kszt. Środ. Przyr. OL PAN 7: 319–327. Schulz, V., Gauder, M., Seidl, F., Nerlich, K., Claupein, W., Graeff-Hönninger, S., 2016: Impact of different establishment methods in terms of tillage and weed management systems on biomass production of willow grown as short rotation coppice. Biomass and Bioenergy 85: 327–334. Sevel, L., Nord-Larsen, T., Raulund-Rasmussen, K., 2012: Biomass production of four willow clones grown as short rotation coppice on two soil types in Denmark. Biomass and Bioenergy 46: 664–672. Spinelli, R., Nati, C., Magagnotti, N., 2008: Harvesting shortrotation poplar plantations for biomass production. Croatian Journal of Forest Engineering 29(2): 129–139. Spinelli, R., Nati, C., Magagnotti, N., 2009: Using modified foragers to harvest short rotation poplar plantations. Biomass and Bioenergy 33(5): 817–821. Stolarski, M., 2006: Opłacalność uprawy na cele energetyczne. Wyd. 2. Regionalne Forum Energetyki Odnawialnej. Przysiek, 46–48.

Kwaśniewski, D., 2010: Koszty produkcji biomasy z upraw polowych. [W:] Produkcja biomasy na cele energetyczne. Wyd. PTIR. Kraków.

Stolarski, M., 2009: Agrotechniczne i ekonomiczne aspekty produkcji biomasy wierzby krzewiastej (Salix spp.) jako surowca energetycznego. Rozprawa habilitacyjna. Wyd. UWM w Olsztynie.

Kwaśniewski, D., Mudryk, K., Wróbel, M., 2010: Zbiór i likwidacja plantacji energetycznych. [W:] Produkcja biomasy na cele energetyczne. Wyd. PTIR. Kraków.

Stolarski, M., Kisiel, R., Szczukowski, S., Tworkowski, J., 2008: Koszty likwidacji plantacji wierzby krzewiastej. Roczniki Nauk Rolniczych 92: 172–177.

Larsson, S., 2006: Od A do Z o wierzbie energetycznej. Czysta Energia 1: 18–19.

Stolarski, M.J., Krzyżaniak, M., Szczukowski, S., Tworkowski, J., Grygutis, J., 2015: Changes of the quality of willow biomass as renewable energy feedstock harvested with biobaler. Journal of Elementology 20(3): 717–730.

Lisowski, A., 2010: Technologie zbioru roślin energetycznych. Wydawnictwo SGGW, Warszawa, 148 p. Matyka, M., 2008: Opłacalność i konkurencyjność produkcji wybranych roślin energetycznych. Studia i raporty IUNG– PIB. Zeszyt 11. Wyd. Dział Upowszechniania i Wydawnictw IUNG-PIB w Puławach, 113–123. Croat. j. for. eng. 38(2017)1

Szczukowski, S., Tworkowski, J., Stolarski, M., 2004: Wierzba energetyczna. Plantpress, Kraków, 46 p. Szczukowski, S., Tworkowski, J., Stolarski, M., Kwiatkowski, J., Krzyżaniak, M., Lejszner, W., Graban, Ł. 2012: Wieloletnie

53


P. Tylek et al.

Root Biomass and Morphological Characterization of Energy Willow Stumps (47–54)

rośliny energetyczne. Multico Oficyna Wydawnicza, Warszawa, 156 p. Szulc, R., Dach, J., 2014: Kierunki rozwoju ekoenergetyki w polskim rolnictwie. Wyd. Inżynierii Rolniczej. Kraków, 120 p. Trzepieciński, T., Stachowicz, F., Niemiec, W., Kępa, L., Dziurka, M., 2016: Development of harvesting machines for willow small-sizes plantations in East-Central Europe. Croatian Journal of Forest Engineering 37(1): 185–199. Tworkowski, J., Kuś, J., Szczukowski, S., Stolarski, M., 2010: Uprawa roślin energetycznych. [W:] Bocian, P., Golec, T.,

Rakowski, J., (red.). Nowoczesne technologie pozyskiwania i energetycznego wykorzystywania biomasy. Wyd. Instytut Energetyki. Warszawa. Tylek, P., 2008: Maszyny do przesadzania starych drzew. [W:] Integrované ťažbovo-dopravné technológie. Technická Univerzita vo Zvolene, 295–302. Wang, Z., MacFarlane, D.W., 2012: Evaluating the biomass production of coppiced willow and poplar clones in Michigan, USA, over multiple rotations and different growing conditions. Biomass and Bioenergy 46: 380–388.

Authors‘ addresses: Paweł Tylek, PhD.* e-mail: rltylek@cyf-kr.edu.pl Prof. Józef Walczyk, PhD. e-mail: rlwalczy@cyf-kr.edu.pl University of Agriculture in Krakow Institute of Forest Utilization and Forest Technology Al. 29-Listopada 46 31-425 Krakow POLAND Prof. Marcin Pietrzykowski, PhD. e-mail: rlpietrz@cyf-kr.edu.pl University of Agriculture in Krakow Institute of Forest Ecology and Silviculture Al. 29-Listopada 46 31-425 Krakow POLAND Prof. Tadeusz Juliszewski, PhD. e-mail: tadeusz.juliszewski@ur.krakow.pl University of Agriculture in Krakow Institute of Machinery Management, Ergonomics and Production Processes Ul. Balicka 116B 30-149 Krakow POLAND

Received: February 29, 2016 Accepted: May 16, 2016

54

Dariusz Kwaśniewski, PhD. e-mail: Dariusz.Kwasniewski@ur.krakow.pl University of Agriculture in Krakow Institute of Agricultural Engineering and Informatics Ul. Balicka 116B 30-149 Krakow POLAND * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Productivity and Time Consumption of Timber Extraction with a Grapple Skidder in Selected Pine Stands Dariusz Kulak, Arkadiusz Stańczykiewicz, Grzegorz Szewczyk Abstract This paper presents the results of the study on work time and productivity of the John Deere 548G-III grapple skidder operating in pine stands on flat terrain. The research covered three types of treatment: late thinning, removal cutting in openings and clear cutting in openings. The average skidding distance was between 120 and 250 m. In the clear-cut stands, a mean duration of a skidding cycle accounted for 12 minutes, while in the thinned ones it was longer, lasting ca. 16 minutes. The investigators performed an estimate of parameters of multiple regression models, which revealed dependences between: Þ duration of skidding cycles and stand conditions, number of logs per load and skidding distance, and Þ productivity achieved in specific skidding cycles and the above-mentioned variables, as well as wood volume per load. The highest productivity within the productive work time of a skidder was recorded in the stand where removal cutting in openings was performed; it exceeded 14 m3×h-1. In other stands, the efficiency was lower, not greater than 9 m3×h-1. However, differences between the productivity achieved within the productive work time and that recorded for the work place time did not exceed 20%, which indicated that skidding operations in the stands under scrutiny were organised properly. Keywords: timber skidding, work efficiency, modelling of work time, productivity

1. Introduction The volume of timber harvested in Poland has significantly increased in the recent years: from 27 mil. m3 in 2000 to nearly 38 mil. m3 in 2013 (Central Statistical Office 2014), though no more than 10% of wood (ca. 4 mil. m3) had been extracted using technologies based on innovative multi-operational forest machines (Karaszewski et al. 2013). In Poland, skidding of timber is mainly performed with the use of agricultural tractors, 3000 of which operate in Polish forests, and properly equipped forest tractors (skidder type) – ca. 1500 machines (Kocel 2013). With regard to skidders, cable models are predominant, although grapple skidders have proved to be considerably more efficient (Bembenek et al. 2011, Mousavi et al. 2013). Having knowledge of productivity of particular skidding Croat. j. for. eng. 38(2017)1

means, operating under diversified conditions, is essential for an appropriate selection and respective matching of these machines with a certain type of forest stand to assure satisfactory financial results (Acar and Yoshimura 1997, Gallis and Spyroglou 2012, Giefing et al. 2012, Marčeta et al. 2014). One of the ways to achieve this goal is to describe the relations between the productivity of skidding operations and properties of the terrain where skidding works were performed (Horvat 2007), selected characteristics of stands under treatments (Ghaffariyan et al. 2013), skidding means employed (Kluender and Stokes 1994), and skidding distance (Naghdi 2005, Maesano et al. 2013). In respect to Europe, semi-suspended skidding of timber is the most popular method, commonly engaged in the central and southern part of the continent

55


D. Kulak et al.

Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

Table 1 Characteristics of the stands under investigation Stand

A

B

C

14.8

2.6

4.7

50°03’21“N 19°35’29“E

50°03’29“N 19°30’47“E

50°03’57“N 19°35’41“E

Studied area, ha Location – geographical coordinates Type of habitat

Fresh upland deciduous forest Fresh upland deciduous forest Fresh mixed coniferous forest

Tree species composition

10 Pine (Pinus sylvestris L.)

10 Pine (Pinus sylvestris L.)

10 Pine (Pinus sylvestris L.)

Age, years

77

77

102

Stocking index

0.9

0.9

0.7

Crown closure

Broken crown

Open crown

Broken crown

36/27

31/26

37/25

430

188

225

0.65±0.23

0.63±0.14

0.81±0.19

Average diameter at breast height, cm / Average height, m Large timber, m3×ha–1 3

Mean volume of harvested trees ± SD, m

(Gil 2000, Kocel 2013). Results of studies on productivity of skidding operations performed under these conditions have been published in many publications, referring to various felling treatments, such as: thinning (Kluender et al. 2007, Spinelli and Magagnotti 2012, Vusić et al. 2013), clear cutting (Zečić 2005, Kluender et al. 2008), or selective cutting (Sabo and Poršinsky 2005, Behjou et al. 2008). However, there are only few papers dealing with the efficiency of skidding operations performed in stands in various manners by one type of skidder. A better recognition of this issue would enable an identification of conditions, under which the skidder achieves the highest productivity level, and consequently, provides the most profitable economic outcomes. This paper aimed to characterise the duration of a skidding cycle and the productivity achieved by a grapple skidder, the John Deere 548G-III in particular, operating in pine stands in various manners. To be more specific, the research goals covered: Þ characteristics of timber loads and skidding distance Þ characteristics of work time structure Þ c haracteristics of a skidding cycle duration, including the determination whether the impact of the manner of use of a forest stand on this variable is statistically significant Þ constructing a mathematical model that would enable an estimate of a skidding cycle duration Þd etermining the productivity of a skidder operating in the stands under investigation and developing a mathematical model for work efficiency.

56

2. Materials and methods 2.1 Research area The research was carried out in southern Poland, in forest stands where various types of treatments were performed: late thinning (stand A), removal cutting in beech undergrowth (stand B), and clear cutting in openings (stand C). All the trial plots were located on flat terrain, 200– 300 m a.s.l. The major features of the stands studied are listed in table 1.

2.2 Machines and technology engaged in logging operations In the stands under analysis, timber was extracted in the long length system (LLS) (Pulkki 2004). Cutting

Fig. 1 Skidder equipped with a forest grapple observed during the investigations Croat. j. for. eng. 38(2017)1


Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

Table 2 General technical specifications of the John Deere 548G-III grapple skidder Unit

Value

Weight

kg

10,750

Length

mm

6330

Width

mm

2640

Height

mm

3010

Engine – type, power

kW

Diesel, 98

2

Grapple – gripping area in closed position

m

0.7

Cable winch – cable length / pulling force

m/kN

50/156

down and delimbing of trees, which were felled in the direction opposite to the direction of skidding, were performed with the use of chainsaws. Longwood was extracted in the long tree system (LTS), using the semisuspended skidding method. This method of skidding employed the John Deere 548G-III grapple skidder, equipped with a cable winch remotely controlled, as displayed in Fig. 1. The major parameters of the skidder used for extraction in the studied stands are listed in Table 2. The skidder was operated by one person. The operator approached the logs as close as possible to minimise the distance of hauling wood with a winch and facilitate forming loads using a grapple. The skidding operations were performed in autumn and winter, when the soil was dry or frozen, not covered with snow.

2.3 Measurements and calculations During the skidding operations, a time study was carried out by means of a digital stopwatch, with the accuracy of 1 sec. The work time measurements covered durations of all operations that were recorded at the felling site, including: Þ t ravel without a load – relocation of a vehicle without a load from the landing to the felling site. The time measurement started at the moment when the skidder left the landing and entered the operational track. The time measurement ended when the skidder stopped at the first log designated for loading and hauling Þ f orming a load – lifting logs with the use of a grapple (possibly, dragging a log to the skidder using a cable) and relocating between particular logs. The time measurement started at the moment when the skidder stopped at the first log, and ended when the entire load was lifted Croat. j. for. eng. 38(2017)1

D. Kulak et al.

Þ s kidding – travel with a load to the landing. The

time measurement started when the skidder, fully loaded, set on its way to the landing, and ended when the entire load was deposited at the landing Þ l og stacking – relocating logs deposited at the landing, with the use of a log stacker, to align the stacks. The time measurement started at the moment when the entire load was laid at the landing, and ended when the last log was aligned properly Þa ligning ground beams – placing ground beams at the landing in an appropriate manner to create a stack foundation for skidded logs Þd aily maintenance – activities connected with the preparation of the skidder for work before its launching, and refilling the fuel Þd elays – unblocking the cable on a winch drum; Þb reaks – time for meal consumption, physiological needs and rest for the skidder operator. The work efficiency was calculated for the productive work time and the work place time, based on the Classification of Time in forest work study (Acuna et al. 2012). Upon completing every skidding cycle, the load was registered to compute the wood volume and mass. A hauling distance of every load was determined with the use of the Garmin 64S GPS device. In the course of the indoor research, the volume and mass of timber extracted during every skidding cycle were calculated, as well as the basic descriptive statistics referring to volumes of individual loads, work times and productivity. For determining the mass of skidded timber in kilograms, the authors assumed, according to Tomczak and Jelonek (2014), that one cubic meter of freshly felled pine wood weighed 750 kg. The significance of differences in duration of skidding cycles in specific stands was determined using the variance analysis (Fisher’s test), followed by the posthoc LSD test. The multiple regression analysis was used to estimate parameters of the equations that described dependences between the duration of skidding cycles and the work efficiency, and the stand conditions under which the skidding operations were performed, selected features of loads and the skidding distance. Afterwards, the above-mentioned parameters were computed. The significance of particular independent variables was assessed by means of the Student’s t-test, while the Fisher’s test was used for the entire model. For determining the relative significance of certain properties of the model, standardised (normalised) ß (beta) coefficients of regression were assumed as a measure of weights, which enabled the

57


D. Kulak et al.

Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

Table 3 Characteristics of skidder loads, skidding distances and work times within skidding cycles Stand

A

Number of cycles

78

Mean

Sum

Minimum

Maximum

Standard deviation

Number of logs per load, pcs.

3.28

256

1

7

1.38

Load volume, m3

2.20

171.82

0.27

6.28

1.12

Skidding distance, m

197

15,378

18

581

127.59

Work times within skidding cycles, min.

16.1

1254

1.6

40.5

9.2

Number of logs per load, pcs.

3.66

465

1

5

0.60

Load volume, m

2.95

374.50

1.55

4.61

0.65

Skidding distance, m

246

31,350

53

456

98.43

Work times within skidding cycles, min.

12.0

1532

3.4

19.1

2.3

Number of logs per load, pcs.

2.94

353

1

5

0.63

Load volume, m

1.86

223.27

0.72

3.17

0.45

Skidding distance, m

124

15,082

27

246

46.16

Work times within skidding cycles, min.

11.5

1389

1.8

18.9

2.9

3

B

127

3

C

121

3. Results

28 h 12 min 46 s in stand C. Within this time, the skidder completed 326 skidding cycles and hauled 1074 logs with a total volume of about 770 m3 and mass of about 577,500 kg. Detailed characteristics of loads of the skidder and skidding distances are gathered in Table 3.

The time studies performed in the stands under scrutiny covered in total 80 h 2 min 9 s, of which 24 h 49 min 5 s in stand A, 33 h 0 min 58 s in stand B, and

The logs were transported to upper landings, situated close to the stands under investigation; therefore, in all the cases, the minimum skidding distance was

authors to compare factors with various units (Stanisz 2007). All the statistical analyses were performed using the Statistica 10 software.

Fig. 2 Share of operations under analysis within work time

58

Croat. j. for. eng. 38(2017)1


Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

D. Kulak et al.

not excessive. The longest mean skidding distance was Skidding cycle duration = 1.23 – 6.63× a – 1.39 × b + 2.60 × c + 0.03× d ± recorded in stand B, whereas, the shortest in stand C. a – 1.39 × b + 2.60 × c + 0.03× d ± 2.82 (min.) (1) In stands A and C, the Skidding cycle duration = 1.23 mean volume of an individual – 6.63× 3 load amounted to ca. 2 m , while in stand B this value Where: accounted for nearly 3 m3. a zero-one variable of the stand category, taking the Fig. 2 shows the shares of particular elements of value »1« for stand B, and »0« for the other two times within the work time. The diagram revealed legstands ible differences between the stand B and the other two b zero-one variable of the stand category, taking the stands in terms of duration of most of the operations value »1« for stand C, and »0« for the other two under analysis. Forming a load on this trial plot took stands slightly over 25% of the work place time and the times c number of logs per load (pcs.) of skidding, travels without a load and log stacking d skidding distance (m). had higher shares. In stands A and C, the time of forming a load was predominant, taking nearly 50% of the The model of multiple regression function prework place time. What is noteworthy is the surprissented above explained the variability in duration of ingly low, not exceeding 8%, share of times of breaks. skidding cycles well, which was reflected by the value Delays that occurred during the logging operations of coefficient of determination R2, accounting for 73%. were not serious and they were fixed by the skidder The values of β coefficients obtained in the analysis operator as they emerged, due to which their impact indicated that the skidding distance had the greatest on the course of works was insignificant. The share of impact on the skidding cycle duration, followed by the the time of repairs was low, ranging from 0.16% in stand-related factors – the conditions encountered in stand C to 2.5% in stand B. stand B and the number of logs per load. The factors Table 3 also contains characteristics of skidding with the weakest influence on the skidding cycle duracycles, enclosing the times of forming a load, skidding, tion in the above-mentioned model were the condilog stacking and travels without a load (returns). tions under which the wood was hauled in stand C, Variance analysis revealed statistically significant where the mean duration of a cycle was the lowest. differences in mean duration of a skidding cycle beThe values of productivity per hour achieved by tween the stands under investigation (p=0.00; F=21.34). the skidder studied, operating in particular stands, are The LSD test proved that the differences between stand displayed in Fig. 3. A and the other two stands (B and C) were significant. Dependences between the skidding cycle duration and the type of a felling site (stand), the skidding distance and the number of logs per load were displayed by means of the multiple regression model (Eq. 1), the parameters of which are presented in Table 4.

Table 4 Regression analysis – duration of a skidding cycle Equation parameters: R=0.86; R2=0.73; R2revised=0.73; F=221.82; p<0.001; Estimation error = 2.82 Parameters of independent variables: b Absolute term

Standard Coefficients of deviation regression 0.68

1.23

t

p

1.79

0.04

a

–0.59

0.41

–6.63

–15.90 <0.001

b

–0.12

0.42

–1.39

–3.25

c

0.43

0.19

2.60

13.43 <0.001

d

0.62

0.01

0.03

17.49 <0.001

Croat. j. for. eng. 38(2017)1

0.01

Fig. 3 Skidding productivity achieved in particular stands

59


D. Kulak et al.

Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

The skidder achieved the highest productivity in stand B; it exceeded 14 m3×h-1 when calculated within the productive work time. On the other two trial plots, the productivity of skidding ranged from 8 to 9 m3×h–1. The differences in productivity achieved within the productive work time and the work place time of a working shift were not large, which might indicate that the logging operations were well organised and executed.

4. Discussion

Productivity of skidding means used for semisuspended skidding of timber is highly diversified (Mederski et al. 2010). It depends on many factors, among which the most important are the following: the volume of extracted trees, number of logs per load, skidding distance, landform and atmospheric conditions, experience of an operator (Gil 2000, Ozturk and Senturk 2010, Sowa and Szewczyk 2013). Therefore, in The model of multiple regression (Eq. 2), describthe existing literature there are publications reporting ing the efficiency of skidding cycles within the producboth, the higher and the lower levels of skidding effitive work time, took the form presented below: ciency when compared with those presented in this paper; the latter did not exceed 14.2 m3×h-1, when cal3 -1 Productivity of a skidding cycle = 14.79+3.72×a-1.79×b-2.73×cculated -0.03×d+5.08× e ±4.53 (m ×h ) within the productive work time. For instance, Bembenek et al. (2011) stated that in a 140 year old 3 -1 a-1.79×b-2.73×c-0.03×d+5.08×e ±4.53 (m ×h ) (2) = 14.79+3.72× clear-cut beech stand, the productivity of skidding performed by the HSM 904D grapple skidder, over a Where: distance of 200–300 meters, was very high, accounting e wood volume per single load (m3) for 18 m3×h-1 within the productive work time. Howother symbols as in equation 1 above. ever, transport operations took place in a finely accessible area, the landform of which was only slightly Characteristics of the above mentioned model are inclined, with a dense network of skid roads. In compresented in Table 5. parison, the HSM 94 grapple skidder operating in the poplar and pine plantations, though on gently deTable 5 Regression analysis – productivity of a skidding cycle scending slopes, and hauling logs over the similar distance (220–285 m), achieved the productivity hardEquation parameters: R=0.73; R2=0.53; R2revised=0.52; ly exceeding 7 m3×h-1 (Mousavi et al. 2013). Whereas F=71.47; p<0.001; estimation error = 4.53 in higher mountainous locations, the MB Trac 900 skidder, operating on steep slopes in winter, reached Parameters of independent variables: the productivity of merely ca. 6 m3×h-1 (Acar and Dinc Standard Coefficients of 2001). Nevertheless, logging operations in the mounb t p deviation regression tains do not necessarily have to be less efficient as proved by other authors. According to Behjou et al. Absolute term 1.03 14.79 14.36 <0.001 (2008), who analysed skidding of hardwood (beech a 0.30 0.65 3.72 5.66 <0.001 and alder), performed down the slope with a 30% angle of inclination, over a distance of 300 meters, using b –0.14 0.64 –1.79 –2.79 0.01 the Timbejack 450C skidder, recorded that the machine c –0.41 0.38 –2.73 –7.22 <0.001 managed to achieve the productivity of 22 m3×h–1. A skidder of the same type, operating under very similar d 0.73 0.01 –0.03 –11.77 <0.001 conditions (in beech and alder stands, down the slope e –0.57 0.44 5.08 11.47 <0.001 with a mean inclination accounting for 30%), over a distance of 900 meters, achieved only slightly lower productivity of 20.1 m3×h-1 (Lotfalian et al. 2011). The model in question explained the variability in In the studies presented in this paper, the skidding productivity of skidding cycles in slightly over 50%. distance accounted for 250 meters. According to Zečić The F test revealed that the model was statistically siget al. (2010), an increase in the distance of skidding nificant, though there must have been other indepenperformed by a skidder from 200 to 1000 meters would dent variables that affected the productivity of skidresult not only in a decrease in efficiency by ca. 50%, ding in its successive cycles. The values of standardised but also in higher unit costs, which might even grow β coefficients of regression indicated that the skidding by 60%. However, skidding over shorter distances distance had the strongest impact on productivity, folwould not affect the productivity so strongly; accordlowed by the wood volume per load and the number ing to simulations presented by Di Gironimo et al. of logs per load. Stand-related factors affected the pro(2015), extending a distance of semi-suspended skidding from 50 to 200 meters would lower its efficiency ductivity of skidding less significantly.

60

Croat. j. for. eng. 38(2017)1


Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

only by ca. 5%. Thus, it may be assumed that hauling operations in the stands studied were performed within optimal distances, which allowed the skidder to achieve a satisfactory level of productivity. One of the factors that might have influenced the productivity recorded was the volume of individual loads. The operator of the John Deere skidder under analysis did not make a full use of the available capacity of the machine. The volumes of loads quoted in this paper ranged from 1.8–2.9 m3 and they were closer to those recorded in young plantations, where volumes of loads oscillated around 2.2 m3, at a mean volume of a single tree smaller than 0.4 m3 (Mosuavi et al. 2012), rather than to clear-cut stands, where an average load contained nearly 6 m3 wood (Mederski et al. 2010). Nevertheless, it is difficult to forecast how the efficiency of skidding would change if the skidder operator formed larger loads, as this operation took the most of time (up to 50%) on all the trial plots under investigation. Adding more logs to a single load would even increase the share of this work time elements. The stands where skidding took place were not optimal in terms of employing a grapple skidder, which proved to be much more effective while operating on large clear-cut sites. With regard to the latter, the time of forming loads would also be the shortest. However, the studies presented in this paper were carried out in pine stands under thinning and complex felling system, which are typical of Polish forest management. The share of time of forming loads was similar to the data reported in the existing literature only in stand B. Mosuavi et al. (2013) stated that this time element was prevalent in a semi-suspended skidding cycle, and constituted 27% of work time for a grapple skidder, while its share for a cable skidder reached 36%. Sabo and Poršinsky (2005), having analysed skidding performed in the mountains, on slopes of various angles of inclination, estimated that depending on the particular stand, the share of time of forming loads ranged from 30 to 32% within the total time of work. In the clear-cut stands (B and C), the mean duration of a skidding cycle was the same (the differences detected appeared to be statistically insignificant), and accounted for ca. 12 minutes, which corresponded to similar values quoted in the existing literature. According to Mederski et al. (2010), who conducted studies in a clear-cut beech stand, the mean duration of a skidding cycle accounted for 10.9 minutes. In the thinned stand A, this time was longer (ca. 16 minutes), though it was within the range given in the related literature. Admittedly, Brinker et al. (1996) reported that the mean duration of a skidding cycle performed in pine plantations, with the use of the Timberjack 240C grapple skidder, over a distance of 300 meters, amounted to 5 minutes. On the other hand, according Croat. j. for. eng. 38(2017)1

D. Kulak et al.

to Zečić et al. (2010), the mean duration of a skidding cycle over the same distance, employing the Timberjack 240C skidder, though performed in fir and beech stands in the mountains, lasted 35 minutes. Due to a great number of variables determining the duration of skidding and its productivity, a multiple regression analysis is often used for modelling these properties. This research covered an elaboration of models that described the duration and the productivity of skidding cycles within the productive work time, based on a skidding distance, number of logs per load, and wood volume per load. Those variables were used for developing an efficiency model. The models also included other variables defining the treatment performed and stand conditions encountered on a certain felling site. The latter is a variable hardly ever used for modelling of work time and productivity, which is mainly due to the fact that studies on productivity of skidding are usually conducted in stands managed within only one category of felling system. Nevertheless, the field research discussed in this paper was carried out in three stands managed within different felling systems, which enabled the authors to include the variable in question into modelling. A similar approach was taken by Nurminen et al. (2006), whose model of skidding efficiency included, among others, a variable describing the type of cutting. The variables most commonly used in models published in the existing literature referred to the number of extracted logs or load volume and skidding distance (Bolding et al. 2009, Borz et al. 2013, Vusić et al. 2013), in some cases complemented with the inclination of the terrain (Zečić and Marenče 2005, Gholami and Majnounian 2008, Ozturk and Senturk 2010), the number of loads (Ozturk 2010), type of skidder (wheeled, crawler) (Maesano et al. 2013) or temperature and soil conditions at which logging operations took place (Horvat et al. 2007). Borz et al. (2014), apart from the variables mentioned above, developed their models of duration of skidding operations, which included the skidding direction, down or up slope. Upon modelling productivity of logging operations based on wide-spread research conducted in North America, Goychuk et al. (2011) suggested to complement these models additionally with a few variables, such as: season of the year, technological system and human factor related to the manner of work performance and experience of a particular operator of the skidder.

5. Conclusions The John Deere 548G-III grapple skidder, performing skidding of longwood in pine stands, achieved a satisfactory level of productivity and, therefore, the authors recommend to engage machines of this type

61


D. Kulak et al.

Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63)

for skidding under conditions in question, in particular, in clear-cut pine stands. Logging operations in thinned stands, where a skidder is forced to maneuver among trees remaining in the stand, are expected to be less efficient. Therefore, an essential role in forest stands within younger age classes is played by a properly designed and, what is even more important, suitably utilised network of forest tracks. The models of multiple regression obtained in this studies, defining the productivity and the time consumption of skidding, may appear very useful in practice since they enable relatively reliable estimate of productivity of logging works, performed not only in stands like those studied, but also in stands with varying conditions, partly similar to those discussed in this paper (Papyrakis and Gerlagh 2007). Taking into account an extremely vast incidence of pine species in Europe and Asia (Sinclair et al. 1999), the aspect of studies presented here appears to be very significant.

6. References Acar, H.H., Dinc, B., 2001: An investigation of winter harvesting on steep terrain in forestry. Turkish Journal of Agriculture and Forestry 25(2): 139–147. Acar, H.H, Yoshimura, T., 1997: A study on the productivity and cost of cable logging in Turkey. Journal of Forest Research 2(4): 199–202. Acuna, M., Bigot, M., Guerra, S., Hartsough, B., Kanzian, C., Kärhä, K., Lindroos, O., Magagnotti, N., Roux, S., Spinelli, R., Talbot, B., Tolosana, E., Zormaier, F., 2012. Good practice guidelines for biomass production studies. In: Magagnotti, N., Spinelli, R. (eds.), CNR IVALSA, Sesto Fiorentino, 51 p. Behjou, F.K, Majnounian, B., Namiranian, M., Dvořák, J., 2008: Time study and skidding capacity of wheeled skidder Timberjack 450C in Caspian forests. Journal of Forest Science 54(4): 183–188. Bembenek, M., Mederski, P.S., Erler, J., Giefing, D.F., 2011: Results of large-size timber extracting with a grapple skidder. Acta Scientarum Polonorum Silvarum Colendarum Ratio et Industria Lignaria 10(3): 5–14. Bolding, M.C., Kellogg, L.D., Davis, C.T., 2009: Productivity and costs of an integrated mechanical forest fuel reduction operation in southwest Oregon. Forest Products Journal 59(3): 35–46. Borz, S.A., Ignea, G., Popa, B., 2014: Modelling and comparing timber winching performance in windthrow and uniform selective cuttings for two Romanian skidders. Journal of Forest Research 19(6): 473–482. Borz, S., Dinulică, F., Birda, M., Ignea, G., Ciobanu, V., Popa, B., 2013: Time consumption and productivity of skidding Silver fir (Abies alba Mill.) round wood in reduced accessibility conditions: a case study in windthrow salvage logging form Romanian Carpathians. Annals of Forest Research 56(2): 363–375.

62

Brinker, R.W., Klepac, J.F., Stokes, B.J., Roberson, J.D., 1996: Effect of tire size on skidder productivity. In Proceedings of a Joint Conference of the Canadian Woodlands Forum, the Canadian Pulp and Paper Association, and the International Union of Forest Research Organizations, 9–11 p. Central Statistical Office, 2014: Forestry. Statistical information and elaborations, Warszawa, 324 p. Di Gironimo, G., Balsamo, A., Esposito, G., Lanzotti, A., Melemez, K., Spinelli, R., 2015: Simulation of forest harvesting alternative processes and concept design of an innovative skidding winch focused on productivity improvement. Turkish Journal of Agriculture and Forestry 39(2): 350–359. Gallis, C., Spyroglou, G., 2012: Productivity linear regression models of tree-length harvesting system in natural coastal Aleppo pine (Pinus halepensis l.) forests in the Chalkidiki area of Greece. Croatian Journal of Forest Engineering 33(1): 115– 123. Ghaffariyan, M.R., Naghdi, R., Ghajar, I., Nikooy, M., 2013: Time prediction models and cost evaluation of cut-to-length (CTL) harvesting method in a mountainous forest. Small-scale Forestry 12(2): 181–192. Gholami, M.J., Majnounian, B., 2008: Productivity and cost of wheeled skidder in Hyrcanian Forest. International Journal of Engineering Science 2(3):99–103. Giefing, D.F., Bembenek, M., Gackowski, M., Grzywiński, W., Karaszewski, Z., Klentak, I., Kosak, J., Mederski, P.S., Siewert, S., 2012: Evaluation of thinning operations in older pine stands. Research methods. Nauka Przyroda Technologie 6(3): 1–23. Gil, W., 2000: Ground extraction of wood by skidders from the aspect of the logging code of practice. Sylwan 144(1): 59–74. Goychuk, D., Kilgore, M.A., Blinn, C.R., Coggins, J., Kolka, R.K., 2011: The effect of timber harvesting guidelines on felling and skidding productivity in Northern Minnesota. Forest Science 57(5): 393–407. Horvat, D., Zečić, Ž., Šušnjar, M., 2007: Morphological characteristics and productivity of skidder ECOTRAC 120V. Croatian Journal of Forest Engineering 28(1): 11–25. Karaszewski, Z., Giefing, D.F., Mederski, P.S., Bembenek, M., Dobek, A., Stergiadou, A., 2013. Stand damage when harvesting timber using a tractor for extraction. Leśne Prace Badawcze 74(1): 27–34. Kluender, R.A., Stokes, B.J., 1994: Productivity and costs of three harvesting methods. Southern Journal of Applied Forestry 18(4): 168–174. Kluender, R., Lortz, D., McCoy, W., Stokes, B., Klepac, J., 1997: Productivity of rubber-tired skidders in southern pine forests. Journal of Forest Products 47(11/12): 53–58. Kluender, R., Lortz, D., McCoy, W., Stokes, B., Klepac, J., 1998: Removal intensity and tree size effects on harvesting cost and profitability. Forest Products Journal 48(1) :54–59. Kocel, J., 2013: Firmy leśne w Polsce. CILP. DGLP Warszawa, 333 p. Lotfalian, M., Moafi, M., Foumani, B.S., Akbari, R.A., 2011: Time study and skidding capacity of the wheeled skidder Timberjack 450C. Journal of Soil Science and Environmental Management 2(7): 120–124. Croat. j. for. eng. 38(2017)1


Productivity and Time Consumption of Timber Extraction with a Grapple Skidder ... (55–63) Maesano, M., Picchio, R., Monaco, A.L., Neri, F., Lasserre, B., Marchetti, M., 2013: Productivity and energy consumption in logging operation in a Cameroonian tropical forest. Ecological Engineering 57: 149–153. Marčeta, D., Petković, V., Košir, B., 2014: Comparison of two skidding methods in beech forests in mountainous conditions. Nova mehanizacija šumarstva 35(1): 51–62. Mederski, P.S., Bembenek, M., Jörn, E., Dieter, D.F., Karaszewsk, Z., 2010: The enhancement of skidding productivity resulting from changes in construction: grapple skidder vs. rope skidder. In: Cavalli, R., Grigolato, S., (eds) Proceeding of FORMEC 2010, Forest Engineering: Meeting the Needs of the Society and the Environment, July 11–14, Padova, Italy, 7 p. Mousavi, R., Nikooy, M., Esmailinezhad, A., Ershadfar, M., 2012: Evaluation of full tree skidding by HSM-904 skidder in patch cutting of aspen plantation in Northern Iran. Journal of Forest Science 58(2): 79–87. Mousavi, R., Nikoye Seyahkal, M., Naghdi, R., 2013: Comparison of timber skidding using two ground-based skidding systems: grapple skidding vs. cable skidding. International Journal of Forest, Soil and Erosion 3(3): 79–86. Naghdi, R., Rafatnia, N., Jalali, G.H.S., 2005: A survey of the efficiency of Timberjack 450C Wheeled skidder in Shafaroud forests in Guilan province. Iranian Journal of Natural Resources 57(4): 675–687. Nurminen, T., Korpunen, H., Uusitalo, J., 2006: Time consumption analysis of the mechanized cut-to-length harvesting system. Silva Fennica 40(2): 335–363. Ozturk, T., 2010: Productivity of MB Trac 900 tractor at beech stands on mountainous areas in Blacksea region. African Journal of Agricultural Research 5(1): 028–033. Ozturk, T., Senturk, N., 2010: Productivity and Time Studies of MB Trac 900 Tractor at Beech Stands on Mountainous Areas in Turkey. Baltic Forestry 16(1): 132–138.

D. Kulak et al.

Pulkki R.E., 2004. Glossary of forest harvesting terminology. [online], www.flash.lakeheadu.ca. Sabo, A., Poršinsky, T., 2005: Skidding of fir roundwood by Timberjack 240C from selective forests of Gorski Kotar. Croatian Journal of Forest Engineering 26(1): 13–27. Sinclair, W.T., Morman, T.J.D., Ennos, R.A., 1999: The postglacial history of Scots pine (Pinus sylvestris L.) in western Europe: evidence from mitochondrial DNA variation. Molecular Ecology 8(1):83–88. Sowa, J.M., Szewczyk, G., 2013: Time consumption of skidding in mature stands performed by winches powered by farm tractor. Croatian Journal of Forest Engineering 34(2): 255–265. Spinelli, R., Magagnotti, N., 2012: Wood extraction with farm tractor and sulky. Estimating productivity, cost and energy consumption. Small-scale Forestry 11(1): 73–85. Stanisz, A., 2007: Przystępny kurs statystyki z zastosowaniem STATISTICA PL na przykładach z medycyny. Tom 2. Modele liniowe i nieliniowe. Kraków: StatSoft Polska, 868 p. Tomczak, A., Jelonek, T., 2014: Green density of Scots pine (Pinus sylvestris L.) sapwood coming from selected stands north-western Poland. Forestry Letters 107(6): 5–9. Vusić, D., Šušnjar, M., Marchi, E., Spina, R., Zečić, Ž., Picchio, R., 2013: Skidding operations in thinning and shelterwood cut of mixed stands–Work productivity, energy inputs and emissions. Ecological Engineering 61: 216–223. Zečić, Ž., Marenče, J., 2005: Mathematical models for optimization of group work in harvesting operation. Croatian Journal of Forest Engineering 26(1): 29–37. Zečić, Z., Krpan, A.P.B., Vukušić, S., 2005: Productivity of C Holder 870 F tractor with double drum winch Ingland 4002 in thinning beech stands. Croatian Journal of Forest Engineering 26(1): 49–57.

Papyrakis, E., Gerlagh, R., 2007: Resource abundance and economic growth in the United States. European Economic Review 51(4): 1011–1039.

Zečić, Ž., Vusić, D., Prka, M., Klepac, S., 2010: Uticaj nagiba traktorskog puta na proizvodnost traktora Timberjack 240C pri privlačenju drvnih sortimenata u prebornim šumama. Šumarski list 134(3/4): 103–114.

Received: February 22, 2016 Accepted: July 22, 2016

Authors’ address: Dariusz Kulak, PhD. * e-mail: rlkulak@cyf-kr.edu.pl Arkadiusz Stańczykiewicz, PhD. e-mail: rlstancz@cyf-kr.edu.pl Grzegorz Szewczyk, DSc. e-mail: rlszewcz@cyf-kr.edu.pl University of Agriculture in Krakow Faculty of Forestry Institute of Forest Utilization and Forest Technology Department of Forest and Wood Utilization Al. 29 Listopada 46 31–425 Krakow POLAND * Corresponding author

Croat. j. for. eng. 38(2017)1

63



Original scientific paper

Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations: a Preliminary Case Study Pierre Ackerman, Chloe Williams, Simon Ackerman, Carla Nati Abstract Recent and increasing use of mechanized cut-to-length (CTL) operations in South Africa has been associated with greater diesel and lubricant requirements than was previously the case with motor-manual or semi-mechanized activities, placing a strain on the environmental and economic sustainability of operations. This case study explores diesel and lubricant consumption of a typical CTL pine saw timber operation, taking into account the stand and terrain factors, with the aim of setting baselines for these consumption rates as well as carbon emissions. Data analyzed was provided by Bosbok Ontginning, a contractor based in Mpumulanga, throughout their clear-fell operations over the 49 compartments from May 2014 to June 2015. The mean diesel consumption rate was found to be 0.64 l m-3 and 0.38 l m-3, while the lubricant consumption was 0.08 l m-3 and 0.03 l m-3 for the harvester and the forwarder, respectively. Carbon emissions from diesel were found to account for less than 1% of the carbon stored in the harvested timber. Statistical analysis showed that tree size, stand density and ground condition were not significant predictors of either diesel or lubricant consumption. Prior research suggests that other factors not included in this study (i.e. machine characteristics, operator habits and productivity) may have a more pronounced effect on diesel consumption. Future studies should therefore be conducted to analyze these factors within South African conditions as well as explore stand and terrain conditions more specifically and over more diverse stand and terrain conditions as well as machine types. Keywords: diesel, lubricant, emissions, carbon, cut-to-length

1. Introduction Commercial forestry has seen an increase in the use of mechanized harvesting in the past few decades (JirouĹĄek et al. 2007). Cut-to-length (CTL) logging, involving a harvester and a forwarder, is one system of harvest that can be fully mechanized (Holtzscher and Lanford 1997, Nurminen et al. 2006). Although it has been used and studied extensively on a global scale, mechanized CTL is a fairly new technology in South Africa. Internationally, mechanized harvesting has contributed to increasing productivity, improving conditions for forest workers and decreasing the demand for manpower in forest operations (Holtzscher and Lanford 1997). However, it has also increased fuel and oil requirements (Athanassiadis 2000, Berg and Karjalainen 2003). Both the fuel consumed by large harCroat. j. for. eng. 38(2017)1

vesting machines as well as the oils and lubricants that they require not only present an expense that should ideally be minimized, but also contribute to emissions (Markewitz 2006, Cosola et al. 2016). This is important because carbon emissions, notably CO2, have been linked to a variety of negative environmental consequences, such as the greenhouse gas effect, acidification, oxidant formation as well as negative health impacts (Athanassiadis 2000). This trend, coupled with global concerns over climate change, makes further investigation into how mechanized harvesting contributes to carbon emissions a priority. In his analysis of emissions from different fuel and oil types, Athanassiadis (2000) found that harvesters consume 1167 l of diesel per 1000 m3 of wood processed, while emitting 4.22–4.25 tons of CO2. He also found

65


P. Ackerman et al. Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72)

forwarders to consume 935 l of diesel per 1000 m3, emitting 3.52–3.55 tons of CO2. Studies have predicted fuel consumption rates between 1.4 l m–3 (Ghaffariyan and Sessions 2012) and 2.0 l m-3 (Sambo 2002). Furthermore, a study by Berg (1997) estimated emissions from harvesting and forwarding combined in a clear cutting system in Sweden to be approximately 3 kg CO2 m–3, whereas Klvač and Skoupý (2009) reported 8.58 kg CO2 m–3 in their study of a clearfelling operation in Ireland. Past studies have shown that both fuel consumption and emissions are affected by many factors, including operator characteristics, stand and terrain variables, as well as machine specifics (Berg 1997, Athanassiadis 2000, Klvač and Skoupý 2009, Ghaffariyan et al. 2015, Cosola et al. 2016). In fact, in their literature review on the carbon footprints of different management regimes, Cosola et al. (2016) found that operations in plantations tended to produce lower emissions due to easier access and working conditions. Although mechanized harvesting is a source of carbon emissions, forestry has been shown to have the potential to play a role in stabilizing atmospheric CO2 as trees sequester carbon into their biomass (Berg and Karjalainen 2003, Tavoni et al. 2007, Cosola et al. 2016). In many countries, carbon sequestration is used to offset greenhouse gas emissions and, if correctly managed, forested land can pool carbon in plant biomass, in organic litter, and sometimes, in soil (Dixon et al. 1994, Jandl et al. 2007). Wood products are an especially stable pool of carbon (Jandl et al. 2007, England et al. 2013, Levasseur et al. 2012). In their life cycle assessment of carbon in wood products harvested from Australian plantations, England et al. (2013) found that the carbon stored in logs that were sustainably harvested nearly offset the amount of carbon released through burning, harvesting and transporting the product. In South Africa, plantations have been found to offset approximately 3.8% of carbon emitted by the country (Christie and Scholes 1995). However, few studies have been conducted on the carbon balances of forest operations in South Africa. Those that have been conducted tend to focus either on machine emissions or on carbon storage in biomass exclusively. As such, this study aims to determine the carbon balance of South African plantations by assessing the emissions associated with mechanized CTL harvesting and comparing these to the carbon stored in the harvested logs. Further, it will determine first (for South African operations) and baseline estimates for mechanized CTL diesel and lubricant consumption with the view of more precise machine and harvesting systems costing. The study will also explore some of the stand and

66

terrain factors that could affect these rates since they present an environmental and economic cost. Objectives: ÞE stimate diesel and lubricant consumption, CO2

emissions and carbon stored in harvested timber related to the case-study that can form baseline estimates for machine costing and harvest planning

ÞD etermine whether tree size, stand density and/

or ground condition are significant predictors of diesel and lubricant consumption rates for the harvester.

2. Materials and methods 2.1 Data collection Data used in this study was provided by Bosbok Ontgnning operating typical CTL pine saw timber clear-felling operations. The contractor is based in the Mpumulanga region of South Africa, an area characterized by cool, dry winters and hot, wet summers (Louwa and Scholes 2002). Mean annual temperature is approximately 14 ºC to 19 ºC, while mean annual rainfall is between 840 mm and 1670 mm and soils are typically ferralitic or podzolic (Louwa and Scholes 2002). Bosbok Ontgnning’s historical records (outlining data recorded by on-board computer systems and costing archives) were used to obtain values concerning volumes harvested and the hours worked in each compartment over the period from May 2014 to June 2015 (14 months). Further, average monthly data was provided, from which diesel consumption volumes were drawn. The protocol for diesel and lubrication consumption data gathering was as follows. At the end of each shift, a service truck with a fuel bowser dispensed diesel and lubrication. The fuel was administered via a fuel meter from the bowser (the metering system was calibrated weekly) and this volume was recorded via a job card for each machine. Seeing that these are scheduled services (daily or shift), machine hours were read and included to the job-card. Information contained on the job-card was then captured to the machine records, which were in turn made available to the authors. Using the calculated machine utilization figures and the volume produced, fuel consumption per PMH or volume was calculated. Machine utilization rates were calculated based on a time study conducted in their operations over three 8 hour shifts according to standard procedures for South African forestry, outlined in Ackerman et al. Croat. j. for. eng. 38(2017)1


Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72) P. Ackerman et al.

occurred between May and October was assumed to be dry, moist between March and May, and wet from November to February. From this, the applicable ground condition was derived using the classification provided by the National Terrain Classification for Forestry (Erasmus 1994), with a rating of 1 equating to »very good« and 5 being »very poor«. Operators had similar experience and were deemed adequately trained for the operations. All machines were advanced in operating hours (Table 1).

(2014). This involved timing the machines as they worked, breaking down their activities into productive work time and delays (accounted for if they were over 30 seconds) in order to better understand the efficiency of the operation. Bosbok Ontginning operates its CTL harvesting activities using two harvesters and two forwarders concurrently (Table 1). For the purposes of this study, it was assumed that the total volume of wood cut by harvester I was extracted by forwarder I and that the wood cut by harvester II was extracted by forwarder II since the harvesters were not always working in the same compartments. In total, 49 compartments were harvested (806.8 ha).

Based on Erasmus’ (1994) national classification standards, ground conditions can range from very good (1) to very poor (5), ground roughness scale extends from smooth (1) to very rough (5), and slope class ranges from level (1) to very steep (7).

The areas studied were planted with Pinus patula and had similar terrain, which was classified using the National Terrain Classification for Forestry (Erasmus 1994). Compartments were characterized by low ground roughness and minimal slopes. The mean stand density was 328 stems ha-1 (SD=76), average tree volume was 1.05 m3 (SD=0.24) and all compartments had received their final thinning. Compartments differed mostly in terms of their ground condition, a measure of the strength of the soil and its trafficability when it is either wet, moist or dry (Table 2). The soil moisture level (i.e. wet, moist or dry) was estimated based on average weather conditions for the time of year in which harvesting occurred. Based on precipitation trends reported in (Louw 1997), harvesting that

2.2 Data analysis Calculations were modeled based on the average values found for both harvesters and both forwarders, thus representing a CTL system using only one machine of each type. Productivities per productive machine hour (i.e. excluding delays) and per scheduled machine hour (i.e. including delays) were calculated according to Ackerman et al. (2014) based on the volumes and working hours provided by the contractor. Since diesel consumption was provided, emissions were calculated using equation (1) developed by the Environmental Protection Agency (2008).

Table 1 Machine and operator specifications relating to harvesters and forwarders studied Harvester I

Harvester II

Forwarder I

Forwarder II

Make and model

John Deere 759JH

John Deere 759JH

John Deere 1710D

John Deere 1710D Eco III

Engine power, kW

179.7

179.7

160

160

Age (in April 2015), PMH

19,095

11,348

18,723

5196

Average operator experience (years)

2.25

3

2.5

2.5

Table 2 Stand and terrain characteristics of harvested compartments, grouped based on ground condition class (standard deviations are shown in brackets) Ground condition

Ground roughness

Slope class

Age, years

Stand density, stems ha-1

Tree volume, m3

Sample size, n

1

1.05 (0.15)

1.53 (0.24)

23.01 (1.49)

287.05 (60.44)

1.15 (0.21)

19

2

1.00 (0.00)

1.54 (0.28)

22.97 (1.28)

402.69 (109.00)

0.87 (0.23)

7

3

1.50 (0.41)

1.75 (0.20)

23.69 (0.49)

284.33 (64.34)

1.12 (0.36)

3

4

1.00 (0.00)

1.55 (0.14)

23.22 (0.94)

343.17 (94.64)

1.07 (0.17)

15

5

1.29 (0.50)

1.57 (0.11)

22.89 (1.91)

327.16 (38.96)

1.16 (0.26)

19

Croat. j. for. eng. 38(2017)1

67


P. Ackerman et al. Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72)

Emissions =

Fuel × FD × FO × CO2(m.w.) C(m.w.)

Where:

analysis was only possible for the harvester and not for the forwarder. Analysis, including basic statistics, correlation analysis and an analysis of covariance (ANCOVA) using Generalized Linear Modelling (GLM) were conducted using STATISTICA 12.6 software (StatSoft, Tulsa). The generalized linear model approach was used as the desired model has multiple predictors of differing sample sizes and the GLM allows for normality and homoscedasticity assumptions to be validated by using Mallows’ cp. Once validated, prediction from the model is the same as that of a simple linear regression.

(1)

Fuel daily diesel volume consumed, l FD carbon content of diesel, 0.731757 kg C l-1 (EPA 2008) FO fraction of diesel oxidized, assumed to be 1.00 (EPA 2008) CO2(m.w.) C(m.w.)

conversion factor for C to CO2 based on their molecular weights, 3.6667 g CO2 g C-1 (EPA 2008)

In order to focus the results of this analysis on the harvesting operation itself, emissions from other phases of the forestry operation (such as secondary transport and processing) were not included. Further, emissions related to preparation of the site, such as road construction, were excluded. In addition, carbon produced during the production phases of diesel, lubricants and harvesting equipment used were not considered in this study.

3. Results The machines harvested and forwarded a total of 255,594 m3 and worked 13,767 SMH over the course of the data collection period. The average productivity of harvesters was 54.13 m3 PMH-1, while forwarders had an average productivity of 45.92 m3 PMH-1 (Table 3). Availability, utilization and productivity figures for the machines can be found in Table 3.

Estimates of carbon storage in round wood logs were attained based on a modified version of Christie and Scholes’ (1995) equation (2). Cp = Vk × ph × Fcarbon

Table 3 The average availability, utilization and productivity of the harvester and forwarder used as well as the entire mechanized CTL system

(2)

Mechanical availability Utilization Mean productivity

Where: Cp Vk pk

amount of carbon stored in timber products, Mg volume of harvested wood timber, m3 density of air dried timber product, Mg m-3 (Malan 2012) Fcarbon fraction of oven-dry mass that is carbon, assumed to be 0.5 (Christie and Scholes 1995)

Statistical analysis aimed to determine whether tree size, stand density and ground condition are significant predictors of diesel and lubricant consumption. However, the available data from Bosbok Ontiginning’s historical records were limiting. Due to unbalanced and incomplete block design, relevant

%

%

m3 PMH-1

Harvester

74.52

68.84

54.13

Forwarder

91.87

78.52

45.92

Full CTL system

83.19

73.68

50.02

On average, harvester I and II consumed 0.64 l m–3 or 23.55 l SMH–1 of diesel and 0.08 l m–3 or 2.62 l SMH–1 of lubricant (Table 4). Forwarders consumed less diesel and lubricant with a rate of 0.38 l m–3 or 13.45 l SMH–1 and 0.03 l m–3 or 1.09 l SMH–1, respectively (Table 4). Further, CO2 emissions of the harvesters from diesel

Table 4 Diesel and lubricant consumption as well as emission estimates (calculated based on EPA (2008)) for the harvester and forwarder used in the mechanized CTL system Diesel consumption lm Harvester

Diesel consumption -1

Lubricant consumption

Lubricant consumption -1

CO2 emissions -1

CO2 emissions

l SMH

-3

lm

l SMH

kg SMH

kg m-3

0.64

23.55

0.08

2.62

63.18

1.71

Forwarder

0.38

13.45

0.03

1.09

36.08

1.02

Full CTL system

1.02

37.00

0.11

3.71

99.26

2.73

68

-3

Croat. j. for. eng. 38(2017)1


Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72) P. Ackerman et al.

were calculated to be 1.71 kg m–3 and of the forwarders they were calculated to be 1.02 kg m–3. Meanwhile, the carbon stored in the harvested logs was calculated to be 260 kg m–3. As such, throughout the study period, carbon emissions from CTL operation were 342,811 kg, while carbon stored in the harvested wood volume was 48,796,931 kg. This translates to emissions from the CTL operation accounting for less than 1% carbon being stored in the harvested wood. As mentioned above, data concerning the forwarder was found to be insufficient to conduct any meaningful statistical tests. For the harvester, correlation analysis of the predictor variables showed average tree volume and stand density (Table 2) to be significantly correlated (p<0,01), but their correlation was not high enough for them to be considered collinear (Pearson r=0.55) (Carey 2012). As such, a GLM was conducted on diesel consumption with stand density, tree size and ground condition as potential predictors (Table 2). For statistical analysis, ground conditions ranging from 1 to 3 (classified according to Erasmus (1994)) were grouped as »good« and conditions 4 and 5 were considered »poor«. Results found that these predictors were not significantly related to diesel consumption. There was actually only a small decrease in the mean diesel consumption rate in good ground conditions versus poor ground conditions (0.71 l m-3 and 0.67 l m-3, respectively). Another GLM was conducted between lubricant consumption and stand density, tree size and ground condition. This analysis also showed that there were no statistically significant relationships between the predictor variables and the rate of lubricant consumption by the harvester. In the case of lubricant, the mean rate of consumption (0.07 l m-3) is identical when comparing good and poor ground conditions.

4. Discussion The emissions from harvesting accounted for less than 1% of the carbon stored in the harvested wood, which is similar to findings in Berg and Karjalainen (2003) and England et al. (2013). The mean diesel consumption values found in this study, which were 0.64 l m–3 for the harvester and 0.38 l m-3 for the forwarder (Table 4), are within the range of those reported in prior studies, although on the lower end (Athanassiadis et al. 1999, Athanassiadis 2000, Berg and Karjalainen 2003, Klvač et al. 2003, Nordfjell et al. 2003, Ghaffariyan et al. 2015). Emissions followed a similar trend as the results of this study fell on the lower end of the range of estimates reported in earlier literature (Berg 1997, Berg and Karjalainen 2003, Klvač and Skoupý Croat. j. for. eng. 38(2017)1

2009). For instance, Berg (1997) estimated that emissions from mechanized clear-cutting were 3 kg m–3 and Dias et al. (2007) estimated 3.12 kg m–3 and 2.31 kg m–3 for harvesters working on Eucalypt and Maritime pine, respectively, as well as 2.43 kg m–3 for forwarders. Meanwhile, this study shows an emission rate of 2.73 kg m–3. This may be due to easier working conditions in the South African plantations studied compared to prior research locations, which mainly took place in natural forests. As noted by Cosola et al. (2016), emissions from harvesting natural stands tend to be higher than from plantations. However, these rates are difficult to directly compare since the studies involved machines with different engines and of various ages, which may affect emissions and efficiency. Few studies have reported lubricant consumption rates, but the ones found in this study (0.08 l m–3 for the harvester and 0.03 l m–3 for the forwarder) are approximately twice as high as those reported in Athanassiadis (2000). According to Athanassiadis et al. (1999), harvesters tend to consume twice as much lubricant as forwarders due to the complexity of the machine and potential (and in some cases frequent) hydraulic hose breakages. Since a variety of factors influence both diesel consumption and emissions, it is quite likely that factors specific to each study had sizeable effects on the obtained results. In fact, studies have found that machine and engine characteristics, terrain conditions as well as operator habits can significantly affect diesel consumption and thus, CO2 emissions (Athanassiadis et al. 1999, Athanassiadis 2000, Makkonen 2004, Klvač and Skoupý 2009, Cosola et al. 2016). However, the results showed that site conditions, notably tree size, stand density and ground condition, had no significant effects on the rate of diesel or lubricant consumption. This contrasts Cosola et al. (2016) review, in which they found that tree volume affected fuel consumption through changes in productivity because harvesting larger trees usually entailed using larger, more powerful machines. Although no studies report the effects of tree size on diesel consumption rates, many have found that diesel consumption is affected by productivity (Nordfjell et al. 2003, Cosola et al. 2016). In comparison to prior time study data (Kellogg and Bettinger 1994, Nurminen et al. 2006, Jiroušek et al. 2007, Eriksson and Lindroos 2014), each machine relatively high productivity within its respective range (54.13 m3 PMH–1 for the harvester and 45.92 m3 PMH–1 for the forwarder) (Table 3) may help to explain their fairly low levels of diesel consumption. Accordingly, it would be expected that both tree size and stand density might only affect diesel and lubricant consumption if the difference in conditions was substantial

69


P. Ackerman et al. Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72)

enough to affect productivity. However, the tree sizes varied minimally in the harvested compartments (Table 2), as can be expected in pine clear-felling age trees, and thus, did not likely affect machine productivity. Although Nordfjell et al. (2003) and Cosola et al. (2016) found that the distance travelled does have an effect on the volume of diesel consumed by a forwarder, this study found that stand density, which could be associated with travelling distance (since the harvester would likely move a greater distance between felling locations to harvest trees that are spaced further apart), did not affect the harvester’s diesel consumption. However, not only were the actual distances travelled not measured in this study, but the GLMs were only conducted on the harvester, which may have a different work pattern than the forwarder. Nonetheless, our results suggest that stand density should not necessarily be considered a proxy for distance travelled in future studies of harvester diesel consumption rates. However, it is likely that an efficient work pattern that minimizes unnecessary travel would also reduce fuel consumption, as noted by Cosola et al. (2016). In terms of terrain factors, some studies have reported that the difficulty of the work being performed affects the rate of emissions from the machine (Berg 1997, Berg and Karjalainen 2003, Nordfjell et al. 2003). Results here show a slight decrease in terms of diesel consumption when comparing harvesting in good ground conditions versus poor ground conditions (0.71 l m-3 and 0.67 l m-3, respectively), which is the opposite trend to that reported by Nordfjell et al. (2003), who noted that difficult conditions translate to lower productivities and lower efficiencies. The difference here is minimal, however, and thus may be due to another factor that was not studied. Markedly, analysis in this study was limited by the data available (small sample sizes and incomplete blocking). As such, it would be useful to collect more data on this operation to further and more accurately assess the effects of these variables. Other factors may have more substantial effects on diesel use and allocation to volume produced. In general, past research has shown that it is the machine itself, the operator’s habits, the amount of time that the machine is working, its productivity, and the entire management approach used for the stand that are most important when considering diesel consumption rates (Athanassiadis et al. 1999, Berg and Karjalainen 2003, Cosola et al. 2016). Machine characteristics seem to be particularly important. Machine size can affect consumption, with larger machines having lower rates of consumption (Athanassiadis et al. 1999) and fuel type changing emissions by up to 80% (Athanassiadis 2000). This may help to explain why the factors ana-

70

lyzed in this study were not significant in determining the diesel or lubricant consumption rates. It is recommended that further research be conducted to investigate the relationships between machine as well as operator factors and diesel consumptions in South Africa. In addition, data presented here only represents a snapshot of typical mechanized CTL operations in South Africa. Terrain conditions were only represented by averages and, ideally, data concerning fuel consumptions should be collected for a longer period of time. Studies using remote sensing technology, such as LiDAR, could be useful in better assessing the specific relationships between terrain characteristics and lubricants, diesel as well as emissions, as has been done with studies assessing the relationships between productivity and slope (Alam et al. 2012, Alam et al. 2013, Strandgard et al. 2014). It may also be useful to conduct further research to discover which terrain factors (such as slope and ground roughness), if any, do have significant effects on diesel consumption rates for mechanized CTL operations in South Africa.

5. Conclusions In this study, carbon emissions from mechanized CTL were found to only represent a small fraction of the carbon stored in harvested saw timber, as has been reported in studies on mechanized harvesting in other countries. In fact, emissions from diesel were calculated to be 2.73 kg m-3 and carbon stored in logs was approximated at 260 kg m-3, so emissions accounted for less than 1% of the carbon stored. The mean diesel consumption rate, found to be 1.02 l m-3 (0.64 l m-3 for the harvester and 0.38 l m-3 for the forwarder), and the lubricant consumption rate of 0.11 l m-3 (0.08 l m-3 for the harvester and 0.03 l m-3 for the forwarder), were on the lower end of the range of values found in prior studies. However, these can be considered baseline figures for mechanized CTL in South African plantation conditions, potentially useful for future machine costing and harvest planning. Further, the results of the GLMs conducted found that tree size, stand density and ground condition, grouped as either good or poor, were not significant predictors of either diesel consumption or lubricant consumption. This may be due to confounding effects of the machine productivity on the calculated diesel and lubricant consumption rates. When comparing to prior research, it is also evident that other factors, such as the machine itself, its engine specifics, the operator’s techniques and work habits, have a more pronounced effect on diesel consumption and thus, carbon emissions, which were not explored in this study due to the lack of specific Croat. j. for. eng. 38(2017)1


Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72) P. Ackerman et al.

data. Markedly, the data used in this study only represents average values over a relatively short period of time. It would be useful to conduct longer term studies with more specific data to further analyze the work condition factors that affect diesel consumption and emissions, both important in terms of making mechanized CTL environmentally and economically sustainable.

Acknowledgments The authors acknowledge and appreciate the willingness of Bosbok Ontginning and York to provide data and stand data.

6. References Ackerman, P., Gleasure, E., Ackerman, S., Shuttleworth, B., 2014: Standards for Time Studies for the South African Forest Industry. (Accessed 17 March 2015) Available at: http:// www.forestproductivity.co.za/?page_id=678. Alam, M.M., Strandgard, M.N., Brown, M.W., Fox, J.C., 2012: Improving the productivity of mechanised harvesting systems using remote sensing. Australian Forestry 75(4): 238– 245. Alam, M., Acuna, M., Brown, M., 2013: Self-levelling fellerbuncher productivity based on lidar-derived slope. Croatian Journal of Forest Engineering 34(2): 273–281. Athanassiadis, D., Lidestav, G., Wästerlund, I., 1999: Fuel, hydraulic oil and lubricant consumption in Swedish mechanized harvesting operations. Journal of Forest Engineering 10(1): 59–66. Athanassiadis, D., 2000: Energy consumption and exhaust emissions in mechanized timber harvesting operations in Sweden. Science of the Total Environment 255(1): 135–143. Berg, S., 1997: Some aspects of LCA in the analysis of forestry operations. Journal of Cleaner Production 5(3): 211– 217.

Dias, A.C., Arroja, L., Capela, I., 2007: Carbon dioxide emissions from forest operations in Portuguese eucalypt and maritime pine stands. Scandinavian Journal of Forest Research 22(5): 422–432. Dixon, R.K., Solomon, A.M., Brown, S., Houghton, R.A., Trexier, M.C., Wisniewski, J., 1994: Carbon pools and flux of global forest ecosystems. Science 263(5144): 185–190. England, J.R., May, B., Raison, R.J., Paul, K.I., 2013: Cradleto-gate inventory of wood production from Australian softwood plantations and native hardwood forests: Carbon sequestration and greenhouse gas emissions. Forest Ecology and Management 302: 295–307. EPA, 2008: Direct Emissions from Mobile Combustion Sources. EPA – United States Environmental Protection Agency Climate Leaders, Washington DC, 31 p. Erasmus, D., 1994: National Terrain Classification System for Forestry: Version 1.0. Institute for Commercial Forestry Research, ICFR Bulletin 11/94. Pietermaritzburg, South Africa. Available online at: http://www.forestproductivity.co.za/wpcontent/uploads/2014/07/Terrain-Classification-SystemForm.pdf. Eriksson, M., Lindroos, O., 2014: Productivity of harvesters and forwarders in CTL operations in northern Sweden based on large follow-up datasets. International Journal of Forest Engineering 25(3): 179–200. Ghaffariyan, M.R., Apolit, R., Kuehmaier, M., 2015: Analysis and control of fuel consumption rates of harvesting systems: a review of international studies. Industry Bulletin-Australian Forest Operations Research Alliance 15: 1–4. Ghaffariyan, M.R., Sessions, J., 2012: Comparing the efficiency of four harvesting methods in a blue gum plantation in south-west Western Australia. CRC for Forestry Harvesting Operations Bulletin 29, 4 p. Holtzscher, M., Lanford, B., 1997: Tree diameter effects on cost and productivity of cut-to-length systems. Forest Products Journal 47(3): 25–30.

Berg, S., Karjalainen, T., 2003: Comparison of greenhouse gas emissions from forest operations in Finland and Sweden. Forestry 76(3): 271–284.

Jandl, R., Lindner, M., Vesterdal, L., Bauwens, B., Baritz, R., Hagedorn, F., Johnson, D.W., Minkkinen, K., Byrne, K.A., 2007: How strongly can forest management influence soil carbon sequestration? Geoderma 137(3): 253–268.

Carey, G., 2012: Chapter 11: GLM: multiple predictor variables. (Accessed 27 November 2015) Available at: http:// psych.colorado.edu/~carey/qmin/qminChapters/QMIN11GLM_Multiple_Predictors.pdf.

Jiroušek, R., Klvač, R., Skoupý, A., 2007: Productivity and costs of the mechanised cut-to-length wood harvesting system in clear-felling operations. Journal of Forest Science 53(10): 476–482.

Christie, S.I., Scholes, R.J., 1995: Carbon storage in eucalyptus and pine plantations in South Africa. Environmental monitoring and assessment 38(2–3): 231–241.

Kellogg, L., Bettinger, P., 1994: Thinning productivity and cost for a mechanized cut-to-length system in the northwest Pacific coast region of the USA. Journal of Forest Engineering 5(2): 43–54.

Cosola, G., Grigolato, S., Ackerman, P., Monterotti, S., Cavalli, R., 2016: Carbon footprint of forest operations under different management regimes. Croatian Journal of Forest Engineering 37(1): 201–217. Croat. j. for. eng. 38(2017)1

Klvač, R., Skoupý, A., 2009: Characteristic fuel consumption and exhaust emissions in fully mechanized logging operations. Journal of Forest Research 14(6): 328–334.

71


P. Ackerman et al. Diesel Consumption and Carbon Balance in South African Pine Clear-Felling CTL Operations ... (65–72) Klvač, R., Ward, S., Owende, P.M., Lyons, J., 2003: Energy audit of wood harvesting systems. Scandinavian Journal of Forest Research 18(2): 176–183. Levasseur, A., Lesage, P., Margni, M., Brandão, M., Samson, R., 2012: Assessing temporary carbon sequestration and storage projects through land use, land-use change and forestry: Comparison of dynamic life cycle assessment with ton-year approaches. Climatic Change 115(3–4): 759–776. Louw, J.H. 1997: A site-growth study of Eucalyptus grandis in the Mpumalanga escarpment area. Southern African Forestry Journal 180(1): 1–14. Louwa, J.H., Scholes, M., 2002: The influence of site factors on nitrogen mineralization in forest soils of the Mpumalanga escarpment area: South Africa. The Southern African Forestry Journal 193(1): 47–63.

eds.). Southern African Institute for Forestry, South Africa, 621–638 p. Markewitz, D., 2006: Fossil fuel carbon emissions from silviculture: impacts on net carbon sequestration in forests. Forest Ecology and Management 236(2): 153–161. Nordfjell, T., Athanassiadis, D., Talbot, B., 2003: Fuel consumption in forwarders. International Journal of Forest Engineering 14(2): 11–20. Nurminen, T., Korpunen, H., Uusitalo, J., 2006: Time consumption analysis of the mechanized cut-to-length harvesting system. Silva Fennica 40(2): 335–363. Sambo, S.M., 2002: Fuel consumption for ground-based harvesting systems in western Canada. FERIC Advantage report 3, 12 p.

Makkonen, I., 2004: Saving fuel in mechanized forestry operations. Forest Engineering Research Institute of Canada, Pointe-Claire, QC. Internal Report IR-2004-08, 10 p.

Strandgard, M., Alam, M., Mitchell, R., 2014: Impact of slope on productivity of a self-levelling processor. Croatian Journal of Forest Engineering 35(2): 193–200.

Malan, F., 2012: Solid wood properties and qualities of South African grown pine and eucalypt species. In: South African Forestry Handbook (Bredenkamp, B., Upfold, S.,

Tavoni, M., Sohngen, B., Bosetti, V., 2007: Forestry and the carbon market response to stabilize climate. Energy Policy 35(11): 5346–5353.

Author’s address: Prof. Pierre Ackerman, PhD. * e-mail: packer@sun.ac.za Chloe Williams e-mail: chloe.c.williams@hotmail.com Department of Forest and Wood Science Faculty of AgriSciences Stellenbosch University Private Bag X1, Matieland Stellenbosch 7602 Simon Ackerman, MSc. e-mail: Simon.Ackerman@icfr.ukzn.ac.za Institute of Commercial Forestry Research PO Box 100281, Scottsville 3209 SOUTH AFRICA

Received: November 11, 2015. Accepted: June 2, 2016.

72

Carla Nati, PhD. e-mail: nati@ivalsa.cnr.it CNR-Ivalsa Via Madonna del Piano 10 I-50019 – Sesto Fiorentino (FI) ITALY * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) in Iranian Caspian Forests Farzam Tavankar, Mehrdad Nikooy, Rodolfo Picchio, Amireslam Bonyad, Rachele Venanzi Abstract Caucasian alder is a large tree and one of the commercial species in the Caspian forests. We investigated the wound healing rate (WHR) and compared the diameter growth (DG) of 66 wounded and 66 unwounded alder trees 15 years after selected logging in these forests. The conditions of the wounds after 15 years were as follows: 56.1% had closed, 25.7% were open, and 18.2% had decayed. The mean WHR was 4.95 cm2/yr, ranging between 0 and 17.3 cm2/yr, and DG of wounded trees was 13.3% lower than in unwounded trees. The WHR and DG were related to the size, intensity and location of the wound, stem diameter (diameter at breast height; DBH), and ratio of wound size to stem basal area. The WHR in middle-aged trees was more than in young and older trees. The reduction in DG of wounded alder trees was only observed in the DBH range below 58 cm. Wounds that were larger than 100 cm2 in area significantly reduced diameter growth of alder trees. Height of wound from ground level had significant negative effects on WHR and DG. Closed wounds had no significant effect on DG, but open or decayed wounds reduced DG by 13.8% and 34.3%, respectively. 72.7% of total logging wounds were caused by skidding operations where 20.8% of them were decayed, and reduced DG by 12.2%. Selective logging needs more careful planning of roads, skid trails, and winching corridors. Keywords: Alder tree, Caspian forests, logging wound, selection cutting, wound occlusion

1. Introduction Logging damage to residual trees during selection cutting may lead to serious economic losses in terms of timber quality at the final harvest (Kiser 2011), wood losses of damaged trees, and tree growth reduction (Vasiliauskas 2001). A number of residual trees are damaged during each selective logging operation in forests (Camp 2002, Picchio et al. 2012, Tavankar et al. 2015a). The Iranian Caspian (also called Hyrcanian) forests are managed as a mixed uneven aged high forest with single and group selective cutting regime. Tree wounding is the most common type of logging damage, representing more than 90% of the total damage (Marchi et al. 2014, Tavankar et al. 2013). Frequency of wounded trees and intensity of wounds during logging operations can have detrimental imCroat. j. for. eng. 38(2017)1

pacts on stand growth and forest sustainability. This depends on several factors such as: logging system (Bragg et al. 1994, Spinelli et al. 2010, Marchi et al. 2014), logging machines (Han and Kellogg 2000), logging season (Limbeck-Lilienau 2003), skill of workers (Wallentin 2007, Nikooy et al. 2010), harvest intensity (Fjeld and Granhus 1998, Gullison and Hardner 1993, Behjo 2014, Tavankar et al. 2015a), ground slope (Tavankar et al. 2015a), stand density (Picchio et al. 2012), and design of the extraction trails (Gullison and Hardner 1993, Nikooy et al. 2012, Danilović et al. 2015). Logging wounds may decrease the quality of residual trees and increase stand mortality through insect and disease infestation (Han and Kellogg 2000). Wounding can cause stem deformities and significant losses of the final crop volume and value (Meadows 1993, Lo Monaco et al. 2015). Logging wounds on re-

73


F. Tavankar et al.

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

sidual trees often become an input port for fungal decay (Vasiliauskas 2001), especially in wounds that are near the ground level or root wounds (Bettinger and Kellogg 1993, Camp 2002). The ability of trees in occlusion of their logging wounds, not only depends on the time elapsed from wound occurrence, it also depends on the site conditions, tree species, tree age, and wound characteristics (Vasiliauskas 2001, Tavankar et al. 2015b). Normally, the required time for healing of logging wounds in fast-growing tree species is lower than in slow-growing tree species (Vasiliauskas and Stenlid 2007). Wound characteristics such as size, location, and intensity are the main factors in wound-healing ability and diameter growth of trees (Tavankar et al. 2015b). Wounding season was also reported as another factor that has effects on wound healing (Limbeck-Lilienau 2003). The amount of decayed logging wounds decreased as the wound height increased (Han et al. 2000). Safeguard ecological and production aspects of forest ecosystems, such as this studied, can result in limiting logging damages to residual trees. This must remain a major objective in selection managing of forests, because the quality of timber starts in the forest. In the Iranian Caspian forest, the future of logging wounds and the effect of wounds on diameter growth of alder trees is unclear. In order to improve logging methods, more knowledge about the long term impact of forest operations is needed (Whitman et al. 1997, Tavankar and Bonyad 2014). The objectives of this study were to: Þ study the characteristics of logging wounds in alder trees Þ investigate healing rate of logging wounds in alder trees over a period of 15 years after logging damage Þ investigate the effect of logging wounds on diameter growth of alder trees Þ find some possible relations between wound characteristics (size, location, and intensity), healing rate and diameter growth of alder trees.

2. Materials and methods 2.1 Study area This study was conducted in the Caspian forests of Iran. These forests are located in the north of the country in south coast of the Caspian Sea, extending from the coastal area to an elevation of 2800 m on the northern aspects of the Alborz mountain belt, and cover an

74

area of about 2 million hectares (Poorzadi and Bakhtiari 2009). The Iranian Caspian forests are rich in biodiversity, and include a lot of stand types and about 80 woody species (Marvie-Mohadjer 2006). These forests are natural and broadleaf. Moreover, they are the only commercial forests in Iran. The study area is located in four adjacent parcels in district 1 of the Asalem Nav watershed in the Iranian Caspian forests. The Nav watershed is located between 37°38’34’’ to 37°42’21’’N and 48°48’44’’ to 48°52’30’’E. Elevation in the study area ranged from 950 to 1430 m a.s.l. The average rainfall ranged from 920 to 1250 mm per year, with the heaviest precipitation in summer and fall. The average daily temperature ranges from a few degrees below 0°C in December, January, and February, and up to +25°C during summer. The soil of the study site is classified as a brown forest (Alfisols) and well-drained. The texture of the soil ranges from clay loam to loamy. The original vegetation of this area is an uneven-aged mixed forest dominated by Fagus orientalis Lipsky (55%) and Carpinus betulus L. (28%), with the companion species Alnus subcordata C.A. May (8%), Acer platanoides L. (4%), Acer cappadocicum Gled. (3%), Ulmus glabra Huds. (1%), and Tilia begonifolia Steven (1%) (Tavankar et al. 2015b). In these forests, selective cutting is the most common silvicultural method. Harvest trees were marked before the cutting operation in July 1999. During December 1999 and January 2000, the marked harvest trees, scattered in the study area, were felled, bucked and topped by chain saw, at a merchantable height or 20 cm diameter under bark (DUB). During April and May 2000, logs were winched from the felling site (downhill) to the back of a skidder on a skid trail (uphill) by cable of the skidder winch. In the final phase of primary transportation, logs were skidded to roadside landings by a Timber-jack 450 C wheeled skidder. The mass of the skidder was 9.8 t, and its width and length were 3.8 and 6.4 m, respectively. The harvesting system was the cut to (commercial) length by groundbased skidding. In particular, timber was extracted mostly in logs of long-length (7.8 m), or seldom in logs of short-length (5.2 m). The DBH of harvested trees ranged between 20 and 135 cm, with an average of 68.2 cm. The skidder drive was limited to the constructed skid trails. The skid trails were planed and constructed before the felling season. The skidder group included a driver, a chaser, and a feller. The winching operation was controlled manually. Tree density, stand basal area, and growing stock above 10 cm diameter at breast height (DBH) before and after selective cutting are shown in Table 1. Croat. j. for. eng. 38(2017)1


Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Table 1 Main dendrometric parameters of the stand before and after selective cutting treatment Stand Density, stem/ha 2

Basal area, m /ha 3

Volume, m /ha

Before cutting

After cutting

Harvested

270.4

255.2

15.2

27.3

23.0

4.3

205.7

182.9

22.8

at wound occurrence time (in 2000) were calculated using Eq. 3 and Eq. 4, respectively. DBH 2 − DBH1 t

(1)

UDG − WDG UDG

(2)

WS1 − WS2 t

(3)

WS SBA

(4)

DG =

RDG =

2.2 Data collection and analysis

Immediately after logging (year 2000), damage caused by mechanical means to residual trees was assessed by systematic sample plots. The grid dimensions were 100×100 m and the plot area was circular at 0.1 ha, and all wounded alder trees (83 stems) were identified, numbered and marked. The position of each damaged tree was also identified on a topographical map using the global positioning system (GPS). The following parameters were recorded for each wounded tree: DBH and diameter at wound height (DWH) measured by dendrometric caliper in mm; wound intensity (i.e. type of damaged tissues: bark, phloem, and wood fibers); cause of wounding (i.e., felling or winching); and the location and size of the wounds. The cause of wound was determined on the basis of wound characteristics such as position, size, type (horizontal or parallel) and intensity. The wound size was determined by measuring the maximum length and width with a ruler (to a ±0.5 mm accuracy) and calculating the ellipsoid surface area (Picchio et al. 2011). Wound sizes were then classified into 4 classes: <25, 25–100, 100–200, and >200 cm2. The position of the wound (average height from the ground) was determined with a tape measuring the distance between the wound center and the ground. The location of the wounds was recorded in 3 classes, <0.3, 0.3–1, and >1 m (Limbeck-Lilenau 2003, Nikooy et al. 2010). Near to each wounded tree, an unwounded alder with similar characteristics (i.e. DBH, height, vitality, crown class of all the trees: i.e. dominant, codominant, subdominant, etc.) was selected and measured, as a control tree. After 15 years (in 2015), 66 pair of trees (wounded and control trees) were identified in the study area. The DBH and condition of wounds were reexamined and classified in three types: closed, open and decayed (Han et al. 2000, Tavankar et al. 2015b). The 15 year period diameter growth of wounded and unwounded trees was calculated using Eq. 1 (Clark and Clark 1992). Reduction of diameter growth was calculated using Eq. 2, wound healing rate (WHR), and the ratio of wound size to stem basal area (RSA)

Where:

Croat. j. for. eng. 38(2017)1

F. Tavankar et al.

WHR =

RSA =

DG diameter growth, mm/yr DBH1 diameter at breast height at the start of interval, mm DBH2 diameter at breast height at the end of interval, mm t time interval between two measurements, years RDG reduction of diameter growth UDG unwounded diameter growth WDG wounded diameter growth WHR wound healing rate, cm2/yr WS1 wound size at the start of interval, cm2 WS2 wound size at the end of interval, cm2 WS wound size, cm2 SBA stem basal area, cm2 After checking for normality (KolmogorovSmirnov test) and homogeneity of variance (Levene test), paired t-test was applied to compare means of DG in wounded and unwounded trees. ANOVA and Duncan tests were used for the effect of wound characteristics on WHR and DG. A nonparametric Chisquared test of contingency tables was applied to determine whether significant differences existed among the number of each wound condition (closed, open, and decayed) and wound intensities (bark, phloem, and wood) for different wound characteristics (Eq. 5). Tests were not conducted if the expected frequency in any cell of the contingency table was <5. Where:

k

(Oij − Eij )2

i=1

Eij

X2 = ∑

(5)

Oij sample number of the ith row and the jth column in the contingency table Eij theoretical number of the ith row and the jth column in the contingency table, and the degree of freedom, df = (k—1) ´ (r–1) Regression analysis was applied to test the following relations between:

75


F. Tavankar et al.

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Table 2 Main average wound characteristics (mean ± SD) referred to year 2000 and statistical analysis results. The data were detailed for cause and reported also in total Cause of wounds

n

Wound size, cm2 * a

Wound intensity

Wound height, m * a

Bark, %

Phloem, %

Wood, %

Felling

18

200.9±172.4

1.61±0.84

33.3

44.5

22.2

Skidding

48

55.7±48.5b

0.29±0.11b

2.1

25.0

72.9

Total

66

95.3±103.3

0.65±0.44

10.6

30.3

59.1

* Difference letters in columns indicate significant differences between the means at a<0.05 by Duncan test

Þ DG and DBH for both damaged and undamaged trees Þ DG and RSA Þ WHR and RSA Þ WHR and DBH Þ WHR and DG. All analyses were performed using SPSS 19 (IBM, NY, USA).

3. Results These wounded alder trees constituted 18.3% of the residual alder trees and 1.54% of the total residual trees in the sample plots. 27.3% of total logging wounds were caused by the felling operation and 72.7% were caused by the skidding operation (Table 2). The mean size and height of felling wounds were significantly greater (p<0.01) than those of skidding wounds. The percentage of intensive wounds by winching operation was more than the percentage of intensive wounds by felling operation, so 12 wounds (25%) of all winching wounds occurred with phloemdamaged intensity, 35 wounds (72.9%) occurred with wood-damaged intensity, and only one wound occurred with bark-damaged intensity. About a third of all felling wounds (33.3%) occurred with bark-damaged intensity, 8 wounds occurred with phloem-damaged intensity, and only four wounds (22.2%) occurred with wood-damaged intensity. The frequency of wounds for DBH classes by cause of wound is shown in Table 3. From the analysis of all the wounds, 7.6% were found on trees with DBH<20 cm, 34.8% on trees with DBH ranging from 20 to 40 cm, 33.3% on trees with DBH ranging from 40 to 60 cm, 15.2% on trees with DBH ranging from 60 to 80 cm, and 9.1% on trees with DBH >80 cm. In each DBH class the frequency of skidding wounds was higher than the frequency of felling wounds. Results of the secondary assessment (in 2015) showed that 56.1% of the wound had closed, 25.7%

76

were open, and 18.2% of the wound had decayed (Table 4). The results of ANOVA tests showed that the wounds condition had significant effects on the WHR (F=292.9; P<0.001) and WDG (F=9.6; P<0.001). The average WHR in closed wounds was significantly greater (p<0.01) than the average WHR in open and decayed wounds. The average of unwounded diameter growth (UWDG; control trees) was estimated to 6.48±1.6 mm/yr, and the average of wounded diameter growth (WDG) was 5.62±1.70 mm/yr (a reduction of 13.3%). Paired samples t-test indicated that there were no significant differences between DG of wounded and unwounded alder trees, but DG of decayed wounds was significantly lower than the DG of unwounded alder trees. Wound conditions (2015) in relation to wound characteristics (2000) are shown in Table 5. The highest percentages of closed wounds were observed for wounds <25 cm2 (81.8%), wounds with a bark-damaged intensity (100%), wounds on the stem >1 m (88.9%), and wounds from the felling (61.1%). The highest percentages of open wounds were observed in the wounds >200 cm2 (50%), and the highest percentages of decayed wounds were observed in the wounds of 100–200 cm2 (44.4%). The Chi-squared tests showed that the wound size, and wound intensity had significant effects on wound conditions (p<0.01). The wound location Table 3 Absolute and percentage wound frequencies for each DBH class by cause of wound DBH, cm

Cause of wound

<20

20–40

40–60

60–80

>80

n

%

n

%

n

%

n

%

n

Felling

2

40.0

7

30.4

4

18.2

3

30.0

2

Skidding

3

60.0

16

69.6

18

81.8

7

70.0

4

Croat. j. for. eng. 38(2017)1


Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Table 4 Averages (±SD) of wound healing rate (WHR), wounded diameter growth (WDG) and reduction in diameter growth (RDG) for each wound condition referred to year 2015 and statistical analysis results. The data were detailed for typology of wound condition and reported also in total Wound condition

Frequency n

%

WHR*

WDG*

2

cm /yr

RDG

mm/yr a

% a

Closed

37

56.1

8.3±2.8

6.5±0.9

0.3±0.1

Open

17

25.7

1.2±0.3b

5.6±1.1b

13.8±3.4

b

c

Decayed

12

18.2

0.3±0.1

4.3±1.1 **

34.3±8.0

Total

66

100

4.95±3.23

5.62±1.70

13.3±5.2

** Significant difference with the mean diameter growth of unwounded trees at a<0.01 by paired t test * Difference letters in columns indicate significant differences between the means at a<0.05 by Duncan test

(height from ground level) had also significant effects on wound conditions (p<0.05), but the cause of wounds had no significant effect on wound conditions. The highest wound healing rate (WHR) was observed in the wounds with a bark-damaged intensity (10.32 cm2/yr), and the lowest wound healing rate was observed in the wounds >200 cm2. The ANOVA tests showed that all independent variables (size, intensity, location and cause of wounds) had significant effects on WHR (p<0.01). With the increase of wound size and wound intensity, WHR decreased, but with the increase of wound height from ground level, WHR increased. The highest percentages of reduction in diameter growth (RDG) were observed in the wounds >200 cm2 (22.6%), wounds with a damaged wood intensity (11.7%), wounds at <0.3 m (14.4%), and wounds from the skidding agent (12.2%). The results of paired samples t test showed that the average of diameter growth in wounded trees (WDG) with wound sizes of 100–200 cm2 (5.52 mm/yr, RDG=14.8%) was significantly lower (p<0.01) than the diameter growth of unwound­ed trees (UWDG=6.48 mm/yr). The results of paired samples t tests also indicated that the ­average values of diameter growth in wounded trees with wounds >200 cm2 (5.01 mm/yr, RDG=22.6%), phloem (5.83 mm/yr, RDG=10.1%), and wood inten­ sities (5.72 mm/yr, RDG=11.7%), in heights <0.3 m (5.55 mm/yr, RDG=14.4%), and wounds from skidding agents (5.69 mm/yr, RDG=12.2%) were significantly lower than the diameter growth of unwounded trees. However, diameter growth of wounded trees Croat. j. for. eng. 38(2017)1

F. Tavankar et al.

with wounds <25 cm2 (6.33 mm/yr, RDG=2.3%), bark intensity (6.44 mm/yr, RDG=0.6%), position 0.3–1 m (6.3 mm/yr, RDG=2.7%), position >1 mm (6.39 mm/yr, RDG=1.4%), and wounds from felling agents (6.23 mm/yr, RDG=3.9%) have no significant difference compared to the average of UWDG. The regression analyses showed that: the wound healing rate (WHR) decreased by increasing the ratio of wound size to basal area (Fig. 1); trees with DBH of 40–60 cm had the highest WHR (Fig. 2); the WHR increased by increasing diameter growth of alder trees (Fig. 3); growth of wounded and unwounded trees was similar when DBH was greater than 58 cm (Fig. 4); diameter growth decreased by increasing ratio of wound size to basal area (Fig 5).

Table 5 Wound conditions (2015) in relation to wound characteristics (2000) and statistical analysis results. The data were detailed for wound size, intensity, position and cause Wounds in year 2000 Characteristics n

Wound conditions in year 2015 Closed Open %

%

Chi-squared

WHR* cm2/yr

Decayed %

Size, cm2

155.9**

<25

11

81.8

18.2

8.13a

25–100

31

74.2

19.3

6.5

6.81a

100–200

18

22.2

33.3

44.4

1.45b

>200

6

16.7

50.0

33.3

1.01b

Intensity

98.2**

Bark

7

100

10.32a

Phloem

20

75.0

20.0

5.0

7.16b

Wood

39

38.5

33.3

28.2

3.00c

Height from ground level, m

43.8**

<0.3

42

47.6

31.0

21.4

3.69c

0.3–1

15

60.0

20.0

20.0

5.88b

>1

9

88.9

11.1

9.94a

Cause of wound

3.7

Felling

18

61.1

27.8

11.1

9.84a

Skidding

48

54.2

25.0

20.8

3.24b

** Significant difference with the mean diameter growth of unwounded trees at a < 0.01 by Chi-squared test * Difference letters in columns indicate significant differences between the means at a < 0.05 by Duncan test

77


F. Tavankar et al.

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Fig. 1 Non-linear regression of wound healing rate (WHR) in relation to the ratio of wound size to basal area (RSA) of alder trees, detailed statistical results shown in Table 6

Fig. 3 Non-linear regression of wound healing rate (WHR) in relation to diameter growth of alder trees, detailed statistical results shown in Table 6

Fig. 2 Non-linear regression of wound healing rate (WHR) in relation to DBH of alder trees, detailed statistical results shown in Table 6

Fig. 4 Non-linear regressions of annual diameter growth of wounded (WDG) and unwounded (UWDG) alder trees and their DBH, detailed statistical results shown in Table 6

4. Discussion

species (8.4%). This is probably in response to the position of alder trees in forest stands, which is closer to disturbed areas such as roads and skid trails (Tavankar et al. 2015a).

4.1. Logging wounds The results showed that 18.3% of residual alder trees sustained damage. This amount of damage was more than the amount of damage to all residual tree

78

About 86% of these wounds were situated on the bottom 1 m of the stem and about 73% were caused by Croat. j. for. eng. 38(2017)1


Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Fig. 5 Linear regression of diameter growth (DG) and ratio of wound size to basal area (RSA), detailed statistical results shown in Table 6 log winching in accordance with the results of other studies (Vasiliauskas 1993, Froese and Han 2006, Naghdi et al. 2008, Kosir 2008, Nikooy et al. 2010, Picchio et al. 2012, Marchi et al. 2014, Tavankar et al. 2013, Lotfalian et al. 2008, Majnounian et al. 2009, Jourgholami 2012, Tavankar et al. 2015a). In particular, the bottom 1 m of the stem represents the most valuable part of the tree and also the most vulnerable part to biological diseases; indeed, it was there that the highest percentage of open and decayed wounds occurred. Most damage occurred in DBH classes of 21–40 and 41–60 cm, 34.8% and 15.2%, respectively. As Table 6 Detailed results of regression analysis showed in Fig. 1–5 for relationship between: wound healing rate (WHR) and ratio of wound size to basal area (RSA), WHR and diameter at breast height (DBH), WHR and diameter growth (DG), wounded diameter growth (WDG) and DBH, unwounded diameter growth (UWDG) and DBH, diameter growth (DG) and RSA in alder trees Variables

N

r2

r2 adjusted

WHR–RSA

66

0.508

0.493

2.30 32.55

<0.001

WHR–DBH

66

0.337

0.316

2.66 16.03

<0.001

WHR–DG

66

0.607

0.594

2.05 48.59

<0.001

WDG–DBH

66

0.365

0.345

1.60 46.53

<0.001

UWDG–DBH

66

0.417

0.398

1.23 22.51

<0.001

DG–RSA

66

0.417

0.407

1.36 45.69

<0.001

Croat. j. for. eng. 38(2017)1

SE

F

p-value

F. Tavankar et al.

shown for beech trees in uneven aged forests (Tavankar et al. 2015a), wounded trees are mostly distributed in the two central DBH classes. The probability of damage depends on the felling intensity and trees density (Picchio et al. 2012), for the typical structure of the uneven aged forest (groups of coetaneous trees), and correct silvicultural treatment, similar to a thinning (more intensive in the central DBH classes). Pre-harvest planning and winching path marking before logging operations can reduce damage to the stand in these forests. Whereas residually damaged trees were highly concentrated near the skid trails as shown in other studies (Naghdi et al. 2009, Ezzati and Najafi 2010), skid trail planning before felling operations can substantially reduce the skidding damage (Naghdi et al. 2008, Majnounian et al. 2009). Nikooy et al. (2010) reported that skilled operation of a skidder can decrease the level of damage. Picchio et al. (2012) studied improved winching techniques designed to decrease stand damage in the forests of central Italy. They reported that the use of a snatch block decreased by one-quarter the frequency of wounded trees from 50% to 36%. Han and Kellogg (2000) suggested that artificial tree protection rigging such as rub pads should be used to prevent damage on stumps and stems. Therefore, training of forest workers and adequate technologies can be effective in reducing logging damage on residual stands. All workers should be made aware of the purpose of the selection cutting and both the minor and major injuries to the residual stands as well as excessive ground disturbance, which may result in significant volume losses (Davis and Nyland 1991) and the natural regeneration of the forest (Picchio et al. 2012). Limiting logging damage to residual trees must remain a major objective in selection managing of forests (Tavankar et al. 2015a).

4.2. Wound healing The mean wound healing rate (WHR) was 4.95 cm2/yr, ranging between 0 and 17.3 cm2/yr, and depended on wound characteristics (size, intensity, and location), tree age, and ratio of wound size to stem basal area (RSA). After 15 years from the occurrence of wounds, 51.1% were closed. In this case, unlike what was observed in similar conditions for beech trees (37.5% closed after 10 years from wound occurrence) (Tavankar et al. 2015b), alder seems to react in a better way in recovering from the wounds, especially wounds caused by skidding impact. The results showed that WHR decreased by increasing wound size and wound intensity, and increased by the increasing height of wound from ground level.

79


F. Tavankar et al.

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

The mean WHR of felling wounds (9.84 cm2/yr) was significantly higher than the WHR of skidding wounds (3.24 cm2/yr). This is due to the position and intensity of skidding wounds. For the extraction wounds, normally the first meter of the stem was damaged and the wounds were large and deep, extending below the bark down into the wood. These are ideal conditions for infection caused by diseases, in particular fungi. The highest coefficient of determination was found between diameter growth and WHR (Adj. r2 = 0.59).

4.3. Diameter growth Diameter growth (DG) of wounded alder trees (5.62 mm/yr) was 13.3% lower than that of unwounded alder trees (6.48 mm/yr). Tavankar et al. (2015b) studied the effects of logging wounds on diameter growth of beech trees (Fagus orientalis) in Caspian forests and reported a reduction of 8.1%. The reduction in diameter growth of wounded trees was closely related to the wound size, intensity and location, DBH and ratio of wound size to basal area. Wounds greater than 100 cm2 significantly reduced DG. Phloem and damaged wood wounds significantly decreased DG (10.1% and 11.7%, respectively). Skidding wounds significantly decreased DG (12.2%), because these wounds were intensive. The effect of wounds on the reduction of DG in the young trees was more intense than in the older trees, so the DG of wounded trees with DBH>58 cm was equal to the DG of unwounded trees.

5. Conclusion In the Caspian forests, the future of logging wounds and the effects of wounds on diameter growth of alder trees are unclear. This study shows how logging wounds could cause a sensitive reduction (13%) in diameter growth. Our results showed that the effect of bole wounds on the diameter growth depends on their severity, location, size, and tree age. Young alder trees were more sensitive to logging wounds. Residual stand damage is an unavoidable risk of selection cutting, but the level of damage should be minimized to assure the quality of the product in the future. The results of this study also indicate that tree skidding has a high potential for residual stand damage, and that intensive wounds occurred during skidding operations. Results show that the percentage of wounds by log skidding was higher than the percentage by tree felling (73% vs. 27%). The wound size by log extraction was smaller than by tree felling (57 vs. 201 cm2), but

80

the percentage of intensive wounds by log extraction was higher than by tree felling (73% vs. 22%). For these reasons it is important to minimize damage, both to the number of trees damaged and the extent of damage to the individual tree. Logging workers must be persuaded, through adequate training, that most damage to residual trees is unnecessary and avoidable.

6. References Behjou, F.K., 2014: Effects of wheeled cable skidding on residual trees in selective logging in Caspian forests. Smallscale Forestry 13(3): 367–376. doi: 10.1007/s11842-013-9259-x Bettinger, P., Kellogg, L., 1993: Residual stand damage from cut-to-length thinning of second growth timber in the Cascade Range of western Oregon. Forest Product Journal 43(1112): 59–64. Bragg, W., Ostrofsky, W., Hoffman, B., 1994: Residual tree damage estimates from partial cutting simulation. Forest Product Journal 44(7-8): 19–22. Camp, A., 2002: Damage to residual trees by four mechanized harvest systems operating in small diameter, mixed conifer forests and steep slopes in northeastern Washington: A case study. Western Journal of Applied Forestry 17(1): 14–22. Clark, D.A., Clark, D.B., 1992: Life history diversity of canopy and emergent trees in a neotropical rain forest. Ecolological Monographs 62(3): 315–344. Danilović, M., Kosovski, M., Gačić, D., Stojnić, D., Antonić. S., 2015: Damage to residual trees and regeneration during felling and timber extraction in mixed and pure beech stands. Šumarski list 139 (5–6): 253–262. Davis, C.J., Nyland, R.D., 1991. Designated skid trails: Results after 18 years. ASAE winter meeting, December 1991, Chicago Illinois, 17–20. Ezzati, S., Najafi, A., 2010: Long-term impact evaluation of ground-base skidding on residual damaged trees in the Hyrcanian forest, Iran. International Journal of Forestry Research 1(1): 1–8. doi:10.1155/2010/183735 Fjeld, D., Granhus, A., 1998: Injuries after selection harvesting in multi-storied spruce stands – the influence of operating systems and harvest intensity. Journal of Forest Engineering 9(2): 33–40. doi: 10.1080/08435243.1998.10702716 Froese, K., Han, H.S., 2006: Residual stand damage from cut-to length thinning of a mixed conifer stand in northern Idaho. Western Journal of Applied Forestry 21(3): 142–148. Gullison, R, Hardner, J., 1993: The effects of road design and harvest intensity on forest damage caused by selective logging: empirical results and a simulation model from the Bosque Chimanes, Bolivia. Forest Ecology and Management 59 (1–2): 1–14. doi:10.1016/0378-1127(93)90067-W Croat. j. for. eng. 38(2017)1


Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82) Han, H.S., Kellogg, L.D., 2000: Damage characteristics in young Douglas-fir stand from commercial thinning with four Timber Harvesting Systems. Western Journal of Applied Forestry 15(1): 27–33.

F. Tavankar et al.

Shafaroud forest of Guilan province. Iranian Journal of Environmental Sciences 60(3): 931–947. (In Persian)

Han, H.S., Kellogg, L.D., Filip, G.M., Brown, T.D., 2000: Scar closure and future timber value losses from thinning damage in western Oregon. Forest Product Journal 50(1): 36–42.

Naghdi, R., Rafatnia, N., Bagheri, I., Hemati, V., 2008: Evaluation of residual damage in felling gaps and extraction routes in single selection method (Siyahkal forest). Iranian Journal of Forest and Poplar Research 16(1): 87–98. (In Persian)

Jourgholami, M., 2012: Operational impacts to residual stands following ground-based skidding in Hyrcanian Forest, northern Iran. Journal of Forestry Research 23(2): 333– 337.

Nikooy, M., Rashidi, R., Kocheki, G., 2010: Residual trees injury assessment after selective cutting in broadleaf forest in Shafaroud. Caspian Journal of Environmental Sciences 8(2): 173–179.

Kiser, J., 2011: Histochemical and geometric alterations of sapwood in costal Douglas-fir following mechanical damage during commercial thinning. Silva Fennica 45(4): 729–741. doi:.org/10.14214/sf.447

Nikooy, M., Ershadifar, M., 2012: Effects of skid trail planning, landing construction and directional felling on normal selective logging in Caspian forest. Formec 8–12 Oct., Dubrovnik, Croatia, 9 p.

Košir, B., 2008: Damage to young forest due to harvesting in shelterwood systems. Croatian Journal of Forest Engineering 29(2): 141–153. Limbeck-Lilienau, B., 2003: Residual stand damage caused by mechanized harvesting systems. In: Proceedings of High Tech Forest Operations for Mountainous Terrain, Sclaegl, Austria, 2003, 1–11. Lo Monaco, A., Calienno, L., Pelosi, C., Balletti, F., Agresti, G., Picchio, R., 2015: Technical properties of beech wood from aged coppices in central Italy. iForest 8(1): 82–88. doi: 10.3832/ifor1136-007 Lotfalian, M., Parsakho, A., Majnounian, B., 2008: A method for economic evaluation of forest logging damages on regeneration and stand (Case study: Alandan and Waston Serries). Journal of Environmental Science and Technology 10(2): 51–62. (In Persian) Marchi, E., Picchio, R., Spinelli, R., Verani, S., Venanzi, R., Certini, G., 2014: Environmental impact assessment of different logging methods in pine forests thinning. Ecological Engineering 70: 429–436. doi: 10.1016/j.ecoleng.2014.06.019 Majnounian, B., Jourgholami, M., Zobeiri, M., Feghhi, J., 2009: Assessment of forest harvesting damage to residual stands and regenerations – A case study of Namkhaneh district in Kheyrud forest. Environmental Sciences 7(1): 33–44. (In Persian) Marvie-Mohadjer, M., 2006: Silviculture. Tehran, University of Tehran. (In Persian) Meadows, J.S., 1993: Logging damage to residual trees following partial cutting in a green ash-sugarberry stand in the Mississippi Delta. In: Proceedings of the 9th central hardwood forest conference, Indiana, US, March, 1993, 8–10. Naghdi, R., Rafatnia, N., Bagheri, I., Hemati, V., 2008. Evaluation of residual damage in felling gaps and extraction routes in single selection method (Siyahkal forest). Iranian Journal of Forest and Poplar Research 16(1): 87–98. (In Persian) Naghdi, R., Bagheri, I., Taheri, K., Akef, M., 2009: Residual stand damage during cut to length harvesting method in Croat. j. for. eng. 38(2017)1

Picchio, R., Neri, F., Maesano, M., Savelli, S., Sirna, A., Blasi, S., Baldini, S., Marchi, E., 2011: Growth effects of thinning damage in a Corsican pine (Pinus laricio Poiret) stand in central Italy. Forest Ecology and Management 262(2): 237–243. doi: 10.1016/j.foreco.2011.03.028 Picchio, R., Magagnotti, N., Sirna, A., Spinelli, R., 2012: Improved winching technique to reduce logging damage. Ecological Engineering 47: 83–86. doi: 10.1016/j.ecoleng.2012.06.037 Poorzadi, M., Bakhtiari, F., 2009: Spatial and temporal changes of Hyrcanian forest in Iran. iForest 2(5): 198–206. doi: 10.3832/ifor0515-002. Spinelli, R., Magagnotti, N., Nati, C., 2010: Benchmarking the impact of traditional small-scale logging systems used in Mediterranean forestry. Forest Ecology and Management 260(11): 1997–2001. doi:10.1016/j.foreco.2010.08.048 Tavankar, F., Majnounian, B., Bonyad, A., 2013: Felling and skidding damage to residual trees following selection cutting in Caspian forests of Iran. Journal of Forest Science 59(5): 196–203. Tavankar, F., Bonyad, A., 2014: Long-term Effects of Logging damages on quality of residual trees in the Asalem Nav forest. Journal of Environmental Studies 40(1): 39–50. (In Persian) Tavankar, F. Bonyad, A.E., Majnounian, B., 2015a: Affective factors on residual tree damage during selection cutting and cable-skidder logging in the Caspian forests, Northern Iran. Ecological Engineering 83(8): 505–512. doi.org/10.1016/j.ecoleng.2015.07.018 Tavankar, T., Bonyad, A.E., Marchi, E., Venanzi, R., Picchio, R., 2015b: Effect of logging wounds on diameter growth of beech (Fagus orientalis Lipsky) trees following selection cutting in Caspian forest of Iran. New Zealand Journal of Forestry Science 45(19): 1–7. doi: 10.1186/s40490-015-0052-9 Vasiliauskas, R., 1993: Wound decay of Norway spruce associated with logging injury and bark stripping. Proc of Lithuanian Forest Research Institute 33: 144–156.

81


F. Tavankar et al.

Effects of Logging Wounds on Caucasian Alder Trees (Alnus subcordata C.A. Mey.) ... (73–82)

Vasiliauskas, R., 2001: Damage to trees due to forestry operations and its pathological significance in temperate forest: a literature review. Forestry 74(4): 319–336. doi: 10.1093/forestry/74.4.319 Vasiliauskas, A., Stenlid, J., 2007: Discoloration following bark stripping wounds on Fraxinus excelsior. European Journal of Forest Pathology 28(6): 383–390. doi: 10.1111/j.1439-0329.1998. tb01192.x

Wallentin, C., 2007: Thinning of Norway spruce. PhD Thesis. Acta Universitatis Agriculture Sueciae, Sweden, 116 p. Whitman, A.A., Brokaw, V.L.N., Hagan, J.M., 1997: Forest damage caused by selection logging of mahogany (Swietenia machrophylla) in northern Belize. Forest Ecology and Management 92(1): 87–96. doi:10.1016/S0378-1127(96)03941-2

Authors’ address: Assist. prof. Farzam Tavankar, PhD.* e-mail: tavankar@aukh.ac.ir Islamic Azad University Department of Forestry Khalkhal IRAN Assist. prof. Mehrdad Nikooy, PhD. e-mail: nikooy@guilan.ac.ir Prof. Amireslam Bonyad, PhD. e-mail: bonyad@guilan.ac.ir University of Guilan Faculty of Natural Resources Somehsara IRAN

Received: February 26, 2016 Accepted: September 16, 2016

82

Prof. Rodolfo Picchio, PhD. e-mail: r.picchio@unitus.it Rachele Venanzi, PhD. e-mail: venanzi@unitus.it University of Tuscia Department of Agriculture and Forests Science Via S. Camillo de Lellis 01100 Viterbo ITALY * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning Anna Cudzik, Marek Brennensthul, Włodzimierz Białczyk, Jarosław Czarnecki Abstract This paper concerns the evaluation of logging systems in terms of the damage to the forest ecosystem. Damages to trees and soil during late thinning conducted in foothills areas in Poland using tree-length and cut-to-length logging systems were assessed. In both stands, the test plots were located within the primary and secondary skid trails. In the study, areas occupied by skid trails were determined as well as the depth of ruts. In order to determine changes in the soil properties at selected measurement points, a soil penetration resistance and a maximum shearing stress were measured. For each logging system, the share of trees damaged during harvesting operations and location of injuries were determined. The studies have shown that a 70% larger area was required to form technological trail with CTL than with TL. After CTL, skid trails were scarred by shallow ruts, and the share of ruts with the depth between 0.16 and 0.25 was three times smaller than after TL. The average increase in penetration resistance of soil in the ruts after TL was 324% and 302% and after CTL 308% and 220%, respectively, for primary and secondary skid trail, in comparison to the values obtained in measurement points located 5 m from the trails. In TL, comparable changes of soil properties were caused by skidder wheels and by hauled wood. The research has shown a greater share of damaged trees after TL. In both logging systems, the most damage was found within the root collar and lower parts of the bole. Keywords: logging system, soil disturbance, damage to trees

1. Introduction The main goal of forest management is to sustain continuous development of forest ecosystems that optimally fulfil their productive and non-productive functions (Gebauer et al. 2012, Jourgholami 2012). In order to achieve this goal, forest management should combine market demand, the economy of wood harvesting and requirements for forest environment protection (Zastocki 2003). In recent years, a continuous increase in the level of wood harvesting mechanization has been recorded. In many cases, manual felling and logging by horses and farm tractors have given way to mechanized harvesting, using specialized felling (harvester) and logging (skidder, forwarder) machines (Ampoorter et al. 2010). Widespread use of specialized forestry equipCroat. j. for. eng. 38(2017)1

ment contributes to increase in productivity and improves work safety, but in some cases could also be associated with the adverse effects on the forest environment (Magagnotti et al. 2012, Picchio et al. 2016). Logging is an example of the strongest human intervention into forest environment, causing many threats to its individual components. Damage to the forest ecosystem arises due to felling and skidding operations, regardless of the technical means used in this process. Many reports have indicated that most damage occurs during wood transportation from the stump area to the landing (Jamshidi et al. 2008, Cambi et al. 2015, Cambi et al. 2016). Disturbance of surface layers, changes in physical and chemical properties of soils and damage to residual trees are the main consequences of logging operations (Modry and Hubeny 2003, Ampoorter et al. 2007). Soil compaction is a seri-

83


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95)

ous disturbance due to physical properties caused by equipment used for skidding and forwarding. As a result of soil compaction, there is a decrease in porosity and increase in bulk density (Demir et al. 2007), penetration resistance and strength of the soil (McFero et al. 2006). Soil compaction causes a reduction of the number and activity of microorganisms (Tan et al. 2008) and edaphic fauna (Marchi et al. 2016, Venanzi et al. 2016) leading to disruption of the chemical processes in the soil (Arocena 2000). The common effects on the soil from ground based forest operations are increased compaction and removal of litter mass in skid trails (Tan et al. 2008). Damage to the soil contributes to the deterioration of conditions for tree growth and reduces the site productivity (Brais 2001, Gomez et al. 2002, Smith 2003). The problem of damage to remaining trees in the stand is the subject of many scientific studies. Damage to the residual stand in forest operations often occurs during timber extraction (Vasiliauskas 1993, Wronsky and Murphy 1994, Košir 2008, Picchio et al. 2012). Most trees damaged due to forest operations are situated close to the extraction trails (Froese and Han 2006, Youngblood 2000). Mechanical injury to residual standing trees are caused by machine traffic and log dragging (Klvač et al. 2010). Damage usually occurs right after treatment, but sometimes develops over time (Legere 2001, Ezzati and Najafi 2010). Damage to residual trees during selection cutting may decrease the quality of residual trees and increase stand mortality from insect and disease infestations (Han et al. 2000, Camp 2002). Range and size of damage to forest ecosystem during harvesting activities depends on a number of factors related to, among others, machine mass, type and size of its tires, used logging technology and number of machine passes or skidding cycles (Alakukku et al. 2003). Not without significance is also the age of the stand, where the logging operations are carried out, the amount of removal trees, season and weather conditions (Lageson 1997, Limbeck-Lilienau 2003). An important role is also played by the level of training and ecological awareness of employees (Bragg et al. 1994, Sirén 2001). A large number of factors affecting the level of damage to the forest ecosystem make the results presented in numerous scientific works not universal. It is primarily associated with natural and technological conditions of the logging process in different countries. There are few studies dealing with the comparison of damage to the forest ecosystem by different logging technologies during late thinning in a selection cutting system. The impacts on the environment, especially on soil and residual trees, is an important

84

aspect to be considered in planning and execution of forest operations (Picchio et al. 2016). The aim of the study was to assess the damage to forest soil and remaining trees after logging operations in late thinning performed using tree-length and cutto-length logging system. The specific goals were: To determine soil disturbance through appointment of the forest area occupied by skid trails and assess damage to the surface layers of the soil; Assess damage to trees remaining in stands, by determining the percentage of damaged trees and the location of injuries.

2. Material and methods The research was conducted in Poland (Lower Silesia) in Forest District Międzylesie, Forest sub-district Biała Woda in two selected forest stands No. 89a and 99b. Tree-length harvesting system (TL) was studied in forest stand No. 89a. TL included felling the tree Table 1 Description of study site Logging system Logging method

Tree-length (TL) Chain saw + Skidder

Felling category

Cut-to-length (CTL) Harvester + Forwarder

Late thinning

Location – stand number

89a

99b

Area, ha

9.34

6.09

Altitude, m

790–870

850–910

Slope, °

13–17

8–12

Ground cover

Herbaceous

Grass-green

Type of habitat

Fresh montane mixed broadleaved forest

Fresh montane mixed coniferous forest

Share, % – Species

80 – Spruce 20 – Beech

100 – Spruce

Stand age, year

90

85

Average diameter at breast height (DBH), m

0.38 – Spruce 0.28 – Beech

0.30

Average tree height, m

26 – Spruce 23 – Beech

23

Stocking of merchantable timber, m3·ha-1

476 – Spruce 76 – Beech

326

Total harvesting volume, 726 m3 (660 Spruce + 66 Beech)

449

Removal from 1 ha, m3

73

Time of harvesting

77 Spring

Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al.

Table 2 Specification of forest machines Harvester Gremo HPRV Head SP 551 LF II

Forwarder

Skidder

Gremo 950 R LKT 81 T

Length, mm

7490

7895

5700

Width, mm

2600

2600

2230

Weight, kg

13,970

11,185

7145

Ground clearance, mm

580

580

540

Height, mm

3445

3445

2780

Number of wheels

8

8

4

Tire dimensions

600/50×22.5

Outreach of boom, m

10

600/50×22.5 16.9-30 6.5

using chain saw and wood extraction by a cable skidder LKT 81T equipped with two-drum winch with pulling force of 80 kN. Cut-to-length harvesting system (CTL) was studied in forest stand No. 99b. The CTL system included a harvester and a forwarder with loading capacity of 10,000 kg. Detailed information on the location and characteristics of the test stands are presented in Table 1. Information concerning the vehicles (machines) used in the logging process are shown in Table 2. A survey of tree damage, rutting, soil penetration resistance and maximum shearing stress was conducted after wood harvesting operations. In both tested forest stands, the area occupied by skid trails was defined. A total length of skid trails was measured with a measuring tape. Rut depth measurements were performed on primary skid trails (more than 10 passes) and secondary skid trails (less than 5 passes). The depth of the ruts was measured at the designated lengths of 100 m, divided into 2 m sections. In each stand, 10 measuring sites were randomly selected (5 measuring sites on primary and 5 on secondary skid trail). Measurements were carried out in randomly selected rut. The depth of ruts was classified as follows: Þ <0.05 m (shallow) Þ 0.05–0.15 m (medium deep) Þ 0.16–0.25 m (deep) Þ >0.25 m (very deep). The degree of tree damage and changes in the surface layer of the soil were determined at designated test sites. Tested areas in both forest stands were representative network of circles with a radius of 12.6 m (area of ca 0.05 ha). In each stand, 10 research plots were selected (5 within the primary skid trail, 5 within the secondary skid trails). Scheme of the test site and Croat. j. for. eng. 38(2017)1

Fig. 1 The scheme of test sites – location measurement plots: L – left rut, B – between the ruts, R – right rut, 2 m from the rut, 5 m from the rut (control)

the location of the measurement plots are shown in Fig. 1. The evaluation of surface damage to the soil structure was carried out based on measurements of soil compaction (penetration resistance) and the maximum shearing stress. Compactness of soil in the layer of 0–0.2 m was measured using Penetrologger Eijkelkamp with a measuring range of 0–10 MPa and accuracy of 1 kPa. The instrument enabled the measurement and recording of data as a function of cone penetration depth. A cone with a base surface of 1 cm2 and an angle apex of 60° was used for tests. Penetrologger was equipped with a Theta probe type ML2x to measure soil moisture with the accuracy of 1% vol. Soil humidity was measured in several points of each tested stand in the surface layer (0–0.05 m). The maximum shearing stress was measured using a shear tester with a wings probe made by Geonor company. The measuring range of the instrument was 0–260 kPa with the measurement accuracy of 2 kPa. Measurements were made at three depths: 0.05; 0.10; 0.15 m. Measurements of penetration resistance and maximum shear-

85


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95)

ing stress at each plot were performed in five repetitions. The percentage of damaged trees during logging operations was determined based on the number of damaged trees relative to the total amount of remaining trees in the stand on each test site. Mechanical damage to residual trees was classified according to their location: Þ root collar and bole at a height <0.3 m from the ground Þ bole at a height of 0.3–1 m Þ bole at a height >1 m. The share of damage was specified for each class (different location) in relation to the total number of damaged trees. The statistical analysis of the obtained results was conducted in the Statistica 12.5 software. For the determination of factors impact, the multifactor analysis of variance (ANOVA) was used; the significance level a was equal to 0.05. Before carrying out the ANOVA tests, the terms of its applicability were verified (the normal distribution by Shapiro-Wilk test and the homogeneity of variance by Levene test). When the number of factor levels was higher than 2, the post-hoc tests (HSD Tukey) were conducted – these tests had to show the significant differences between each of the factor level.

3. Results The total length of skid trails in both tested forest stands was different and amounted to 1920 m in the stand 99b and 1500 in the stand 89a. The width of skid trails amounted 3.0 m in CTL and 3.5 m in TL. Fig. 2 presents areas occupied by logging trails in the analyzed forest stands, where different systems of wood logging were used. After TL, skid trails occupied 5.6% (0.53 ha) of the total area, trails were located at a distance of about 60 m from each other. Mechanized logging using harvester and forwarder requires a dense network of trails – after CTL, skid trails occupied 9.5% (0.58 ha) of the total stand area. The 22–32 m distance between the trails was primarily dictated by the location of trees to be felled and to a smaller extent by the length of the harvester boom. The soil humidity, measured during the tests, in both forest stands was comparable and amounted to 20–33% vol. Lower moisture values (20–26% vol.) was obtained in the undisturbed soil, higher soil moisture (24 –33% vol.) was observed within the skid trails and local terrain cavity.

86

Fig. 2 The area occupied by skid trails in relation to the total area of study forest stands The treatment effect on the share of the ruts of the total length of skid trails is shown in depth classes in Fig. 3. The data indicate that the extent of skid trails damage depends on the logging system and the number of machine passes. It was found that the greater diversity of ruts depth occurred in tree-length logging system (wood transported by skidder LKT 81T), than in cut-to-length logging system. In both logging systems, on the prevailing trail lengths shallow ruts occurred (<0.05 m); their share after TL was 49% and 64%, and after CTL 75% and 87%, respectively, on primary and secondary skid trails. In CTL, the share of ruts deeper than 0.05 m was significantly smaller compared to TL system. At the trails, where wood extraction with skidder was carried out, a share of ruts with depth of 0.16–0.25 m was three times greater than on the trails in stand 99b. After CTL, there were no very deep ruts (>0.25 m), while after TL, ruts in this depth class accounted for 4–6% of the skid trail length. The very deep ruts were located in the local cavity, where the ground was characterized by higher humidity. The differences in the range of skid trail damage were the result of differences in the size (width) of tires and number of wheels of forest machines. Tires with greater width had less tendency to penetrate deep into the soil. Greater number of wheels (8 wheels), in the case of forwarder, could result in a lower contact pressure. In addition, less damage to trails in CTL was probably the result of local trail surface protection by branches coming from harvesting operations. Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al.

Fig. 3 The percentages of different rut depth on skid trails in tested stands (arithmetic mean ÂąSD); a) Tree-length; b) Cut-to-length Fig. 4a, b and Fig. 5a, b show the course of changes in soil penetration resistance as a function of cone penetration depth in the analyzed harvest areas. Presented courses are related to measurements on research plots located at the primary and secondary skid trails. After TL (Fig. 4a, b), the maximum values of soil penetration resistance were exceeded by 5 MPa on the primary skid trail and by 4 MPa on the secondary skid trail. The average increase in soil penetration resis-

tance in the layer 0–0.2 m, both on the primary and secondary skid trail, was similar and amounted to 324% and 302% compared to the values obtained for undisturbed soil at a distance of 5 m from the trails. Soil penetration resistance values measured at points spaced about 2 to 5 m from the trails (both primary and secondary) do not differ significantly, which indicates that the range of changes in soil properties concerns only skid trails. The largest increase in soil pen-

Fig. 4 Relationship between penetration resistance and soil depth in stand 89a: a) primary skid trail, b) secondary skid trail Croat. j. for. eng. 38(2017)1

87


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95)

Fig. 5 Relationship between penetration resistance and soil depth in stand 99b: a) primary skid trail, b) secondary skid trail etration resistance on the primary skid trail has been demonstrated in 0–0.13 m depth, and on the secondary skid trail in the layer 0–0.10 m. It was observed that, after TL, the impact of the skidder wheels and drawn wood led to a comparable increase in penetration resistance. Courses of changes in the penetration resistance of the soil within the skid trails (primary and secondary) after CTL are shown in Fig. 5a, b. In case of cut-tolength logging system, the changes in soil properties were observed mainly in the ruts. The increase in penetration resistance of soil resulted from the weight of the machine and load of transported wood. The soil penetration resistance on the surface between the ruts was much lower than in the ruts, and in terms of value closer to the results obtained at a distance of 2 m from the skid trails. On the secondary skid trail, lower values were observed of soil penetration resistance than on the primary trail. In the analyzed depth range (0–0.2 m), the average increase in penetration resistance of soil in the ruts was 220% and 308%, respectively, for secondary and primary skid trail, in comparison to the values obtained in measurement points located 5 m from the trails. At a depth of 0.2 m in the ruts on the primary skid trail, the values of soil penetration resistance reached 4 MPa, while on the secondary skid trail these values do not exceed 3 MPa. After CTL, lateral range of changes of soil properties was negligible. It was indicated by the value of soil penetration resistance obtained at the measuring points outside the ruts.

88

Different waveform character presented in Fig. 4 and Fig. 5 indicate various susceptibility of the soil to compaction under the wheels of machines with different contact pressure. The shearing stress strength of the soil is the parameter used to determine conditions of transferring a driving force from vehicle wheels to the ground. The values of maximum shearing stress, measured after TL, are presented in Fig. 6a, b. It was observed that the shearing stress strength of the soil, both on the surface of skid trail and beyond it, increased with the depth. Similarly as in the case of penetration resistance, the range of influence of skidder wheels and transported wood was limited to the trails surface. A higher shearing stress was found on the primary skid trail (average differences between values obtained on the primary and secondary skid trail amounted to 16%). On the primary skid trail, the highest values of maximum shearing stress were obtained in the furrow formed by hauled wood (between ruts), at depth of 0.05 and 0.10 m. On the secondary trail, used with less intensity, the highest values of maximum shearing stress in the furrow caused by hauled wood were found only at a depth of 0.05 m. The greatest increase in values of maximum shearing stress, compared to the values obtained at the control point (5 m from the trail), was observed at a depth of 0.05 m (values even higher than 700%). At greater depths (0.10 and 0.15 m), the values of the maximum shearing stress were smaller than the values obtained on the undisturbed soil. The obtained results confirmed the fact that the Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al.

Fig. 6 The values of maximum shearing stress of soil measured in stand 89a: a) primary skid trail, b) secondary skid trail (arithmetic mean ± SD)

greatest changes in the soil structure occur in the surface layers at the contact point between the vehicle wheels and transported wood with the soil. This tendency was observed for both types of skid trail (primary and secondary). Fig. 7 presents the values of the maximum shearing stress measured within the trails after CTL. On the secondary skid trail, the values of the analyzed

parameter were significantly lower than on the primary trail. It was shown for both types of skid trail that the biggest changes of the soil structure occurred as a result of the impact of vehicles wheels (in the ruts), while on the surface between the ruts and 2 m from the skid trails the values of maximum shearing stress were significantly lower. This fact proves that the vertical impact of machinery on the soil is domi-

Fig. 7 The values of maximum shearing stress of soil measured in stand 99b; a) primary skid trail, b) secondary skid trail (arithmetic mean ± SD) Croat. j. for. eng. 38(2017)1

89


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95)

In the designated research areas, damages to trees remaining in the stand after the harvesting operations were assessed. These damages were primarily caused by wood extraction. In stand 89a, where treelength logging system was used, tree damages were caused by skidder wheels and hauled logs. In cut-tolength logging, the residual trees were damaged by wheels of machines (harvester/forwarder) and by hydraulic boom during loading of logs on forwarder trailer.

Fig. 8 Percentage share of damaged trees (arithmetic mean ± SD) nant, while the range of lateral changes is clearly smaller. The greater the number of passes, the higher the strength of the soil and the higher the values of maximum shearing stress of soil. Increase in the values of the maximum shearing stress was aligned to all measure depths; for the primary trail it was about 285% and for the secondary trail below 200% in relation to the undisturbed soil (5 m from the trails).

Fig. 9 Percentage share of tree damage classes (arithmetic mean ± SD)

90

Fig. 8 shows the percentage of damaged trees in the area of technological trail on both analyzed forest stands. It was observed that more standing trees were damaged when tree-length logging system was used. At the research plots located within primary skid trail in stand 89a, the damaged trees accounted for an average of 13.5%. The largest share of damaged trees was 16.1% (5 trees damaged out of 31 growing trees), while the smallest share was 10% (3 trees damaged out of 30 growing trees). Within the secondary skid trail, the share of damaged trees was significantly lower – an average of 6.8% (the lowest 3.33% and the highest 12.5%). In forest stand 99b, where cut-tolength logging system was used, the share of damaged trees within the primary skid trail was 6.3%, while within the secondary skid trail the share of damaged trees was 4%. Based on the methodology applied, damaged trees were classified according to their location. Assessment of logging technology was also conducted based on this classification. As the results obtained within the primary and secondary skid trail at both analyzed forest stands were comparable, the average values of the measurements obtained on research plots were presented. The share of individual damage class in relation to the total number of damaged trees is shown in Fig. 9. Both after TL and CTL, major damage to trees occurred on root collar and in the lowest parts of the bole, their share being 56% in treelength logging and 79% in cut-to-length logging. This type of tree damage was caused by the wheels of forest machinery. In case of tree-length logging system, damaging the bole at a height of 0.3–1.0 m accounted for 41% (they were caused by skidder and hauled wood). In the case of cut-to-length logging system, such injuries accounted for 16% and the damages were caused by the harvester/forwarder and grapple mounted on the hydraulic arm. Damage to the bole located at a height above 1.0 m was the result of tree felling and in tree-length system, it accounted for 3%, while in cut-to-length it accounted for 5%. The results of ANOVA tests are shown in Table 3. Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al.

Table 3 Results of statistical analysis; the significance level a=0.05; x – arithmetic mean; SD – standard deviation Analyzed parameter

Factor

<0.05 m

Share of ruts depth (in total length)

0.05–0.15 m Logging system 0.16–0.25 m >0.25 m Logging system

TL-location

Factor level TL

A

56.60

8.22

CTL

80.80B

6.91

TL

25.00A

4.69

CTL

B

14.30

4.71

TL

13.70A

2.75

CTL

4.50B

1.71

TL

5.00A

1.41

CTL

B

0.00

0.00

TL

2.38A

1.09

CTL

B

1.40

0.71

Left rut

3.27A

0.28

Between the ruts

A

3.23

0.31

Right rut

3.25A

0.27

B

0.05

5 m from trail

B

1.04

0.06

Left rut

2.19A

0.46

B

0.06

Right rut

A

2.20

0.43

2 m from trail

0.85B

0.08

5 m from trail

0.83B

0.05

2 m from trail

Penetration resistance

Between the ruts CTL-location

Logging system

TL

Shearing stress

Logging system

Rut collar, bole <0.3m Share of damage classes

Bole 0.3–1 m

Logging system

Bole >1 m

A

56.19

B

35.92

95.43

Left rut

131.51A

36.91

A

35.03

A

137.29

Right rut

134.58

35.67

2 m from trail

38.78B

16.70

5 m from trail

34.98B

14.44

A

35.86

Between the ruts

B

40.09

12.40

Right rut

90.99A

36.86

B

12.61

B

2 m from trail

Share of damaged trees

0.92

57.49

Left rut CTL-location

1.12

CTL Between the ruts

TL-location

±SD

x

86.84

34.96

5 m from trail

34.56

13.88

TL

10.20A

4.87

B

CTL

4.90

2.60

TL

56.00A

1.05

CTL

79.00B

1.33

TL

40.90A

0.99

B

CTL

15.80

1.03

TL

3.00A

0.47

CTL

5.00B

0.48

p-value <0.000001 <0.000001 <0.000001 <0.000001 <0.000001

<0.000001

<0.000001

<0.000001

<0.000001

<0.000001

0.001385 <0.000001 <0.000001 <0.000001

*the letters at arithmetic values denote separate homogeneous groups

Croat. j. for. eng. 38(2017)1

91


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95)

Based on statistical analysis of the first parameter, it can be stated that the logging system was a significant factor. For penetration resistance and shearing stress, besides the logging system, the location of measure was also a relevant factor. In this way, the range of soil structure changes, caused by logging, could be determined. It has been shown that all factors were significant for shearing stress values and for penetration resistance values. To determine the location as a factor, post-hoc tests had to be carried out. By Tukey’s HSD tests, conducted for TL and CTL, two different homogeneous groups have been identified. Further analysis showed that the logging system was statistically significant both for percentage of damaged trees and for shares of damages in all analyzed parts of the tree.

4. Discussion The area occupied by skid trails in tree-length logging system with wood extraction using skidder was 5.6% of the total stand area. A larger area under skid trails was occupied in stands where cut-to-length logging system was used. The obtained results were confirmed by other authors. Jones et al. (1996) showed that the trails for skidder accounted for 5.4% and 5.3% for planned and unplanned operations, respectively. Jackson et al. (2001), in a study conducted in tropical forests in Bolivia, showed that the area occupied by the primary and secondary logging trails accounted for 19% of the study area. In this case, the skid trails were established before harvesting operations and approximately 24% of the created skid trails were not used to transport logs. Picchio et al. (2012) showed that during thinning operations, using full-tree harvesting system with wood extraction by tractor with winch, skid trails occupied 2.7% of the study area. When using cut-tolength logging system in this study, the logging trail accounted for 9.5% of the stand area. Our results are different from those presented by others authors. According to Elliasson (2005) and Wågberg (2001), the area designated for harvester and forwarder passes accounted for 12.5% and 12.1%, respectively. The study of these authors was carried out at the final felling of trees, while our research was related to areas under trails after selective felling within late thinning. Therefore, the trail network did not have to be so dense, the trails did not have to be formed at distances equal to twice the length of the harvester crane as was the case in final cutting. Analyses of measured rut depths confirm that the rut depth increased with the increase of traffic intensity. Our results are in accordance with the findings of many researchers (McNabb et al. 2000, Eliasson and

92

Wasterlund 2007). The deepest ruts on the examined areas were formed in land cavities, where the soil had higher humidity. This is in accordance with the results obtained by Jourgholami and Majnounian (2011), who showed that the increase in soil humidity results in a significant increase in rut depth. The aim of the measurements of soil penetration resistance and the maximum shearing stress was to demonstrate changes in soil properties resulting from its compaction. In the present study, the values of analyzed parameters, obtained in ruts created by wheels of machines and by hauled wood, were significantly higher than the values obtained on the undisturbed soil. The maximum values of penetration resistance in the soil layer of 0–0.2 m were 4–5 MPa. Ampoorter et al. (2010) showed that the effect of the number of passes on the increase of soil compactness depends on soil humidity – the higher humidity, the greater the increase in compactness. In our research, we also observed a significant increase in soil penetration resistance and values of maximum shearing stress with the number of passes (significant differences between the values obtained on the primary and secondary skid trails with regard to the control plots). Taking into account differences in wood harvesting circumstances, it is hard to compare the results coming from different countries. Many studies have shown that trees damaged by forest operations may range between 4–21% of the total post-harvest stand (Vasiliauskas 2001, Picchio et al. 2011). The results presented in most publications referred to pine stands. Cervinkova (1980) and Dimitri (1983) showed that different tree species vary in resistance to damage, with spruce being more sensitive than pine. As shown in this study, the percentage of trees damaged by logging operations, with the use of harvester and forwarder, is comparable with the results presented by other authors. Based on the research conducted in Latvia, Epalts (1989) indicated 6.5% of damaged trees, Froding (1992) from Sweden – 4% in pine stand and 7.2% in spruce stand, Suwała (1999) – 4.5% in thinning of pine stand. Damage to trees after late thinning operations using tree-length logging system (felling by chain saw and extraction with cable skidder) in the foothill area in Poland (stand 89a) accounted for an average of 13.5% for the research plots located at primary skid trail and 6.8% on surfaces within the secondary skid trail. In studies of late thinning in pine forest, Suwała (1999) showed that after the operation damaged trees accounted for 12.8% of the remaining trees (the trees were felled by chain saw and extracted by cable skidder – distance between the trails was 60 m). In his studies conducted in the forests in Iran, Jourgholami (2012) Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al.

showed that wood transport using a skidder in the stands caused damage to 16.4% of growing trees. Generally, it can be stated that using of the short wood logging system (harvester and forwarder) during late thinning causes less damage to trees than using tree-length system. This is in accordance with the findings of Suwała et al. (2000). The results of researches conducted in lowlands of Poland have shown that the share of damaged pine trees in late thinning in the short-wood system is distinctly lower (4.1–5.8%) than in tree-length system (8.6–11.9%). Our research showed that the damage to residual growing trees is mainly located in the lower parts of the bole, at the height below 1 m. Bettinger and Kellogg (1993), Athanassiadis (1997), Naghdi et al. (2008), Jourgholami (2012) also found in their study that the damage to trees during harvesting operations is related to the lower parts of the bole. This study reported major damage in the root collar and the bole at a height of less than 0.3 m (59% – in tree-length logging system and 79% in cut-to-length system). Vasiliauskas (1993) presented similar results – he showed that in spruce stands harvested by partial and shelterwood cuttings, only 15% of all tree wounds were situated higher than 0.5 m, and over 60% of the trees were damaged at the root collar and at 0.3 m height from the ground.

5. Conclusions Cut-to-length logging system requires a denser trail network, the area under trail occupied 9.5% of the stand. In late thinning using tree-length logging system, the area under skid trails accounted for 5.6% of the total stand area. Much less damage to the trail was found in the short wood harvesting system, the ruts were shallower than after wood transportation by skidder; The range of changes of the analyzed soil properties, regardless of the used logging system, was mainly limited to the direct impact of the vehicle wheels and transported wood. On the trails in study stand 89a, greater increase in the maximum shearing stress values and soil penetration resistance was found after wood hauling with skidder than after wood logging using harvester and forwarder; Larger share of damaged trees was found in treelength logging. Higher share of damaged trees was observed at the primary skid trails. Mechanical injuries of remaining trees, for both logging systems, concerned the trees growing at skid trails. These wounds were mainly located on the root collars and lower parts of the boles. Croat. j. for. eng. 38(2017)1

6. References Alakukku, L., Weisskopf, P., Chamen, W.C.T., Tijink, F.G.J., Linden, J.P., Pires, S., Sommer, C., Spoor, G., 2003: Prevention strategies for field traffic-induced subsoil compaction: a review Part 1. Machine/soil interactions. Soil and Tillage Research 73(1–2): 145–160. Ampoorter, E., Goris, R., Cornelis, W.M., Verheyen, K., 2007: Impact of mechanized logging on compaction status of sandy forest soils. Forest Ecology and Management 241(1–3): 162–174. Ampoorter, E., Van Nevel, L., De Vos, B., Hermy, M., Verheyen, K., 2010: Assessing the effects of initial soil characteristics, machine mass and traffic intensity on forest soil compaction. Forest Ecology and Management 260(10): 1664–1676. Arocena, J.M., 2000: Cations in solution from forest soils subjected to forest floor removal and compaction treatments. Forest Ecology and Management 133(1–2): 71–80. Athanassiadis, D., 1997: Residual stand damage following cut-to-length harvesting operations with a farm tractor in two conifer stands. Silva Fennica 31(4): 461–467. Bettinger, P., Kellogg, L., 1993: Residual stand damage from cut-to-length thinning of second growth timber in the Cascade Range of western Oregon. Forest Products Journal 43(11–12): 59–64. Bragg, W., Ostrofsky, W., Hoffman, B., 1994: Residual tree damage estimates from partial cutting simulation. Forest Products Journal 44(7–8): 19–22. Brais, S., 2001: Persistence of soil compaction and effects on seedling growth in northwestern Quebec. Soil Science Society of America Journal 65(4): 1263–1271. Cambi, M., Certini, G., Fabiano, F., Foderi, C., Laschi, A., Picchio, R., 2016: Impact of wheeled and tracked tractors on soil physical properties in a mixed conifer stand. iForest – Biogeosciences and Forestry 9(1): 89–94. Cambi, M., Certini, G., Neri, F., Marchi, E., 2015: The impact of heavy traffic on forest soils: a review. Forest Ecology and Management 338: 124–138. Camp, A., 2002: Damage to residual trees by four mechanized harvest systems operating in small diameter, mixed conifer forests and steep slopes in northeastern Washington: a case study. Western Journal of Applied Forestry 17(1): 14–22. Cervinkova, H., 1980: Problems of wound decay in conifer stand of eastern Europe. In: Proceeding of the 5th International Conference on Problems of Root and Butt Rott in Conifers, Kassel, Germany, August 1978, 276−282 p. Demir, M., Makineci, E., Yilmaz, E., 2007: Investigation of timber harvesting impacts on herbaceous cover, forest floor and surface soil properties on skid road in an oak (Quercus petrea L.) stand. Building and Environment 42(3): 1194–1199. Dimitri, L., 1983: Wound decay following tree injury in forestry: establishment, significance and possibilities of its prevention. Forstwissenschaftliches Centralblatt 102: 68−72.

93


A. Cudzik et al. Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) Eliasson, L., 2005: Effects of forwarder tyre pressure on rut formation and soil compaction. Silva Fennica 39(4): 549–557. Eliasson, L., Wasterlund, W., 2007: Effects of slash reinforcement of strip roads on rutting and soil compaction on a moist fine-grained soil. Forest Ecology and Management 252(1–3): 118–123. Epalts, A., 1990: The impact of mechanized thinnings on the remaining stand. In: Proceedings of IUFRO Conference »Machine design and working methods in thinnings«. Hyytiälä, Finland, 17-22 September 1989, 11−21 p. Ezzati, S., Najafi, A., 2010: Long-term impact evaluation of ground-base skidding on residual damaged trees in the Hyrcanian forest. Iran International Journal of Forestry Research 1(1): 1–8. Fröding, A., 1992: Thinning damage to coniferous stand in Sweden (Bestándsskador vid gallring). Sveriges Lantbruksuniversitet. Garpenberg, 1−50 p. Froese, K., Han, H.S., 2006: Residual stand damage from cut-to length thinning of a mixed conifer stand in northern Idaho. Western Journal of Applied Forestry 21(3): 142–148. Gebauer, R., Neruda J., Ulrich R., Martinková M., 2012: Soil Compaction – Impact of Harvesters’ and Forwarders’ Passages on Plant Growth, Sustainable Forest Management – Current Research, Dr. Julio J. Diez (Ed.), ISBN: 978-953-510621-0, InTech, Available from:http://www.intechopen.com/ books/sustainable-forest-management-current-research/ impact-of-heavymachines-on-the-soil-root-system-andplant-growth, 179–196 p. Gomez, A., Powers, R.F., Singer, M.J., Horwath W.R., 2002: Soil compaction effects on growth of young ponderosa pine following litter removal in California’s Sierra Nevada. Soil Science Society of America Journal 66(4): 1334–1343. Han, H.S., Kellogg, L.D., Filip, G.M., Brown, T.D., 2000: Scar closure and future timber value losses from thinning damage in western Oregon. Forest Products Journal 50(1): 36–42. Jackson, S.M, Fredericksen, T.S., Malcolm, J.R., 2002: Area disturbed and residual stand damage following logging in a Bolivian tropical forest. Forest Ecology and Management 166(1–3): 271–283. Jamshidi, R., Jaeger, D., Raafatnia, N., Tabari, M., 2008: Influence of two ground-based skidding systems on soil compaction under different slope and gradient conditions. International Journal of Engineering Science 19(1): 9–16. Johns, J.S., Barreto, P., Uhl, C., 1996: Logging damage during planned and unplanned logging operations in the eastern Amazon. Forest Ecology and Management 89(1–3): 59–77.

Klvač, R., Vrána, P., Jiroušek, R., 2010: Possibilities of using the portable falling weight deflectometer to measure the bearing capacity and compaction of forest soils. Journal of Forest Science 56(3): 130–136. Košir, B., 2008: Damage to young forest due to harvesting in shelterwood systems. Croatian Journal of Forest Engineering. 29(2): 141–153. Lageson, H., 1997: Effects of thinning type on the harvester productivity and on the residual stand. Journal of Forest Engineering 8(2): 7–14. Legere, G., 2001: Reduction of stem damage by integrating skidding with declaiming. Forest Engineering Research Institute of Canada 2: 1–19. Limbeck-Lilenau, B., 2003: Residual stand damage caused by mechanized 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, Schlaegl – Austria. Magagnotti, N., Spinelli, R, Güldner, O., Erler, J., 2012: Site impact after motor-manual and mechanised thinning in Mediterranean pine plantations. Biosystems engineering 113(2): 140–147. Marchi, E., Picchio, R, Mederski, P.S., Vusić, D., Perugini, M., Venanzi, R., 2016: Impact of silvicultural treatment and forest operation on soil and regeneration in Mediterranean Turkey oak (Quercus cerris L.) coppice with standards. Ecological Engineering 95: 475–484. McFero Grace III, J., Skaggs, W.R., Cassel, D.K., 2006: Soil physical changes associated with forest harvesting operations on an organic soil. Soil Science Society of America Journal 70(2): 503–509. McNabb, D.H., Startsev, A.D., Nguyen, H., 2000: Soil wetness and traffic level effects on bulk density and air-filled porosity of compacted Boreal forest soils. Soil Science Society of America Journal 65(4): 1238–1247. Modrý, M., Hubený, D., 2003: Impact of skider and high-lead system logging on forest soils and advanced regeneration. Journal of forest science 49(6): 273–280. Naghdi, R., Rafatnia, N., Bagheri, I., Hemati, V., 2008: Evaluation of residual damage in felling gaps and extraction routes in single selection method (Siyahkal forest). Iranian Journal of Forest and Poplar Research 16(1): 87–98. Picchio, R., Neri, F., Maesano, M., Savelli, S., Sirna, A., Blasi, S., Baldini,S., Marchi, E., 2011: Growth effects of thinning damage in a Corsican pine (Pinus laricio Poiret) stand in central Italy. Forest Ecology and Management 262(2): 237–243.

Jourgholami, M., 2012: Environmental Impacts to Residual Stand Damage due to Logging Operations in Hyrcanian Forest. Notulae Scientia Biologicae 4(3): 65–69.

Picchio, R., Neri, F., Petrini E., Verani, S., Marchi, E., Certini G., 2012: Machinery-induced soil compaction in thinning two pine stands in central Italy. Forest Ecology and Management 285: 38–43.

Jourgholami, M., Majnounian, B., 2011: Effects of wheeled cable skidder on rut formation in skid trail – a case study in Hyrcanian forest. Journal of Forestry Research 22(3): 465−469.

Picchio, R., Spina, R., Calienno, L., Venanzi, R., Lo Monaco, A., 2016: Forest operations for implementing silvicultural treatments for multiple purposes. Italian Journal of Agronomy 11(s1): 156–161.

94

Croat. j. for. eng. 38(2017)1


Damage to Soil and Residual Trees Caused by Different Logging Systems Applied to Late Thinning (83–95) A. Cudzik et al. Sirén, M., 2001: Tree Damage in Single-Grip Harvester Thinning Operations. Journal of Forest Engineering 12(1): 29–38. Smith, C.W., 2003: Does soil compaction on harvesting extraction roads affect long-term productivity of Eucalyptus plantations in Zululand, South Africa? South African Forestry Journal 199(1): 41–54. Suwała, M., Jodłowski, K., Rzadkowski, S., 2000: Tree damage and soil disturbances at wood harvesting. In: Proceedings of the International Scientific Conference Forest and wood technology vs. environment. Brno, Czech Republic, 20−22 November 2000, 357−365 p. Suwała, M., 1999: Damages of trees and soil caused at timber harvesting in late thinnings of pine stands. Prace Instytutu Badawczego Leśnictwa 873, 1−86 p. Tan, X., Chang, S.X., Kabzems, R., 2008: Soil compaction and forest floor removal reduced microbial biomass and enzyme activities in a boreal aspen forest soil. Biology and Fertility of Soils 44(3): 471–479. Vasiliauskas, R.,1993: Wound decay of Norway spruce associated with logging injury and bark stripping. In: Proceeding of Lithuanian Forest Research Institute 33, 144–156 p.

Vasiliauskas, R., 2001: Damage to trees due to forestry operation and its pathological significance in temperate forests: a literature review. Forestry 74(4): 319−336. Venanzi, R., Picchio, R., Piovesan, G., 2016: Silvicultural and logging impact on soil characteristics in Chestnut (Castanea sativa Mill.) Mediterranean coppice. Ecological Engineering 92: 82–89. Wågberg, C., 2001: Miljöeffekter och omfattning av spårbildning vid slutavverkning (Environmental impact and extent of log hauling tracks at clear cutting). Department of Silviculture, Swedish University of Agricultural Sciences, Umeå, Student’s Reports 48, 28 p. Wronski E.B, Murphy G., 1994: Responses of forest crops to soil compaction. In Soil Compaction in Crop Production, B.D. Soane and C. van Ouwerkerk (Ed.), p. 317−342. Elsevier, Amsterdam. Zastocki, D., 2003: Wpływ stosowanych przez Zakłady Usług Leśnych środków zrywkowych na uszkodzenia gleby i pozostających drzew w sosnowych drzewostanach przedrębnych (The effect of means used in timber extraction by Forest Services Companies on soil disturbance and tree damage in premature pine stands). Sylwan 147(4): 52–58.

Authors’ address:

Received: July 28, 2016 Accepted: September 15, 2016 Croat. j. for. eng. 38(2017)1

Anna Cudzik, PhD. * e-mail: anna.cudzik@up.wroc.pl Marek Brennensthul, PhD. e-mail: marek.brennensthul@up.wroc.pl Prof. Włodzimierz Białczyk, PhD. e-mail: wlodzimierz.bialczyk@up.wroc.pl Jarosław Czarnecki, PhD. e-mail: jaroslaw.czarnecki@up.wroc.pl Wrocław University of Environmental and Life Sciences Faculty of Life Sciences and Technology Institute of Agricultural Engineering ul. J. Chełmońskiego 37a 51-630 Wrocław POLAND * Corresponding author

95



Original scientific paper

Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance in North Mountainous Forest of Iran Ahmad Solgi, Ramin Naghdi, Petros A. Tsioras, Ulrik Ilstedt, Ali Salehi, Mehrdad Nikooy Abstract Harvest traffic with heavy equipment causes damage to forest soils. Whereas increased soil damage has been reported with increasing harvest equipment traffic and on increasing slope gradients, it is unclear how much soil damage is caused by different directions of skidding. We examined the effects of traffic frequency, skid trail slope and skidding direction on the dry bulk density and total porosity of skidding trail soil in an Iranian temperate forest. The studied treatments included combinations of three different traffic frequencies (3, 7, and 12 passes of a rubber-tired skidder), three levels of slope (<10%, 10–20% and >20%) and two skidding directions (uphill and downhill). The impact on soil properties was greatest during the skidder initial passes. On steep slopes, only three skidder passes were required to cause substantial increases in soil bulk density relative to control plots, regardless of skidding direction. Independently of the traffic frequency and trail slope, uphill skidding caused substantially greater increases in dry bulk density and greater decreases in soil porosity than did downhill skidding. Total porosity was significantly lower on steep slopes than on gentle slopes regardless of traffic intensity and skidding direction. In general, fewer uphill skidder passes were required to achieve substantial soil disturbance than was the case for downhill skidding, possibly because skidders move more slowly when travelling upwards and uphill skidding places greater loads on the skidder rear axle. Keywords: bulk density, downhill skidding, soil compaction, total porosity, uphill skidding

1. Introduction Forest operations involving heavy machinery and animals may have negative effects on the remaining stand (Jourgholami and Majnounian 2013, Wang 1997). Some of the most common adverse impacts relate to the degradation of soil properties. Forest soils are susceptible to compaction because they are loose with a high organic matter content and are generally low in bulk density, high in porosity, and low in strength (Froehlich et al. 1985, Solgi et al. 2015a). Increased soil compaction and decreased soil porosity can reduce site productivity (McMahon 1995, Najafi and Solgi 2010). Consequently, soil compaction caused by logging activities can adversely affect the future regeneration and growth of trees (Sakai et al. 2008). Croat. j. for. eng. 38(2017)1

The degree of soil compaction is influenced by several factors including site and soil characteristics (Ampoorter et al. 2007), soil moisture (Greacen and Sands 1980, Jakobsen and Greacen 1985, Naghdi and Solgi 2014), harvesting system (Han et al. 2009, Jamshidi et al. 2008, Warkotsch et al. 1994) and type of equipment used (Greacen and Sands 1980), amount of slash (Gerasimov and Katarov 2010, McDonald and Seixas 1997), number of machine passes (Naghdi et al. 2007), properties of tires used on forest machines and harvesting vehicles (Sakai et al. 2008), and pressure applied to the soil (Eliasson 2005). It is, therefore, important for forest managers to understand how these factors interact, so that they can organize their operations in a way that will minimize soil disturbance (Demir et al. 2007, Sakai et al. 2008, Naghdi and Solgi 2014, Solgi and Najafi 2014).

97


A. Solgi et al. Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106)

Soil compaction is an unavoidable consequence of skidding operations, and can vary in intensity and distribution (Solgi et al. 2015b). Most of the physical soil disturbance caused during forest management activities is directly attributable to the movement of heavy equipment. Consequently, the impact of machine traffic on soil compaction has been studied in some detail. Traffic intensity plays an important role in soil compaction because deformations can increase with the number of passes, which may lead to excessive soil disturbance (Naghdi et al. 2007, Solgi et al. 2015a). In addition, slope steepness reportedly has a strong effect on soil disturbance during timber harvesting (Krag et al. 1986). During skidding on a terrain trail, a vehicle’s weight (and that of its load) will be distributed unevenly across its axles, with the rear axle typically being most heavily weighted. Najafi et al. (2009) reported that both the magnitude (extent) and depth of soil disturbance increased with the slope of the trail because of this effect. Finally, the direction of machine traffic (i.e. uphill vs. downhill) can modify the impact of vehicle traffic on the soil. Jourgholami et al. (2014a) concluded that the direction of forwarding had an effect on soil compaction. However, the effects of skidding direction on soil disturbance have received less attention than other aspects of machine traffic. A goal of forest managers is to minimize the negative impacts of vehicular passes during harvesting, especially because the effects of disturbance can persist for decades (Rab 2004). Estimation of the extent of disturbance from different direction of skidding is critical for selecting the best direction to minimize the negative impacts, especially in developing countries, where limited financial capacity prohibits the purchase of more modern and sophisticated equipment (Nikooy et al. 2013). The aim of this study is to fill this need by examining the impact of skidding direction (i.e. uphill vs. downhill) on soil physical properties on the skid trail in a mountainous beech stand of Iran. Our working hypothesis was that soil disturbance would increase with changing skidding direction from downhill to uphill skidding.

Table 1 Soil texture classes at a depth of 0–10 cm for skid trails. The range of particle size was <0.002, 0.002–0.05 and 0.05–2 mm for clay, silt, and sand, respectively Soil particle size distributions, g 100 g–1

98

Clay

Silt

Sand

28

39

33

Soil texture Clay loam

2. Materials and Methods 2.1. Site description The research was conducted during August 2014 in Shenrood forest, Guilan province, northern Iran between 36°13’N and 36°15’N and 53°10’E and 53°15’E (Fig. 1). The area is predominantly covered by oriental beech (Fagus orientalis Lipsky) stand. The canopy cover, mean diameter, mean height and stand density were 80%, 29.72 cm, 22.94 m and 170 trees ha-1, respectively. The soil class of our study area was classified as Cambisol according to the World Reference Base (WRB). The texture of the soil along the studied skid trail was classified as clay-loam on the basis of a particle size analysis using the Bouyoucos hydrometer method (Kalra and Maynard 1991). The description of the soil size distribution in skid trail is presented in Table 1. The average depth of soil to the bedrock was 70 cm. These areas are susceptible to erosion due to their steep mountainous conditions, heavy rainfall, and marl and limey sandstone sediments. Marls, due to their special constitution (35% lime and 65% clay), have low infiltration capacities and thus are susceptible to intense run-off and erosion. The study area has an elevation of approximately 800 m above sea level and a northerly aspect. The average annual rainfall recorded at the closest national weather station, located 20 km from the research area, was 860 mm. The maximum mean monthly rainfall of 120 mm usually occurs in October, while the minimum rainfall of 25 mm occurs in August. The mean annual temperature is 15 °C, with the lowest values occurring in February. At the time of skidding, the weather conditions were dry and warm, and the average gravimetric soil moisture content was 21%. The soil had not been driven on before the study began.

2.2. Forest operations and machine specifications At the study site, a combination of group selection and single tree selection silvicultural treatments were performed. In the Hyrcanian forests, harvesting and silvicultural operations are most common in autumn and winter, while log extraction is usually completed during the spring and summer period. Harvesting operations were carried out motor-manually, using chainsaw and axes (especially in thinning operations) for the tree-felling and processing. The logs were later transported from the forest stand to the roadside by means of a 4WD Timberjack 450C rubber-tired skidder which has an unloaded weight of 10.3 tons with 55% of the axle weight on the front axle and 45% on the rear axle. This model is the most common machine used for mechanized forest operations throughout northern Croat. j. for. eng. 38(2017)1


Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106) A. Solgi et al.

Fig. 1 Location of study area Table 2 Technical details of the Timber jack 450C rubber-tired cable skidder

Fig. 2 Rubber-tired skidder (Timberjack 450C) used in logging operations in a mountain forest of Iran Iran (Nikooy et al. 2013). The skidder was equipped with the engine model 6BTA5.9 (engine power of 132 kW) and was fitted with tires with the size of 24.5–32 inflated to 220 kPa (Fig. 2 and Table 2).

2.3. Experimental design and data collection The effects of the number of skidder passes, skid trail slope and direction of skidding on soil properties were determined along a skidding trail and in undisCroat. j. for. eng. 38(2017)1

Specifications

Timberjack 450C

Empty weight, kg

10,257

Number of wheels

4

Front tyres

24.5–32

Rear tyres

24.5–32

Average ground pressure, kPa

220

Engine power, kW

132

Manufacturing year

1998

Manufacturing location

Canada

turbed areas. A skid trail was selected with a longitudinal slope steepness ranging from 0 to 32% and no lateral slope. For the purposes of our study, the skidder, always loaded to its maximum load capacity, was used to extract 3 to 4 m long logs on this skid trail. With regard to the longitudinal profile and maximum gradient of the skid trail, three slope classes were considered (<10%, 10–20%, and >20%). The slope class <10% included trail sections that ranged from 3–7% in

99


A. Solgi et al. Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106)

gradient, gradient 10–20% contained a section within the range of 13–16%, whereas gradient class >20% contained sections within the range of 24–28%. Traffic frequencies of the loaded skidder were three passes for light traffic, seven passes for moderate traffic, and 12 passes for heavy traffic sections. Therefore, treatment plots included the combination of two levels of skidding direction, three levels of trail slope, and three levels of traffic frequency, thus forming 18 combinations of skidding direction, traffic frequency, and slope classes and each treatment was replicated twice (for a total of three tests), thus totaling 54 sample plots. Moreover, for control purposes, soil samples were taken from undisturbed areas that were not affected by skidding and at least 50–60 m (approximately 2 to 3 tree lengths) away from the skid trail to reduce side impacts. In total, 66 soil sampling plots were established: 12 control plots and 54 sample plots. Each plot measured 10 m long by 4 m wide, with a 5 m buffer zone between plots to avoid interactions (Fig. 3). In each plot, four sampling lines were drawn extending across the plot width, perpendicular to the skidder

direction of travel, with 2 m buffer zones between the lines in order to avoid interactions. Three samples of 0–10 cm soil were collected at different points along each such line, one from the skidder left-hand wheel track (LWT), one between the wheel tracks (BWT), and one from the right-hand wheel track (RWT). Soil samples weighing on average 310 g were collected with a soil hammer and rings (inside diameter 5 cm, length 10 cm). Samples were then placed in polyethylene bags, sealed, labeled, and transported to the laboratory, where they were promptly weighed to obtain wet mass. Soil samples were dried in an oven under 105°C until constant mass. Their moisture content was determined gravimetrically after drying (Kalra and Maynard 1991). Total porosity was calculated using Eq. (1):

AP = (1–Db / 2.5)

(1)

Where: AP total apparent porosity Db soil bulk density, and 2.65 (g cm–3) is particle density. Soil bulk density was calculated using Eq. (2):

Db = Wd / VC

(2)

Where: Wd weight of dry soil VC volume of soil cores (196.25 cm3).

2.4. Statistical Analysis One-way and three-way ANOVA was used to assess the significance of observed differences in average bulk density and total porosity under different traffic levels, trail slopes, and directions of skidding, and to assess the significance of interaction effects. Tukey’s HSD test was used to determine the significance of differences between average bulk densities and total porosities for different treatments (Zar 1999). All statistical calculations were performed using SPSS version 11.5.

3. Results Soil bulk density and total porosity were measured as 0.69 g.cm-3 and 73%, respectively, and soil texture was found to be Clay-Loam along the general harvesting area (Table 1).

3.1. Dry bulk density Fig. 3 Layout of an experimental plot

100

Following skidding, soil bulk density increased to 0.84–1.52 g cm-3 and was significantly affected by skidCroat. j. for. eng. 38(2017)1


Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106) A. Solgi et al.

Table 3 Effects of skidding direction, skid trail slope and traffic frequency in the north mountainous forest of Iran. Analysis of variance (P values) of the effects of number of passes (NP), slope gradient (SG) and skidding direction (SD) class on dry bulk density and total porosity Source of variable

p values Dry bulk density

Total porosity

NP

0.005

0.005

SG

0.005

0.005

SD

0.005

0.005

NP×SG

0.476

0.815

NP×SD

0.005

0.005

SG×SD

0.005

0.413

NP×SG×SD

0.517

0.005

*p values less than 0.05 are given in bold; variable interactions are denoted with »x«

downhill skidding were lower, ranging from 0.84 g cm–3 (3 passes and a slope of <10%) to 1.45 g cm–3 (12 passes and a slope of >20%) (Table 4). Regardless of skidding direction, dry bulk density increased consistently with increasing traffic intensity on all slope gradients (Table 4) and with increasing slope gradient at all traffic intensities (Fig. 4). Heavy traffic on slope gradients over 20% resulted in the highest dry bulk densities for both directions of skidding. Heavy traffic on slope gradients <10% resulted in dry bulk densities similar to those after moderate traffic on steeper slopes. Similarly, moderate traffic on slope gradients <10% resulted in bulk density values similar to those after light traffic on steeper slopes. Table 4 Effect of skidder passes on dry bulk density (g cm-3) Number of

direction

passes

(0–10)

(10–20)

(>20)

3

0.95c

1.03c

1.22c

7

1.17b

1.26b

1.38b

12

1.33a

1.39a

1.52a

3

0.84c

0.89c

1.07c

7

1.04b

1.11b

1.25b

12

1.25a

1.28a

1.45a

Uphill

ding direction, traffic intensity, slope gradient, the interaction of skidding direction × slope gradient, and skidding direction × traffic intensity (Table 3). The average soil bulk density in the undisturbed area was 0.69 g cm–3, while that on the skid trail ranged from 0.95 g cm–3 (for 3 passes and a slope of <10%) to 1.52 g cm–3 (12 passes and a slope of >20%) for uphill skidding. The values observed for sites subjected to

Slope, %

Skidding

Downhill

The majority of soil compaction along the skidding trail occurred during the skidder first few passes: the bulk density of the soil after the first three skidder passes was substantially greater than in control plots, regardless of slope steepness or skidding direction (Table 5). Table 5 Dry bulk density increase (%) per skidding direction and slope gradient class compared to the previous traffic intensity level Number of

direction

passes

(0–10)

(10–20)

(>20)

3

37.7

8.4

17.9

7

69.6

7.7

7.5

12

92.7

4.5

6.7

3

21.7

5.9

20.2

7

73.5

6.7

12.6

12

81.2

2.4

10.4

Uphill

Downhill

Fig. 4 Effect of skid trail slope on dry bulk density Croat. j. for. eng. 38(2017)1

Slope, %

Skidding

101


A. Solgi et al. Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106)

This indicates that the direction of skidding during field operations has important effects on compaction. However, the difference in bulk density between the two directions of skidding decreased as the number of passes increased (Fig. 6).

3.2. Total porosity Average total porosity was significant related to skidding direction, traffic intensity, slope gradient, the interaction of skidding direction × traffic intensity, and the interaction of all three factors (Table 3). The average total soil porosity was 73% on the undisturbed area but on the skid trail it ranged from 42% (after 12 passes on a slope of >20%) to 62% (after 3 passes on a slope of <10%) for uphill skidding. Downhill skidding produced slightly higher total porosity values than uphill skidding, ranging from 44% (after 12 passes on a slope of >20%) to 66% (after 3 passes on a slope of <10%) (Table 6). Fig. 5 Average bulk densities of skid trail sections with different skidding directions

Table 6 Effect of skidder pass numbers on total porosity (%) Number of

direction

passes

(0–10)

(10–20)

(>20)

3

62.35a

59.35a

53.94a

7

55.62ab

53.41ab

48.51ab

12

49.34b

48.63b

42.81b

3

66.54a

65.82a

58.54a

7

59.42ab

54.19ab

52.86ab

12

53.27b

53.04b

44.13b

Uphill

Downhill

Fig. 6 Differences in soil bulk density between trail sections with similar slopes used for uphill and downhill skidding at the same traffic frequency

For all numbers of skidder passes and trail slopes, the dry bulk density increase due to uphill skidding was considerably greater than that caused by downhill skidding (Fig. 5).

102

Slope, %

Skidding

Regardless of skidding direction, total porosity decreased consistently with increasing traffic intensity on all slope gradients and with increasing slope gradient at all traffic intensities (Table 6 and Fig. 7). Heavy traffic on slope gradients over 20% resulted in the lowest total porosity for both skidding directions. Heavy traffic on slope gradients <20% resulted in total porosity values similar to those after moderate traffic on steeper slopes. Similarly, moderate traffic on slope gradients <20% resulted in total porosity values similar to those after light traffic on steeper slopes. Regardless of skidding direction, the initial passes decreased total porosity the most; subsequent passes resulted in relative decreases that were more modest, particularly on steeper slopes (Fig. 6). In addition, the total porosity of trail sections subjected to uphill skidding was significantly lower than Croat. j. for. eng. 38(2017)1


Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106) A. Solgi et al.

Fig. 7 Effect of slope on average total porosity

Fig. 8 Average total porosities of skid trails with different skidding directions that of otherwise equivalent sections subjected to downhill skidding (Fig. 8).

4. Discussion Compaction of the soil increases bulk density, reduces soil porosity, decreases infiltration rates, and Croat. j. for. eng. 38(2017)1

lowers soil permeability (Froehlich et al. 1981). It has been well established that these changes in soil physical properties increase surface runoff and erosion, and create less-favorable soil environments for plant growth (Greacen and Sands 1980, Solgi et al. 2014). Careful planning of landing locations and skid trail systems before logging commences is intended to reduce damage to the soil and the residual stand. The dry bulk density of trail sections used for uphill skidding was substantially higher than that for sections used for downhill skidding. The greater compaction caused by uphill skidding can be explained by the greater load on the skidder rear axle (Najafi et al. 2009, Jamshidi et al. 2008, Solgi et al. 2015b). In sections subjected to uphill skidding, steeper slopes were associated with higher bulk densities and lower total porosities, possibly because of the skidder lower speed when climbing steeper slopes. Jourgholami et al. (2014b) reported that differences in levels of compaction between uphill and downhill forwarding may be due to an uneven distribution of the load between the skidder axles or increased wheel slippage and vibration during upslope forwarding relative to downslope forwarding. Soil compaction was also affected by trail slope, with dry bulk density increasing faster at higher slope levels. This is consistent with previous findings (Jamshidi et al. 2008, Solgi et al. 2013) and can be attributed to the difficulty of operating a skidder on steep terrain, which causes the machine to slip continuously and spend relatively long periods of time in each trail section, puddling and dragging the soil (Gayoso and Iroume 1991). When a skidder moves slowly on a steep slope, the top soil is more extensively vibrated and, therefore, more heavily disturbed than would be the case in a flatter area (Naghdi and Solgi 2014). For downhill skidding, the dry bulk density increased with slope steepness independently of the traffic frequency. This may be because machines are driven more slowly on steeper trails for reasons of safety and to minimize the risk of accidents. As shown in Table 5, the bulk density of the skidding trail soil is substantially greater than that in the control plots, even when only three skidder passes have been made. This is broadly consistent with the results of Ampoorter et al. (2007), who reported that more than half the total impact of skidding occurs after only three passes have been made and that further increases in the number of passes have much more modest effects on the bulk density (Ampoorter et al. 2007, Jamshidi et al. 2008, Solgi and Najafi 2014). According to Jamshidi et al. (2008), this may be because the compaction of the trail soil during the first skidder

103


A. Solgi et al. Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106)

passes increases its shear strength, raising its loadbearing capacity. However, the threshold value above which further passes cause only minimal increases in bulk density is probably quite variable, because it will depend on multiple factors including the harvesting technique used, the system chosen, and the equipment used (e.g. animal extraction vs. mechanized extraction), among other things. The bulk density threshold above which soil is too dense for plant roots to penetrate is considered to be between 1.40 and 1.55 g cm-3 for soils with light or medium textures (Kozlowski 1999). Our results indicate that the bulk density approached this critical level after 7 skidder passes in the case of uphill skidding and 12 passes for downhill skidding (Table 4). Naghdi and Solgi (2014) reached similar conclusions after examining various combinations of slope gradient, number of passes, and soil moisture levels. The total porosity of the skid trail was considerably lower than that of the undisturbed control areas. For both skidding directions, total porosity decreased in conjunction with increases in skidder traffic frequency and skid trail slope. This study focused on the topmost 10 cm of soil, which is more sensitive to disturbance by machine traffic than deeper soil layers (Carter et al. 2007). Increasing the number of skidder passes from three to seven had important effects on the total porosity. However, further increases from 7 to 12 passes had negligible effects.

5. Conclusions This work reports the magnitude of changes in physical soil properties associated with different levels of skidder traffic, skidding directions, and slope gradients. Soil compaction during skidding operations increased the bulk density of the skid trail soil and reduced its total porosity. The initial three machine passes exerted the highest effect on soil properties, regardless of the skidding direction or the slope gradient. Increasing traffic frequency has been found to increase the magnitude of soil disturbance, but at a reducing rate. This suggests that unnecessary machine traffic should be kept to a minimum necessary level. On the contrary, the relationship between soil compaction and skidding direction implies that particular attention should be paid to uphill skidding, which causes greater levels of soil compaction than downhill skidding. The slope of the skid trail is another important factor; the results presented herein suggest that slopes of more than 20% should be avoided on skid trails where possible.

104

The results presented herein demonstrate the importance of the examined soil compaction factors and suggest the need to develop practices that will minimize soil compaction. The adverse effects of both skidding direction and slope gradient could potentially be addressed to at least some extent by training machine operators to improve and optimize their driving style. Such training could help minimizing soil compaction, especially in areas where steep slopes are prevalent. Forest managers can also contribute to forest soil protection by means of redesigning the forest road network and careful planning of forest operations. The combination of workforce training and targeted managerial initiatives are expected not only to reduce the impacts of ground skidding but also bring other benefits in the form of increased operational efficiency, reduced operational costs and higher safety during forest operations.

6. References Ampoorter, E., Goris, R., Cornelis, W.M., Verheyen, K., 2007: Impact of mechanized logging on compaction status of sandy forest soils. Forest Ecology and Management 241(1): 162–174. Demir, M., Makineci, E., Yilmaz, E., 2007: Investigation of timber harvesting impacts on herbaceous cover, forest floor and surface soil properties on skid road in an oak (Quercus petrea L.) stand. Building and Environment 42(3): 1194–1199. Eliasson, L., 2005: Effects of forwarder tyre pressure on rut formation and soil compaction. Silva Fennica 39(4): 549–557. Froehlich, H., Aulerich, D., Curtis, R., 1981: Designing skid trail system to reduce soil impacts from tractive logging machines. For Res Lab, Oregon State University, Res Pap 44, 15 p. Froehlich, H.A., Miles, D.W.R., Robbins, R.W., 1985: Soil bulk density recovery on compacted skid trails in central Idaho. Soil Science Society of American Journal 49(4): 1015–1017. Gayoso, J., Iroume, A., 1991: Compaction and soil disturbances from logging in southern Chile. Annals of Forest Science 48(1): 63–71. 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. Greacen, E.L, Sands, R., 1980: Compaction of forest soils: a review. Australian Journal of Soil Research 18(2): 163–189. Han, S.K., Han, H.S., Page-Dumroese, D.S., Johnson, L.R., 2009: Soil compaction associated with cut-to-length and whole-tree harvesting of a coniferous forest. Canadian Journal of Forest Research 39(5): 976–989. Jakobsen, B.F., Greacen, E.L., 1985: Compaction of sandy forest soils by forwarder operations. Soil and Tillage Research 5(1): 55–70. Croat. j. for. eng. 38(2017)1


Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106) A. Solgi et al. Jamshidi, R., Jaeger, D., Raafatnia, N., Tabari, M., 2008: Influence of two Ground-Based skidding systems on soil compaction under different slope and gradient conditions. International Journal of Forest Engineering 19(1): 9–16.

Najafi, A., Solgi, A., 2010: Assessing site disturbance using two ground survey methods in mountain forest. Croatian Journal of Forest Engineering 31(1): 47–55.

Jourgholami, M., Majnounian, B., 2013: Traditional mule logging method in Hyrcanian Forest: a study of the impact on forest stand and soil. Journal of Forest Research 24(4): 755–758.

Nikooy, M., Esmailnezhad, A., Naghdi, R., 2013: Productivity and cost analysis of skidding with Tiimberjack 450C in forest plantations in Shafaroud watershed, Iran. Journal of Forest Science 59(7): 261–266.

Jourgholami, M., Majnounian, B., Abari, M.E., 2014a: Effects of tree-length timber skidding on soil compaction in the skid trail in Hyrcanian forests. Forest Systems 23(2): 288–293.

Pagliai, M., Febo, P., La Marca, M., Lucamonte, G., 1992: Compaction effects caused from different types of tyres on macroporosity and soil structure. Riv Ingegner Agr 3: 168–176.

Jourgholami, M., Soltanpour, S., Abari, M.E., Zenner, E.K., 2014b: Influence of slope on physical soil disturbance due to farm tractor forwarding in a Hyrcanian forest of northern Iran. iForest 7(5): 342–348.

Rab, M.A., 2004: Recovery of soil physical properties from compaction and soil profile disturbance caused by logging of native forest in Victorian central highlands, Australia. Forest Ecology and Management 191(1–3): 329–340.

Kalra, Y.P., Maynard, D.G., 1991: Methods and manual for forest soil and plant analysis. Forestry Canada, Re NORX-319. Northern Forestry Center.

Sakai, H., Nordfjell, T., Suadicani, K., Talbot, B., Bollehuus, E., 2008: Soil compaction on forest soils from different kinds of tires and tracks and possibility of accurate estimate. Croatian Journal of Forest Engineering 29(1): 15–27.

Kozlowski, T.T., 1999: Soil compaction and growth of woody plants. Scandinavian Journal of Forest Research 14(6): 596– 619. Krag, R., Higginbotham, K., Rothwell, R., 1986: Logging and soil disturbance in southeast British Columbia. Canadian Journal of Forest Research 16(6): 1345–1354. Marsili, A., Servadio, P., Pagliai, M., Vignozzi, N., 1998: Changes of some physical properties of a claysoil following passage of rubber- and metal tracked tractors. Soil and Tillage Research 49(3): 185–199. McDonald, T.P., Seixas, F., 1997: Effects of slash on forwarder soil compaction. Journal of Forest Engineering 8(2): 15–26. McMahon, S., 1995: Accuracy of two ground survey methods for assessing site disturbance. Forest Engineering. Logging Industry Research Organization Rotirua, New Zealand. International Journal of Forest Engineering 6(2): 27–33. Naghdi, R., Bagheri, I., Akef, M., Mahdavi, A., 2007: Soil compaction caused by 450C Timber Jack wheeled skidder (Shefarood forest, northern Iran). Journal of Forest Science 53(7): 314–319. Naghdi, R., Solgi, A., 2014: Effects of skidder passes and slope on soil disturbance in two soil water contents. Croatian Journal of Forest Engineering 35(1): 73–80. Najafi, A., Solgi, A., Sadeghi, S.H., 2009: Soil disturbance following four wheel rubber skidder logging on the steep trail in the north mountainous forest of Iran. Soil and Tillage Research 103(1): 165–169.

Croat. j. for. eng. 38(2017)1

Solgi, A., Najafi, A., Sam Daliri, H., 2013: Assessment of Crawler Tractor Effects on Soil Surface Properties. Caspian Journal of Environmental Science 11(2): 185–193. Solgi, A., Najafi, A., 2014: The impacts of ground-based logging equipment on forest soil. Journal of Forest Science 60(1): 28–34. Solgi, A., Najafi, A., Sadeghi, S.H., 2014: Effects of traffic frequency and skid trail slope on surface runoff and sediment yield. International Journal of Forest Engineering 25(2): 171– 178. Solgi, A., Naghdi, R., Nikooy, M., 2015a: Effects of skidder on soil compaction, forest floor removal and rut formation. Madera y Bosques 21(2): 145–153. Solgi, A., Naghdi, R., Tsioras, P.A., Nikooy, M., 2015b: Soil compaction and porosity changes caused during the operation of Timberjack 450C skidder in northern Iran. Croatian Journal of Forest Engineering 36(2): 77–85. Wang, L., 1997: Assessment of animal skidding and ground machine skidding under mountain conditions. International Journal of Forest Engineering 8(2): 57–64. Warkotsch, P.W., VanHuyssteen, L., Olsen, G.J., 1994: Identification and quantification of soil compaction due to various harvesting methods – A case study. Southern African Forest Journal 170(1): 7–15. Zar, J.H., 1999: Biostatistical analysis, 4th edition, Prentice Hall, Upper Saddle River, NJ, USA. 662 p. plus appendices.

105


A. Solgi et al. Combined Effects of Skidding Direction, Skid Trail Slope and Traffic Frequency on Soil Disturbance ... (97–106)

Authors’ address: Ahmad Solgi, PhD. student e-mail: solgi_ahmad231@yahoo.com Assoc. prof. Ramin Naghdi, PhD. * e-mail: rnaghdi@guilan.ac.ir Assoc. prof. Ali Salehi, PhD. e-mail: asalehi@guilan.ac.ir Assist. prof. Mehrdad Nikooy, PhD. e-mail: Nikooy@guilan.ac.ir University of Guilan Faculty of Natural Resources Department of Forestry PO.Box 1144 Sowmeh Sara, Guilan IRAN Petros A. Tsioras, PhD. e-mail: ptsioras@for.auth.gr Aristotle University of Thessaloniki Department of Harvesting and Technology of Forest Products Lab of Forest Utilization POB 227 GREECE

Received: November 18, 2015 Accepted: July 27, 2016

106

Assoc. prof. Ulrik Ilstedt, PhD. e-mail: ulrik.ilstedt@sek.slu.se Department of Forest Ecology and Management The Swedish University of Agricultural Sciences (SLU) SE-901 83 Umeå SWEDEN * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety Abolfazl Jaafari, Javad Rezaeian, Masoud Shafipour Omrani Abstract This study produces a slope failure susceptibility map for evaluation of the Caspian Forest for its capacity to support road construction and timber harvesting. Fifteen data layers were used as slope failure conditioning factors, and an inventory map of recent failures was used to detect the most susceptible areas. Five different datasets of conditioning factors were constructed to compare the efficiency of one over the other in susceptibility assessment. Slope failure susceptibility maps were produced using an adaptive neuro-fuzzy interface system (ANFIS) and geographical information system (GIS). The accuracy of the maps was then evaluated by the area under curve (AUC). The validation results suggest that the ANFIS model with input conditioning factors of slope degree, slope aspect, altitude, and lithology performed the best (AUC=83.74%) among the various ANFIS models explored here. The five ANFIS models have performed reasonably well, and the maps allow development of prudent hazard mitigation plans for the safety in forestry operations. Keywords: ANFIS, GIS, Caspian Forest, landslide susceptibility, forest road, timber harvesting

1. Introduction Landslides are one of the most devastating natural hazards in mountainous terrains. Although the action of gravity is the primary driving force (Gorsevski et al. 2006), landslides are also aggravated by human activities such as mining, agriculture, and forestry operations. With respect to forestry operations (timber harvesting and road construction activities), landslide often increases with long-term consequences and has been reported worldwide (e.g., Sessions et al. 1987, Duncan et al. 1987, Larsen and Parks 1997, Allison et al. 2004). When damaging landslides occur on forestlands, it is not unusual to hear appeals for a broad ban on forestry operations. However, such a ban would be very costly to many forest landowners and it would impact their contributions to state and local economies. Therefore, apart from regular hazard reduction plans, landslide susceptibility (LS) assessments should also be developed and implemented for the safety in forestry operations. Landslide hazard reduction plans, which are generated as the site is handed over to a contractor, are important tools to ensure everybody understands how to deal with different levels of LS across the working site. Croat. j. for. eng. 38(2017)1

In the Caspian Forest in northern Iran, landslides and slope failures are a common problem because naturally formed slopes are disturbed by forestry operations. History has shown that roads with improper terrain stability assessment in this area can cause significant slope failures. This trend is expected to continue and may increase in the future; some estimates suggest that significant portions of the Caspian Forest are prone to mass wasting, and forestry operations in this forest can accelerate landslide rates and magnitudes (Jaafari et al. 2014). Therefore, understanding of LS is needed to evaluate forestry strategies including alternate choices of road location, choice of road standards, choice of transport mode, and understanding whether timber harvesting on and around steep slopes is reasonable. The effectiveness of slope stability studies is apparent from the high prediction results of LS assessment reports from models such as logistic regression (e.g., Pourghasemi et al. 2013), knowledge-based analytical hierarchy process (e.g., Pourghasemi et al. 2012, Pourghasemi et al. 2013), fuzzy logic (e.g., Pourghasemi et al. 2012, Akgun et al. 2012), artificial neural networks (ANNs) (e.g., Conforti et al. 2014), support vector machine (e.g., Pradhan 2013) and adaptive neuro-

107


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

fuzzy interface system (ANFIS) (e.g., Vahidnia et al. 2010, Sezer et al. 2011, Bui et al. 2012, Pradhan 2013). In the case of ANFIS, developed by Jang (1993), only minor applications of landslide-related studies have been reported (Bui et al. 2012). ANFIS is a multilayer feed-forward network, in which each node performs a particular function on incoming signals and has a set of parameters pertaining to this node (Jang 1993). ANFIS combines fuzzy logic and ANNs by using the mathematical properties of ANNs in tuning a rulebased fuzzy inference system that approximates how the human brain processes information (Akib et al. 2014). The ANFIS model is implemented as a first order Takagi and Sugeno’s type fuzzy inference system (Takagi and Sugeno 1983) that consists of 2 fuzzy ifthen rules: Rule 1: If x is A1 and y is B1 then f1 = p1x + q1y + r1

(1)

Rule 2: If x is A2 and y is B2 then f2 = p2x + q2y + r2

(2)

Where: x, y are inputs A, B corresponding term set f output p, q, r constant The main objective of an ANFIS model is to determine the optimum values of the equivalent fuzzy inference system parameters by applying a learning algorithm using input–output datasets. The parameter optimization is done in such a way that during the training session, the error between the target and the actual output is minimized. Further information on ANFIS can be found in Jang (1993). LS assessment involves handling, processing and interpreting a large amount of territorial data. Geographical Information Systems (GIS) are very useful in susceptibility assessment (Ayalew et al. 2005), because they allow frequent updating of the database related to spatial distribution of landslide events and their predisposing factors, as well as the susceptibility assessment procedures (Conforti et al. 2014). In recent years, the use of GIS-based approaches to study landslides has been frequently reported. These include GIS-based frequency ratio, index of entropy, and weights of evidence models (Jaafari et al. 2015a, Jaafari et al. 2014), and GIS-based multicriteria decision analysis (Feizizadeh and Blaschke 2013). Bui et al. (2012) used a GIS-based ANFIS model for LS mapping in Vietnam. Their results showed that ANFIS is a robust method for landslide modeling. Pradhan (2013) compared the ability of the decision tree, support vector machine and ANFIS models to do LS mapping within a GIS environment. The results showed that all the

108

models faired reasonably well, however, the success rate showed that ANFIS had better prediction capability. This paper addresses the slope failure (landslide) susceptibility assessment in the Caspian Forest using an ANFIS suitable to GIS-based analysis. The study tackles the main causal factors and delimits the most susceptible zones for slope failure as a useful tool for the engineers involved in road construction and timber harvesting. The susceptibility maps are also compared with the known landslide locations according to the area under the curve (AUC) of receiver operator characteristic (ROC) curve to test the reliability and accuracy of the modeling approach. The susceptibility assessment presented here enables forest practitioners to avoid areas where forestry operations could cause slope failure, help identify where monitoring programs are necessary, and adopt appropriate policies to guide more efficient forestry operations.

2. Materials and methods 2.1 Study area The study area is situated in Mazandaran Province, which shares a border with Golestan and Guilan Provinces in the north of Iran. The study area has an approximate area of 52 km2 and is located between 36º29’10˝ N and 36º32´50˝ N latitude and 51º40´60˝ E and 51º48´20˝ E longitude (Fig. 1). The area is a part of the Educational and Experimental Forest of Tarbiat Modares University in the Caspian Forest with slope variations between flat and >80°, and altitudes between 160 and 2190 m. Slope shapes vary but frequently represent convex elements. They mainly feature concave valleys. In this area, the stream network flows from the north-east to the west with a dendritic pattern. Due to proximity of the Caspian Sea, the study area enjoys a humid and mild climate with average annual precipitation between 414 to 879 mm. The average summer and winter temperature was 22.5 and 10ºC, respectively (Jaafari et al. 2015b). The vegetation cover is quite continuous and is formed by deciduous trees. According to the geologic map of the area, prepared by Geological Survey of Iran (GSI), the major portion of the study area is underlain by dolomitic limestone. The Alborz fault is the most important fault in the area and is a reverse fault that follows the westeast orientation and dip toward the south. This fault is active, and most earthquakes and landslides in Mazandaran Province are the result of displacements and activity of this fault (Darvishzadeh 2004). Croat. j. for. eng. 38(2017)1


Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

A. Jaafari et al.

Fig. 1 Location of study area with landslide inventory map

2.2 Spatial database 2.2.1 Landslide inventory map The landslide inventory map of the study area was compiled by inheriting the landslide locations from interpretation of aerial photographs and field-based inspections. Aerial photographs show that historical landslides could be mapped via breaks in the forest canopy, denuded vegetation on the slope, bare soil, and other typical geomorphic characteristics (Pradhan 2013, Jaafari et al. 2014). Given the abundant over and understory vegetation in the study area, multiple field surveys and observations were conducted to produce a more detailed and reliable landslide inventory map (Jaafari et al. 2014). Shallow landslides were dominant, but large deep-seated landslides were also observed. In recent years, 103 landslides were detected and mapped within 52 km2 to assemble a database to evaluate the spatial distribution of slope failures in the study area (Fig. 1). Croat. j. for. eng. 38(2017)1

2.2.2 Slope failure (landslide) conditioning factors The recognition and mapping of an appropriate set of instability factors related to slope failures requires previous information on the main causes of landslides (Guzzetti et al. 1999). In this study, the

Fig. 2 General structure of ANFIS

109


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

Fig. 3 Geo-environmental parameter maps of the study area: slope degree, slope aspect, altitude, plan curvature, topographic wetness index, stream power index, sediment transport index, lithology, distances to faults, distances to streams, rainfall, normalized difference vegetation index, plant community, timber volume, and canopy

landslide conditioning factors (LCFs) were selected among the most commonly used in the literature to assess slope failures susceptibility; in particular, the results of field surveys suggested that slope degree, slope aspect, altitude, plan curvature, topographic wetness index (TWI), stream power index (SPI), sediment transport index (STI), lithology, rainfall, dis-

110

tance to faults, distance to streams, normalized difference vegetation index (NDVI), forest canopy, forest plant community, and timber volume match very well with the landslide distribution in the study area. The calculation and significance of these factors in landsliding has explicitly been presented in Pourghasemi et al. (2013), Jaafari et al. (2014), Jaafari Croat. j. for. eng. 38(2017)1


Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

et al. (2015a), and Wang et al. (2016). Fig. 3 shows the LCFs used in this study. Slope degree, slope aspect, altitude, plan curvature, TWI, SPI, and STI layers were created from a 20 m Digital Elevation Model (DEM) using ArcGIS and SAGA GIS. The geological map was prepared by GSI on a 1:100,000 scale. Distance to faults and distance to streams were computed using spatial analyst tool of ArcGIS. The rainfall map was prepared using the mean annual precipitation data from the meteorological stations for the study area over the last 20 years (Jaafari et al. 2014). Extensive investigations by the Tarbiat Modares University on the study area have been the major source of data associated with NDVI, forest plant community, forest canopy, and timber volume used in the present study. As the raster dataset has enriched the capability for spatial analysis, all factor layers were converted into raster format. Given the extent of the study area and the landslide distribution,

Dataset_5

(Model_5)

Dataset_4

(Model_4)

(Model_3)

Dataset_3

(Model_2)

Dataset_2

Dataset_1

Factor

(Model_1)

Table 1 The factor list of the datasets from 1 to 5

Slope angle

ü

ü

ü

ü

ü

Slope aspect

ü

ü

ü

ü

ü

Altitude, m

ü

ü

ü

ü

ü

Lithology

ü

ü

ü

ü

ü

Rainfall, mm

ü

ü

ü

ü

NDVI

ü

ü

ü

ü

Plan curvature

ü

ü

ü

TWI

ü

ü

ü

SPI

ü

ü

ü

STI

ü

ü

ü

Distance to streams, m

ü

ü

Distance to faults, m

ü

ü

Forest canopy,%

ü

Timber volume, m3/ha

ü

Plant community

ü

Croat. j. for. eng. 38(2017)1

A. Jaafari et al.

grid cells having a spatial resolution of 20×20 m (Bui et al. 2012, Jaafari et al. 2014, Jaafari et al. 2015a) were selected as the mapping unit. This was small enough to capture the spatial characteristics of landslide susceptibility and large enough to reduce computing complexity. A series of tests was also performed considering different input datasets from the LCFs. The purpose of selecting various datasets was to explore the influence of parameter enrichments on the performance of the ANFIS models, and the importance of the added parameter on the landslide assessments (Pradhan 2013). Table 1 shows that dataset_1 includes a maximum number of LCFs, and it continues to narrow down to dataset_5. The idea behind this kind of grouping came from the nature and the availability of data and resources of each LCF. Some factors, such as forest canopy, timber volume, and plant community, are costly to collect across forestlands in Iran due to the landscape heterogeneity and unavailability of supporting tools such as accurate high-return LiDAR data for all areas and frequent changes over a short time period due to forestry operations. Thus, they were only included in dataset_1. In contrast, the preparatory factors (e.g. slope, aspect, altitude and lithology) that are not expected to change significantly over a short time period (e.g. 50 years) are very easy to quantify using fairly simple GIS operations. These factors were, therefore, considered for inclusion in all datasets. The inclusion of other factors in different datasets also follows this instruction.

2.3 Training and validation dataset In landslide modeling, the landslide inventory map needs to be split into two subsets for training and validation. Without splitting, it would not be possible to validate the results (Jaafari et al. 2014). When splitting data, there is no rule of thumb for the relative sizes of the two subsets (Pradhan 2013). Here, the inventory map was randomly divided into two datasets. Part_1 that contains 80% of the data (82 landslides) was used in the training phase of the five ANFIS models. Part_2 is a validation dataset with the remaining 20% of the data (21 landslides) used to validate the models and to estimate their accuracy. All 82 landslide locations in the part_1 dataset denote the presence of landslides and were assigned to a value of 1. The same number of points denoting the absence of landslide were randomly sampled from the landslide free area and assigned a value of 0. Values for the 15 LCFs were then extracted to build a training dataset (Bui et al. 2012, Pradhan 2013). This dataset contains a total of 164 points, with one target variable denoting the land-

111


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

slide presence/absence and the 15 LCFs. This dataset was further randomly partitioned into two subsets including: training and checking to develop the ANFIS models. The training set was used to adjust the connections weights, membership functions and model parameters. The checking set was used to check the performance of the model through the training process and to stop the training to avoid over fitting. This method of data division is recommended to control over fitting of the models (Jang et al. 1997). In this study, approximately 70% (116 points) of the extracted database was randomly selected as the training dataset, and the remaining 30% (48 points) as the checking dataset. The commercially available Neuframe software (Neusciences 2000) was used to select the datasets at random. Due to the different scales of the input variables, and in order to increase the speed and accuracy of data processing, input data need to be normalized from 0 and 1 before using them in the ANFIS model. For this purpose, the extracted values from LCFs were normalized using the normalization formula as follows:

Xnorm =

Xi − Xmin Xmax − Xmin

(3)

Where: data that should be normalized Xi Xmax, Xmin the maximum and minimum value of original data, respectively Xnorm normalized value of Xi.

2.4 Development the ANFIS models for the spatial prediction of slope failure In this study, a type_3 ANFIS model (Takagi and Sugeno 1983) was used to produce susceptibility maps of the study area. In this type of ANFIS model, the output of each rule is a linear combination of input variables added by a constant term (Jang 1993). The final output is the weighted average of each rule’s output (Buragohain and Mahanta 2008). The general structure of a type_3 ANFIS model with two inputs of x1 and x2, and one output of y is shown in Fig. 2 (Erenturk 2009). From this figure, it can be seen that the model contains five layers: the first layer actualizes the fuzziness of inputs, the second layer calculates the firing strength of each rule, the third layer normalizes the firing strengths, the fourth layer determines the consequent parameters of the rule, and the fifth layer computes the output of the fuzzy system by summing the outputs of the fourth layer. A total of five ANFIS models were constructed to produce LS maps of the study area. To implement ANFIS, MATLAB programming language version R2011a

112

was used. GENFIS1 and GENFIS2 functions are two available methods that have been widely used to generate the initial fuzzy inference system (FIS). The GENFIS1 generates an initial Sugeno-type FIS for ANFIS training using a grid partition, and GENFIS2 uses subtractive clustering to generate the initial Sugenotype FIS. As proposed by Chiu (1997), due to the large number of input variables considered in this study, the GENFIS2 function was used to generate the initial FIS for ANFIS training by first applying subtractive clustering on the data. GENFIS2 accomplished this by extracting a set of rules that models the data behavior. After constructing the Sugeno-type FIS for the five ANFIS models, each model was trained by considering 200 epochs. Finally, the output data obtained from the models were converted to a GIS grid data to create the slope failure susceptibility maps.

2.5 Validation and comparison of susceptibility maps Prediction modeling does not have a scientific significance without computing the validity of the results. Here, the susceptibility assessment results were tested using known landslide locations. Testing was performed by comparing the known landslide location data with the landslide susceptibility map. To validate the results of the susceptibility assessment, the AUC of the ROC curve was used (Bui et al. 2012, Pourghasemi et al. 2012, Pradhan 2013, Pourghasemi et al. 2013, Jaafari et al. 2014, Jaafari et al. 2015a, Ezzati et al. 2016). The ROC curve is a graphical representation of the trade-off between the false-negative and false-positive rates for every possible cutoff value. By tradition, the plot shows the false-positive rate (FPR) on the X axis (Eq. 4) and the true-positive rate (TPR) on the Y axis (Eq. 5).

 TN  X = FPR = 1 −    TN + FP 

(4)

 TP  X = TPR =    TP + FN 

(5)

Where: TN (true negative) and TP (true positive) are the number of pixels that are correctly classified, whereas FP (false positive) and FN (false negative) are the numbers of pixels erroneously classified. The area under the ROC curve (AUC) characterizes the quality of a forecast system by describing the system’s ability to anticipate the correct occurrence or non-occurrence of pre-defined »events« (Pourghasemi et al. 2013). The best method has a curve with the largest AUC; the AUC varies between 0 and 1, where 1 Croat. j. for. eng. 38(2017)1


Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

indicates perfect prediction and, 0.5 indicates random predictions. Larger ROC value suggests better the compatibility between dependent and independent variables. The quantitative-qualitative relationship

A. Jaafari et al.

between AUC and prediction accuracy can be classified as follows: 0.9–1, excellent; 0.8–0.9, very good; 0.7–0.8, good; 0.6–0.7, moderate; and 0.5–0.6, poor (Hosmer et al. 2013).

Fig. 4 Susceptibility map produced by: (a) model_1, (b) model_2, (c) mode_3, (d) model_4, (e) model_5 Croat. j. for. eng. 38(2017)1

113


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

3. Results The susceptibility maps produced by the five ANFIS models are shown in Fig. 4a–e. In every map, the susceptibility classes of I, II, III, IV and V indicate the likelihood of slope failure (landslide) initiation, ranging from very low to very high susceptibility. A de-

tailed interpretation of susceptibility classification is presented in Table 2. This shows that each susceptibility class provides a relative ranking of the likelihood of a slope failure following road construction and/or timber harvesting. For example, the first class implies very low susceptibility to slope failure and the area characterized by this class is safe for forestry operations. The results of validation of the five ANFIS models using ROC-AUC are shown in Figs. 5 and 6. The results show that all the models have good prediction

Fig. 5 Prediction rate curves for the susceptibility maps produced in this study

Fig. 7 The landslide susceptibility classes delimited by the five ANFIS models Table 2 Detailed slope failure susceptibility classification Susceptibility class

Interpretation Safe

Fig. 6 Success rate curves for the susceptibility maps produced in this study

114

Very low likelihood of failures following road construction or timber harvesting Low instability

I

Normal road construction and timber harvesting will not significantly decrease terrain stability Moderate likelihood of failures following road construction or timber harvesting

II

Minor failures expected in road cuts High likelihood of failures following road construction or timber harvesting Very high likelihood of failures following road construction or timber harvesting

III IV V

Croat. j. for. eng. 38(2017)1


Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

capabilities, with the best results of the model_5 (AUC success rate =86.19%, AUC prediction rate =83.74%), ­f ollowed by the model_4 (AUCsuccess rate=82.23%, AUCprediction rate=75.81%). In addition, a comparison between the five susceptibility classes delimited by the different ANFIS models is presented in Fig. 7. The result suggests that the moderate, high and very high susceptibility classes cover more than 60% of the study area.

4. Discussion

A. Jaafari et al.

sults reported by Sezer et al. (2011) and Pradhan (2013) suggest that the increase in the number of LCFs has a positive impact on the overall prediction performance of LS assessment using ANFIS. The results are quite different according to various researchers and study areas. This is because there is no common guiding principle for selecting LCFs (Ayalew et al. 2005). They are usually selected based on the landslide types, the failure mechanisms, the map scale of analysis, the characteristics of the study area, and data availability (Glade and Crozier 2005).

4.1 Landslide susceptibility mapping

4.2 Landslide susceptibility maps for the safety in forestry operations

Modeling LS across a forestland is challenging because of geological, topographical and environmental complexities. Although various methods for LS assessment have been proposed, the evaluation of predictive ability of these methods in forestlands still lags. This study evaluated the predictive ability of ANFIS for modeling LS across a forestland subjected to forestry operations. Five ANFIS models developed herein offer the possibility to compare the distribution landslide of hazard with different sets of LCFs. When the ROC curves of these five models were considered together, their overall performances were close to each other. Performance validation indicated that the most successful ANFIS model is model_5, which has much fewer attributes than models 1–4. Therefore, it can be concluded that the altitude, slope angle, aspect, and lithology are most suitable LCFs for LS assessment in the study area. Moreover, these results suggest that the other LCFs are a possible source of bias because they decreased the prediction accuracy. There is always a trade-off between the quality of the data, the resources involved, and the reliability of the LS assessment. To achieve the best quality relation, it is very important to invest in landslide inventory and LCFs databases (van Westen et al. 2008). Selection of LCFs is crucial for the quality of LS models (Irigaray et al. 2007). Although some methods, such as linear correlation, Kolmogorov–Smirnov test and Genetic Algorithm (Irigaray et al. 2007, Kavzoglu, et al. 2015) have been suggested to support the optimal selection of LCFs, the standard guideline is still debated. According to Remondo et al. (2003a, 2003b), the best LS models can be produced only with the DEMderived factors. They concluded that some of the LCFs, including lithology and land cover (vegetation), improve predictions only slightly. Other factors, such as regolith thickness, do not improve the predictions at all probably because the variables are not represented accurately enough. However, the different re-

As pointed out by van Westen et al. (2006), the susceptibility classes categorized with such terms as »very high«, »high«, »moderate«, »low« and »very low« risk should be defined based on the experience of the experts with the support of sufficient models and depend on the likelihood that a slide will occur and the consequences that such an event would have for the elements at risk. In this study, each susceptibility map was assigned a set of symbol (I to V) to indicate the likelihood of slope failure (landslide) initiation. A detailed interpretation of susceptibility classification for the relative ranking of the likelihood of slope failures following road construction and/or timber harvesting has also been provided. This interpretation of susceptibility classes can be considered as a safety plan by which safety is managed on the area, as this plan indicates that each part of the area poses certain risks to road construction and timber harvesting. It is worth pointing out that the assignment and interpretation of the susceptibility classes are subjective and specifically reflect forest management considerations applied by managers who make decisions for management purposes. Therefore, contractors involved in forestry operations must have their own operational safety plans. These plans, which must be updated by contractors on a regular basis, should include safety and health policy, responsibilities, risk assessments and controls (Ryan et al. 2004). Moreover, the nature of the forestry operations implies that there can often be several different operational works close to each other. Therefore, other interpretations can also be added to the susceptibility symbol to support each part of the forestry operations. These may include soil erosion potential, risk of sediment delivery to streams, and the potential for landslide debris to enter streams (BCMOF and BCMOE 1999, Schwab and Geertsema 2010). Due to the dynamic nature of forestry operations (e.g. a road with steep cuts is constructed in a slope

Croat. j. for. eng. 38(2017)1

115


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

that was considered to be of low susceptibility), the LS maps are subject to change. The single most important contributor to long-term effectiveness of the produced LS maps is the establishment of monitoring systems to observe the changes and note when and how these changes occur. However, given that a monitoring program within a mountain forest is difficult and costly, the results of this study suggest that it be limited to the highly susceptible zones identified here. Moreover, monitoring programs can improve the confidence in predictive ability of the ANFIS models developed here. These investigations were beyond the situation and scope of this study, but they are important components that benefit more efficient planning of forestry operations.

5. Conclusion This study analyzed the potential of slope failure in a mountain forest using ANFIS models within a GIS environment. The outcome of GIS-based ANFIS application was a set of susceptibility maps, that could be used to predict the stability of slopes from 15 basic factors including slope degree, slope aspect, altitude, plan curvature, TWI, SPI, STI, lithology, rainfall, distance to faults, distance to streams, NDVI, forest canopy, forest plant community, and timber volume. The results of this study suggest that all of the five ANFIS models have performed reasonably well with AUC values over 70%. Therefore, they can be used to develop prudent hazard mitigation plans for safe forestry operations. However, the best model can only be produced with altitude, slope angle, aspect, and lithology. Forest engineers can tailor the use of these models based on their circumstances. The susceptibility assessment of slope failure is an essential resource of knowledge of the study area for its capacity to support safe forestry operations. Unfortunately, such studies are far from common in the mountainous forestlands subjected to forestry operations. This makes comparative analyses difficult. Thus, it is important to apply the method proposed here to different environmental settings.

Acknowledgement A part of this study was orally presented at the 47th International Symposium on Forestry Mechanization (FORMEC) held in France on September 2014. This study was partially supported by Tarbiat Modares University. The author gratefully acknowledges Abdullah Abbasi, Sattar Ezzati, Hamed Asadi, and Mostafa Adib for the collaboration in field surveys.

116

6. References Akgun, A., Sezer, E.A., Nefeslioglu, H.A., Gokceoglu, C., Pradhan B., 2012: An easy-to-use MATLAB program (MamLand) for the assessment of landslide susceptibility using a Mamdani fuzzy algorithm. Computers and Geosciences 38(1): 23–34. Akib, S., Mohammadhassani, M., Jahangirzadeh, A., 2014: Application of ANFIS and LR in prediction of scour depth in bridges. Computers and Fluids 91: 77–86. Allison, C., Sidle, R.C., Tait, D, 2004: Application of decision analysis to forest road deactivation in unstable terrain. Environmental Management 33(2): 173–185. Ayalew, L., Yamagishi, H., 2005: The application of GISbased logistic regression for landslide susceptibility mapping in the Kakuda–Yahiko Mountains, Central Japan. Geomorphology 65 (1–2): 15–31. BCMOF and BCMOE (B.C. Ministry of Forests and B.C. Ministry of Environment), 1999: Mapping and assessing terrain stability guidebook, 2nd edn. Forest Practices Code of British Columbia, Victoria. Bui, D.T., Pradhan, B., Lofman O., Revhaug, I., Dick O.B., 2012: Landslide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro fuzzy inference system and GIS. Computers and Geosciences 45: 199–211. Buragohain, M., Mahanta, C., 2008: A novel approach for ANFIS modelling based on full factorial design. Applied Soft Computing 8(1): 609–625. Chiu, S.L., 1997: An efficient method for extracting fuzzy classification rules from high dimensional data. Advanced Computational Intelligence 1(1): 1–7. Conforti, M., Pascale, S., Robustelli, G., Sdao F., 2014: Evaluation of prediction capability of the artificial neural networks for mapping landslide susceptibility in the Turbolo River catchment (northern Calabria, Italy). Catena 113: 236–250. Darvishzadeh, A., 2004: Geology of Iran. Amirkabir Publisher, Tehran (In Persian) Duncan, S.H., Ward, J.W., Anderson, R.J., 1987: A method for assessing landslide potential as an aid in forest road placement. Northwest Science 61(3):152–159. Erenturk, K., 2009: ANFIS-based compensation algorithm for current-transformer saturation effects. Power Delivery, IEEE Transactions on 24(1): 195–201. Ezzati, S., Najafi, A., Bettinger, P., 2016. Finding feasible harvest zones in mountainous areas using integrated spatial multi-criteria decision analysis. Land Use Policy 59: 478–491. Feizizadeh, B., Blaschke, T., 2013: GIS-multicriteria decision analysis for landslide susceptibility mapping: comparing three methods for the Urmia lake basin, Iran. Natural Hazards 65(3): 2105–2128. Croat. j. for. eng. 38(2017)1


Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

A. Jaafari et al.

Glade, T., Crozier, M.J., 2005: A review of scale dependency in landslide hazard and risk analysis. In: Glade T, Anderson M, Crozier MJ, editors. Landslide hazard and risk. England: John Wiley and Sons Ltd. 75–138.

Pourghasemi, H.R., Pradhan, B., Gokceoglu, C., 2012: Application of fuzzy logic and analytical hierarchy process (AHP) to landslide susceptibility mapping at Haraz watershed, Iran. Natural Hazards 63(2): 965–996.

Gorsevski, P.V., Gessler, P.E., Boll, J., Elliot, W.J., Foltz, R.B., 2006: Spatially and temporally distributed modeling of landslide susceptibility. Geomorphology 80(3): 178–198.

Pradhan, B., 2013: A comparative study on the predictive ability of the decision tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS. Computers and Geosciences 51: 350–365.

Guzzetti, F., Carrara, A., Cardinali, M., Reichenbach, P., 1999: Landslide hazard evaluation: a review of current techniques and their application in a multi-scale study, Central Italy. Geomorphology 31(1): 181–216. Hosmer, D.W. Jr., Lemeshow, S., Sturdivant, R.X., 2013: Applied logistic regression. John Wiley & Sons. Irigaray, C., Fernández, T.E., Hamdouni, R., Chacón, J., 2007: Evaluation and validation of landslide-susceptibility maps obtained by a GIS matrix method: examples from the Betic Cordillera (southern Spain). Natural Hazards 41(1):61–79. Jaafari, A., Najafi, A., Pourghasemi, H.R., Rezaeian, J., Sattarian, A., 2014: GIS-based frequency ratio and index of entropy models for landslide susceptibility assessment in the Caspian forest, northern Iran. International Journal of Environmental Science and Technology 11(4): 909–926. Jaafari, A., Najafi, A., Rezaeian, J., Sattarian, A., Ghajar I., 2015a: Planning road networks in landslide-prone areas: A case study from the northern forests of Iran. Land Use Policy 47: 198–208. Jaafari, A., Najafi, A., Rezaeian, J., Sattarian, A., 2015b: Modeling erosion and sediment delivery from unpaved roads in the north mountainous forest of Iran. International Journal on Geomathematics 6(2): 343–356. Jang, J.S., 1993: ANFIS: adaptive-network-based fuzzy inference system. Systems, Man and Cybernetics, IEEE Transactions on 23(3): 665–685. Jang, J.S.R., Sun, C.T., Mizutani, E., 1997: Neuro-Fuzzy and Soft Computing: A Computational Approach to Learning and Machine Intelligence (Matlab Curriculum Series: Prentice Hall. Kavzoglu, T., Sahin, E.K., Colkesen, I., 2015: Selecting optimal conditioning factors in shallow translational landslide susceptibility mapping using genetic algorithm. Engineering Geology 192: 101–112. Larsen, M.C., Parks, J.E., 1997: How Wide is a Road? The Association of Roads and Mass-Wasting in a Forested Mountain Environment. Earth Surface Process and Landforms 22(9): 835–848. Neusciences., 2000: Neuframe Version 4.0. Neusciences Corp., Southampton. Hampshire, United Kingdom. Pourghasemi, H.R., Moradi, H.R., Aghda, S.F., 2013: Landslide susceptibility mapping by binary logistic regression, analytical hierarchy process, and statistical index models and assessment of their performances. Natural Hazards 69(1): 749–779. Croat. j. for. eng. 38(2017)1

Remondo, J., González, A., Díaz de Terán, J.R., Cendrero, A., Fabbri, A., Chung, C.J.F., 2003b: Validation of landslide susceptibility maps; examples and applications from a case study in Northern Spain. Natural Hazards 30(3):437–449. Remondo, J., González-Díez, A., Díaz de Terán, J.R., Cendrero, A., 2003a: Landslide susceptibility models utilizing spatial data analysis techniques. A case study from the Lower Deba Valley, Guipúzcoa (Spain). Natural Hazards 30(3): 267–279. Ryan, T., Phillips, H., Ramsay, J., Dempsey, J., 2004: Forest Road Manual: Guidelines for the design, construction and management of forest roads. Dublin: National Council for Forest Research and Development. 170 p. Schwab, J.W., Geertsema, M., 2010: Terrain stability mapping on British Columbia forest lands: an historical perspective. Natural Hazards 53(1): 63–75. Sessions, J., Balcom, J.C, Boston, K., 1987: Road location and construction practices: effects on landslide frequency and size in the Oregon Coast Range. Western Journal of Applied Forestry 2(4): 119–124. Sezer, E.A., Pradhan, B., Gokceoglu, C., 2011: Manifestation of an adaptive neuro-fuzzy model on landslide susceptibility mapping: Klang valley, Malaysia. Expert Systems with Application 38(7): 8208–8219. Takagi, T., Sugeno, M., 1983: Derivation of fuzzy control rules from human operator’s control actions In: Proceedings of IFAC Symposium on Fuzzy Information, Knowledge Representation and Decision Analysis, 55–60. Vahidnia, M.H., Alesheikh, A.A., Alimohammadi, A., Hosseinali, F., 2010: A GIS based neuro-fuzzy procedure for integrating knowledge and data in landslide susceptibility mapping. Computers and Geosciences 36(9): 1101–1114. van Westen, C.J., Castellanos, E., Kuriakose, S.L., 2008: Spatial data for landslide susceptibility, hazard, and vulnerability assessment: an overview. Engineering Geology 102(3): 112–131. van Westen, C.J., van Asch, T.W., Soeters, R., 2006: Landslide hazard and risk zonation—why is it still so difficult?. Bulletin of Engineering geology and the Environment 65(2): 167–184. Wang, Q., Li, W., Wu, Y., Pei, Y., Xie, P., 2016: Application of statistical index and index of entropy methods to landslide susceptibility assessment in Gongliu (Xinjiang, China). Environmental Earth Sciences 75(7): 1–13.

117


A. Jaafari et al.

Spatial Prediction of Slope Failures in Support of Forestry Operations Safety (107–118)

Authors’ address: Abolfazl Jaafari, PhD. * e-mail: ajaafari@gmail.com Young Researchers and Elite Club Karaj Branch, Islamic Azad University Karaj IRAN

Received: April 29, 2016 Accepted: October 15, 2016

118

Javad Rezaeian, PhD. e-mail: j_rezaeian@ustmb.ac.ir Masoud Shafipour Omrani, Msc. e-mail: ie.shafipour@gmail.com Mazandaran University of Science and Technology Department of Industrial Engineering Babol IRAN * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads Kevin Boston, Ben Leshchinsky, Erica Kemp, Robin Wortman Abstract Due to the dispersed nature of forestry operations in much of the world, only a subset of a given forest road network are used in any year. Specifically, spur roads are generally only used when harvesting operations are adjacent to roadways; otherwise, they remain unused or only have infrequent administrative traffic. During these periods of light use, a substantial amount of organic litter may build up on the roads. As this detritus accumulates and decays, it creates conditions that encourage the growth of unwanted vegetation in the roadway, accelerating the contamination of the surface aggregate. This organic material can degrade the road by retaining moisture and creating a less tractive road surface. Contemporary forest practices control this unwanted vegetation by using a combination of grading or herbicides, at significant expense. One potential alternative treatment is the utilization of a rotary-mounted asphalt broom for vegetation and debris removal. A series of field trials were performed on wet, contaminated forest roads, in which we evaluated vegetation, debris removal effectiveness, and tire slip on segments of road before and after sweeping. The combined effects of wire and synthetic bristles on the rotary broom proved effective in both increasing traction and removing unwanted debris and vegetation from the road surface. Application of this technique was expedient, and did not result in significant loss of surface aggregate, removing on average less than 1% of the aggregate surface. Keywords: forest road maintenance, rotary brooms

1. Introduction Harvesting patterns and systems dictate the use of road systems. Adjacency or »green-up« rules promote the dispersion of even-age harvest activities throughout forested landscapes (Boston and Bettinger 2006). In the western United States, these adjacency rules are commonly included as part of state forest practice regulations, as well in the Sustainable Forestry Initiative and Forest Stewardship Council certification schemes. These rules limit the opening size in a range from as small as 8 hectares in California to as large as 97 hectares in Washington (Boston and Bettinger 2001, Boston and Bettinger 2006). Adjacency rules are not limited to US private forest practices, as they are often a component of both the Forest Stewardship Council and Sustainable Forestry Initiative certification programs. In the Australian state of Victoria, the maximum opening size is 40 hectares (Boston and Bettinger 2006), while in Sweden restrictions limit opening sizes to less than 20 hectares (Carvajal et al. 2013). Croat. j. for. eng. 38(2017)1

These spatial planning rules result in greater dispersion of harvesting activities, causing many roads to be inactive for extended periods of time. During these times of inactivity, the surfaces of forest roads can accumulate a large amount of organic material from a variety of sources. This material can fall from neighboring trees or be deposited by the wind from the surrounding forest. As organic matter accumulates, it facilitates the growth of grasses and other vegetation that can further contaminate the aggregate surfacing. In areas with abundant moisture, such as the Oregon Coast Range, unused roads can become quickly overgrown with vegetation between each use. Some companies have employed »daylighting« around their roads, which removes the trees near the roads to reduce the amount of material on the road and allow the road to dry more quickly (Kochenderfer 1970). However, the resulting space could support additional wood production, and the tradeoff between width of the buffer strip and the impact on the road has not been explored.

119


K. Boston et al.

The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

The accumulation of organic material on the roadway can result in a variety of hazards and management issues, notably: loss of traction, loss of drainage capabilities, increased fire risk, and increased grading and maintenance costs. A slick, thin film of wet vegetation and debris results in a reduction in traction, a driving hazard. When the leaf litter is wet, light vehicles may experience tire spin, which often results in a temporary loss of traction on the steeper grades (personal observation Boston 2015). Furthermore, this vegetation can impair road drainage systems by fouling voids between aggregate surfacing as well as concentrating water in the vehicle wheel-path, creating a preferential flow channel that can facilitate erosion. Vegetation, especially grasses, can increase susceptibility to fire. In the late summer or autumn in a Mediterranean climate, one characterized by dry summers and wet winters, these grasses dry out and can be ignited by the exhaust systems from vehicles (Wilson 1979). Modern exhaust systems can produce temperatures in excess of the ignition temperature of vegetation commonly found growing on forest roads. The USDA Forest Service recommends that people avoid parking vehicles on dry vegetation (USFS 2014) due to this increased fire risk. Repair and maintenance of roads covered in organic debris often rely on mechanical grading for vegetation removal or resurfacing. Additional aggregate is often placed on top of the road, which can cost as much as $20.00 per cubic yard in regions where highquality aggregate is scarce (Sessions et al. 2006). Often, an herbicide treatment is used as well to inhibit or kill road vegetation. However, the use of herbicides is largely unavailable for publicly managed land in the United States, due to public concern about the environmental impact of these chemicals. Thus, there is a need for a low-cost means of removing organic debris and vegetation from forest roads while maintaining the aggregate surfacing. The purpose of this project is twofold: Ăž quantify the amount of material that falls on forest roads from the surrounding forest in the span of one year Ăž determine if a self-propelled, rotary asphalt broom can effectively remove this accumulated organic material without damaging the surface of the aggregate road and therefore offer an alternative to grading roads.

2. Literature review Aggregate roads are the primary means to access natural resources in much of the world where year-

120

round hauling is required. Grading is the primary means for maintaining these roads, although often at significant cost. Grading has both benefits and disadvantages: it removes ruts and restores road grade, but simultaneously breaks any natural armoring of the surface layer and thus increases sediment movement from the road. Sediment curves show an initial plume of sediment that reduces with time following road grading (Sugden and Woods 2007). Purposeful selection of only those road segments that require grading, rather than grading the entire road, may reduce road maintenance costs and lessen environmental impacts (Thompson et al. 2007). Street sweeping with a mechanical rotary broom is a proven means of removing debris and litter from paved streets, and is sometimes used to help remove storm water pollution from paved streets. This technique was shown to remove 84.5 kg of sediment from streets, while eliminating the discharge of heavy metals like copper or zinc that accumulate from brake pads, from a 10 meter wide, 260 meter long strip (German and Svensson 2002). However, the impact of sweeping has never been documented for aggregate roads. One goal for this project is to determine the effectiveness of sweeping for the purpose of removing the accumulation of organic material on aggregate roads in a forested environment.

2.1 Litter accumulation Litter accumulation can vary significantly by tree species and site index, as both influence the efficiency of biomass production. Fried et al. (1990) collected tree litter from the Oregon coast range, finding that sites dominated by big-leaf maple (Acer macrophyllum) produced the most biomass, with a range between 40 to 276 grams of dry matter per square meter. They found that big-leaf maple leaves were the most common element found per unit of biomass during the year. In winter, the litter traps contained a mixture of big-leaf maple and Douglas-fir (Pseudotsuga menziessii), comprised of a material that was a combination of leaves, needles, and small branches (Fried et al. 1990). There are no known estimates for litter fall when a road divides the plot, as the opening may alter wind patterns.

2.2 Traction Wheel slip is defined as 1 minus the ratio between the linear speed of the tire and rotational or angular velocity of a wheel, which is the product of the wheel radius and angular speed, Eq. 1 (Wong 2001). As the resisting forces increase, the tractive effort will need to proportionally increase. Thus, there will be a resulting increase in wheel slip under steady-state conditions to Croat. j. for. eng. 38(2017)1


The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

produce that slip (Wong 2001). The maximum tractive effort is often produced when the wheel slip is between 15 and 20%. Wheel slip beyond 20% can result in an unstable condition (Wong 2001). Thus, wheel slip will increase with increasing grade if the truck is a constant speed, as greater thrust is needed to overcome resistance from steeper grades. This will continue until the tire spins and the vehicle stalls. One method to reduce necessary slip is to increase the adhesion between the tire and road surface. The accumulation of leaves has been identified as a major cause in loss of adhesion between the wheel and rail for trains in the United Kingdom (Gallardo-Hernandez and Lewis 2008), but no study has evaluated the impact of this material on vehicle performance on aggregate roads. Roads that have high slip or wheel spin when a vehicle passes are a safety hazard. Slip is defined as:

Slip = 100 × (1–V/rw)

(1)

Where: V linear velocity of the vehicle at the tire center r rolling radius of the tire, and ω angular speed of the tire.

3. Methods A series of field tests were performed along a 650 m (2000 foot) segment of single-lane, aggregate-surfaced forest road in the Oregon State University Dunn Forest. This road was selected because it had new surface aggregate placed on the road in 2006, but had not yet experienced hauling or heavy truck traffic at the time of the study. This lack of traffic, as well as grades between 2% and 16% and surrounding plant communities of Douglas-fir and big-leaf maple, presented a scenario representative of unused forest roads found in western Oregon. No attempt was made to generate a sample from the larger population of roads in western Oregon, as further research is needed to infer a larger range of site conditions. Twenty one-square meter litter traps were installed in September 2013 with 30 meter spacing along the margins of the road to measure the litter fall on the road. Adequate clearance was created to allow managerial traffic access without disturbing the traps. Litter trap installation was timed to begin monitoring prior to significant autumn litter fall, with monitoring continuing until June 2014. Material was collected at approximately monthly intervals. The northwest corner of each litter trap contained a plot center, and a 4.52 square meters per hectare (20 square feet per acre) BAF prism was used to measure the basal area surrounding the plots. The tree species, diameter, and total height Croat. j. for. eng. 38(2017)1

K. Boston et al.

were measured for each tree located in the variable radius plot. In April 2014, sweeping of the roads was performed. Three passes were used with a 2.4 m (8 foot) self-propelled rotary asphalt broom. The broom condition was nearly new, with less than 25 hours of cumulative use. The 18,642 watt (25 horsepower) engine was able to turn the 2.4 m broom at 3000 rpm. The broom had a mixture of nylon and metal bristles. During the sweeping, the bristles were in contact with the road surface. Changes in the road surface were evaluated postsweeping for each of the seven five-meter road segments with constant grades. Each road segment was further divided into one-meter segments and marked using paint on the edge of the road, in order to better view the tire slip while the vehicle motion was being recorded. The effectiveness of the sweeping was evaluated using two methods. The first method estimated the change in the vegetation cover. A sampling guide composed of six 7.5 cm squares was randomly assigned to a location within each 5 m test strip. Photographs were taken before and after sweeping at the same location. The change in the vegetative debris cover of the road following sweeping was evaluated using ocular estimates. Results were compared with plant-density guides of known coverage to support these estimates. The second method measured treatment efficacy for traction improvement, and involved vehicle testing to evaluate wheel slip before and after sweeping. The ve-

Fig. 1 Litter in leaf trap based on basal area of surrounding forests

121


K. Boston et al.

The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

Fig. 2 Amount of litter from leaf traps hicle, a 750 kg two-wheel drive pickup truck with 76 cm (30-in) diameter tires, was driven at constant speeds of 5 and 10 mph. Three passes were made at each speed. The tire pressure was constant for all four tires at 379 kPa (55 psi). The right rear tire had unique marks placed on it to clearly identify the number of rotations of the tire. Each pass was recorded using timed video recordings (accurate to milliseconds), which evaluated quantification of wheel rotations and longitudinal velocity of the vehicle across the 5 meter test strip; thus, wheel slip could be calculated for each pass. These evaluations were performed before and after sweeping to determine the change in wheel slip due to sweeping.

0.482; furthermore, this relationship explains only 8% of the variation in the amount of litter captured by the traps. Thus, there is a poor relationship between dry material collected on the traps and total basal or conifer basal area near each trap.

4.2 Cover changes There was a statistically significant difference in the percent cover between the swept and non-swept sections with a p-value of 0.0023. The pre-swept cover

4. Results 4.1 Litter accumulation The leaf litter deposited in the litter traps was a combination of leaves from the broadleaf trees, needles from the conifer trees, seeds, cones, and small branches. The bulk of the autumn litter consisted of big-leaf maple leaves, with more conifer needles and branches being deposited in the winter (see Fig. 1). The average dry mass of the material deposited on the trap was 309 grams per square-meter with a standard deviation of 133 grams per square meter (Fig. 2). The minimum dry weight was 138 grams of dry litter per square meter and the maximum dry weight was 640 grams of dry litter per square meter. A linear regression was performed, but no significant relationship was shown between total or conifer basal area and leaf litter deposited in traps with a p-value of

122

Fig. 3 Changes in percent of road covered by organic material Croat. j. for. eng. 38(2017)1


The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

K. Boston et al.

Fig. 6 Comparison of slip between treated and untreated sections from the seven road grade categories

Fig. 4 Pre-treatment condition

analysis documented the accumulation of organic material on the road. Seventy percent of the roadway was covered in organic matter before treatment, with a standard deviation of 10.6%. Following the treatment, the percent cover was reduced to 9.4% with a standard deviation of 7.87% (Fig. 3). Fig. 4 and 5 are representative pictures of before and after sweeping, which show respectively the accumulation of moss, grasses, and tree debris found on these unused roads and the effectiveness of the sweeper at removing this material from the road bed. Aggregate loss from sweeping was approximately 4.5 kg per linear meter of road (with an average road width of 3 meters that is less than 1% of the aggregate amount on a typical road section). Of this material, a grain size distribution was collected for each section (Fig. 7), demonstrating a mean grain size of 0.5 cm, notably smaller than the aggregate surfacing material (5 cm in diameter). Attrition (i.e., material removed) was primarily comprised of material between 0.07 and 1.27 centimeters in diameter.

4.3 Changes in slip

Fig. 5 After treatment condition Croat. j. for. eng. 38(2017)1

The average slip was measured from the video recording of the truck driving each of the seven sections of the road. The pre-treatment slips measured between 5 and 20%. Slip increased, as expected, with increasing grade. Following the sweeping, the average slip was reduced in all cases. It ranged between 2 and 19% (Fig.

123


K. Boston et al.

The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

Fig. 7 Grain size distribution of material lost during sweeping 6). The greatest reduction of slip occurred at the 16% road grade, where the post-treatment slip was lowered by nearly 50% (Fig. 6). Other sections show a lesser reduction of slip, potentially due to free water left on the aggregate surfacing post-sweeping. The result was a statistically significant difference between the swept and non-swept sections of road with a p-value value of 0.013.

Fig. 8 Distribution of weight of material lost during sweeping (per kg lost)

124

5. Discussion The litter deposition along the road showed similar quantities to those described by Fried et al. (1990), who reported between 40 and 276 grams per square meter of litter in plots that were not bisected by a road. Our results showed a mean value of 138 grams per square meter. This suggests that roads with a basal area between 9 and 41 m2/ha will result in litter falls that are similar to full forests. However, basal area or percent crown closure were not good predictors of the litter accumulation on this site. The high variability of the material in the traps makes prediction difficult, as small branches from windfall constituted the majority of material in the litter traps by weight. These traps appeared to be influenced by the complex wind pattern from the forest, with the variety of nearby openings that tend to occur in a working landscape. The differences in the percent road surface following the treatment demonstrated the effectiveness of rotary brooms at removal of the organic material, including rooted grasses and moss, from forest roads. Saturation of the surface material may have facilitated easier removal of this material due to increased adhesion between the bristles of the broom and the litter. Further testing is needed to determine if the broom is equally effective during dry soil conditions. Few stones were dislodged from the surface during sweeping, typically about 4.5 kg of material per linear meter of road were moved. The majority of material removed was smaller grains, falling within the range Croat. j. for. eng. 38(2017)1


The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

of fine to coarse sand. The surface of the road was therefore firm when sweeping was completed. Future research should monitor the sediment production from the road and changes in water flow patterns, to determine if sweeping modifies these processes. There was a significant reduction in slip on the forest roads after sweeping, especially on the moderately steeper grades. The cleaned aggregate road provides a more tractive surface with greater friction for tires than the saturated vegetation that usually covers the roads in western Oregon during the winter. During one of the pre-treatment test drives on the 16% grade section, the vehicle exhibited a loss of the traction with high tire spin and rear-end fishtailing. Following treatment, this temporary loss of control was not observed.

6. Conclusion A road in mature forested surroundings can accumulate large amounts of organic material, which can cover much of the road’s surface. Total basal area and basal area by conifer were not satisfactory predictors of biomass accumulation on this road for these tree species. Sweeping with a rotary asphalt broom, with a combination of wire and synthetic fibers, proved to be an effective treatment for removal of vegetation and organic material from the surface of the road. It removed a majority of moss, grasses and debris on the road surface. Additionally, it did not remove larger stones from the road surface. The result was a road with clean aggregate that offered an improved driving surface, demonstrated by a reduction in the wheel-slip needed to travel on seven test sections. This test demonstrates that sweeping may be a useful treatment for remediating forest roads that are covered in organic debris, as it offers a cost effective and safe alternative to expensive conventional grading or the controversial application of herbicides.

K. Boston et al.

Boston, K., Bettinger, P., 2006: An economic and landscape evaluation of the green-up rules for California, Oregon, and Washington (USA). Forest Policy & Economics 8(3): 251–266. Carvajal, R., Constantino, M., Goycoolea, M., Vielma, J.P., Weintraub, A., 2013: Imposing connectivity constraints in forest planning models. Operations Research 61(4): 824–836. Fried, J.S., Boyle, J.R., Tappeiner II, J.C., Cromack Jr., K., 1990: Effects of bigleaf maple on soils in Douglas-fir forests. Canadian Journal of Forest Research 20(3): 259–266. German, J., Svensson, G., 2002: Metal content and particle size distribution of street sediments and street sweeping waste. Water Science & Technology 45(6): 191–198. Gallardo-Hernandez, E.A., Lewis, R., 2008: Twin disc assessment of wheel/rail adhesion. Wear 265(9–10): 1309–1316. Kochenderfer, J.N., 1970: Erosion control on logging roads in the Appalachians. Research Papers. Northeastern Forest Experiment Station (NE–158). Luce, C.H., Black, T.A., 1999: Sediment production from forest roads in western Oregon. Water Resources Research 35(8): 2561–2570. Lugo, A.E., Gucinski, H., 2000: Function, effects, and management of forest roads. Forest ecology and management 133(3): 249–262. Sessions, J., Boston, K., Thoreson, R., Mills, K., 2006: Optimal policies for managing aggregate resources on temporary forest roads. Western Journal of Applied Forestry 21(4): 207– 216. Sugden, B.D., Woods, S.W., 2007: Sediment production from forest roads in western Montana. Journal of the American Water Resources Association 43(1): 193–206. Retrieved from http://ezproxy.humboldt.edu/login?url=http://search.proquest.com/docview/201309732?accountid=11532. Thompson, M., Boston, K., Arthur, J., Sessions, J., 2007: Intelligent Deployment of Forest Road Graders. International Journal of Forest Engineering 18(2): 15–23. Wilson, C.C., 1979: Roadsides-Corridors with High Fire Hazard and Risk. Journal of Forestry 77(9): 576–580.

7. References

Wong, J.Y., 2001: Theory of ground vehicles. John Wiley & Sons.

Boston, K., Bettinger, P., 2001: The economic impact of greenup constraints in the southeastern United States. Forest Ecology and Management 145(3): 191–202.

USDA Forest Service (http://www.fs.usda.gov/detail/kisatchie/home/?cid=fsbdev3_024701, 2014). Accessed June 15, 2014.

Croat. j. for. eng. 38(2017)1

125


K. Boston et al.

The use of a Rotary Asphalt Broom to Groom Aggregate Forest Roads (119–126)

Authors’ address:

Received: July 31, 2015 Accepted: September 05, 2016

126

Assoc. prof. Kevin Boston, PhD. * e-mail: Kevin.Boston@oregonstate.edu Assist. prof. Ben Leshchinsky, PhD. e-mail: Ben.Leschinsky@oregonstate.edu Erica Kemp, MsC. e-mail: Erika.Kemp@oregonstate.edu Robin Wortman e-mail: Robin.Wortman@oregonstate.edu Oregon State University Corvallis, OR 97331 USA * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Different Organizational Models of Private Forest Owners as a Possibility to Increase Wood Mobilization in Slovenia and Serbia Špela Pezdevšek Malovrh, Peter Kumer, Predrag Glavonjić, Dragan Nonić, Jelena Nedeljković, Bratislav Kisin, Mersudin Avdibegović Abstract The importance of renewable energy resources has increased over the last decades due to the European Union renewable energy policy and particularly its climate change mitigation objectives. There is a need to mobilize additional wood resources from private forests in order to meet ambitious renewable energy targets and the demand for wood. Due to the conditions prevailing in privately owned forests in Slovenia and Serbia characterized by a large number of still disorganized private forest owners with fragmented and small-scaled forest properties, wood mobilization strongly depends on owners’ organization and cooperation. The purpose of this study is to determine the possibilities for wood mobilization from private forest properties in Serbia and Slovenia, and propose organizational models on this basis and experience from the selected case countries. Surveys were conducted in Slovenia (n=622) and Serbia (n=248) on random samples of private forest owners. Analysis of wood mobilization potentials in Serbia and Slovenia showed that the harvesting intensity in private forests is below the potentials, therefore the preconditions to increase the level of wood mobilization exist. The main obstacles to the increase in the current level of wood mobilization in Serbia are biodiversity and the protective forest function, as well as high acquisition costs, also stated as the main obstacle in Slovenia. Moreover, it appeared that the majority of private forest owners in both countries believe that better logistics and infrastructure and interest association of private forest owners are potential solutions leading to an increase in the level of mobilization. Four models of private forest owner organization are proposed and they take into account the characteristics and attitudes of owners as well as activities in supply chain, including timber sales arrangement, construction and maintenance of forest roads, harvesting, measurement and quality assessment of timber, transportation, invoicing and payments. Keywords: renewable energy resources, wood mobilization, private forest owners, organizational models

1. Introduction The importance of renewable energy sources has increased over the last decade particularly due to the European Union renewable energy policy and its climate change mitigation objectives. Energy security issues, rural policies as well as income and employment generation related to bioenergy production have all played important roles (Hatemaki et al. 2014). Renewable energy policies have consequently been developing rapidly, culminating with the European Croat. j. for. eng. 38(2017)1

Union (EU) Renewable Energy Directive 2009/28/EC (hereafter EU-RED), which sets mandatory targets to all member states. The EU will have reached a 20% share of energy from renewable sources by 2020 (Directive 2009, Mantau et al. 2010, Blennow et al. 2014, Halder et al. 2014, Posavec et al. 2015). In addition to the EU-RED, the importance of renewable energy sources has also been recognized in the EU Forest Strategy (EC 2013) and Climate and Energy Framework for 2030 (EC 2014). The EU Forest Strategy argues that forest-based biomass »is gaining market

127


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

interest« providing »opportunities to maintain or create jobs and diversify income in a low-carbon green economy« (EC 2013). Furthermore, it is noted that »according to the National Renewable Energy Action Plans, biomass will still be the main source of renewable energy in 2020« (EC 2013). In addition to other activities, strategic orientations defined by this document include the following: a) the exploration and promotion of a fuller use of wood as a sustainable, renewable, climate and environment friendly raw material and b) the assessment of potential wood supplies and facilitation of increased sustainable wood mobilization (EC 2013). In order to meet ambitious renewable energy targets, it is necessary to imply a far more intensive use of forest resources (Schwarzbauer 2010) and mobilize additional wood resources (mainly from fragmented private forests) to meet the demand for wood (Rauch and Gronalt 2005, Lindstad et al. 2015). Based on EU-RED and the recognized importance of renewable energy sources, EU countries (including Slovenia) have developed and implemented their National Renewable Energy Action Plans. In addition to the promotion of production and use of energy wood from forests (Beurskens and Hekkenberg 2011), they include national policies and policy recommendations for the development of renewable resources. Policy recommendations have been successfully converted to measures leading to an increased mobilization of wood. Serbia initiated the process of harmonization of national legislation with the EU policy concerning renewable energy as part of its pre-accession negotiations. Therefore, Serbia adopted the National Renewable Energy Action Plan until 2020 (NREAP 2013), defining clear objectives in terms of conditions for energy production from renewable energy sources. In these strategic plans private forests were addressed in terms of their wood mobilization potentials. Considering that private forests in Slovenia and Serbia are characterized by a large number of still disorganized private forest owners (hereinafter PFOs) of fragmented and small forest properties and their continuous fragmentation (Glück et al. 2010, Glück et al. 2011, Pezdevšek Malovrh et al. 2011), wood mobilization will strongly depend on owner readiness to supply woody biomass to the energy market (Posavec et al. 2015, Nonić et al. 2015). In addition, Blennow et al. (2014) report that despite great potentials and needs for additional wood mobilization, a growing number of PFOs in Europe (mostly with fragmented forest properties) do not participate in market wood supply. Therefore, PFOs cannot be expected to supply the amounts of woody biomass for energy required to

128

meet the forest biomass share of EU 2020 renewable energy targets. The most important problems affecting wood mobilization are forest property fragmentation, the lack of PFOs organizations and insufficient motivation of PFOs for harvesting (EC 2008). Rauch and Gronalt (2005) state that the problem of low wood mobilisation is the result of structural disadvantages in small-scale forest properties, bad market position of PFOs, the lack of forest management knowledge and experience, low volumes supplied per forest owner, low machine utilisation and difficulties in promotion. Additional reasons for the low level of wood mobilization from small-scale forest properties are the lack of time needed for wood felling (Suda and Warkotsch 2002), the increase of felling costs, the age of PFOs (Bolkesjo and Baardsen 2002), the low level or the lack of profits from forest management, incomes independency from forestry, the lack of knowledge and skills related to forest management and the lack of cooperation among PFOs (Stern et al. 2013). In addition, there are new types of PFOs who do not want to fell trees, since they primarily value their forest as a place for leisure or hunting (Boon et al. 2004, Hogl et al. 2005, Ní Dhubháin et al. 2007, Pezdevšek Malovrh et al. 2015, Živojinović 2015). Fundamental approaches that can lead to increased wood mobilization are based on PFOs cooperation and the formation of more PFOs associations and cooperatives (Glück 2002, Nichiforel and Schanz 2009, Becker 2010, Schwarzbauer et al. 2010, Mendes et al. 2011). In addition, it is necessary for PFOs to overcome their distrust towards the existing organizations primarily in the new EU member states as a result of past negative experience caused by general collectivism (EC 2008). Schwarzbauer et al. (2010) argue that wood mobilization is particularly high in formal forms of cooperation (different forms of partnerships, associations or PFO cooperatives). Moreover, Becker (2010) highlighted cluster initiative and local forest management cooperatives as forms of PFO organization with the highest significance for wood mobilization. The classification of PFOs into groups, their attitudes (Schaffner 2008) toward wood mobilization (Huber et al. 2013) and motivation have great importance for the solving of this problem. The aim of this paper is to determine the possibilities for wood mobilization from private forest properties in Serbia and Slovenia on the basis of: 1) PFOs characteristics, 2) wood potential for mobilization as the difference between the increment and harvesting rate and 3) the attitudes of PFOs toward wood mobilization. Based on the results and models existing in the selected case countries, different PFOs organizaCroat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

tion models are proposed in order to increase wood mobilization. The following activities in the supply chain were analyzed in the determination of the models: timber sales arrangement, construction and maintenance of forest roads, harvesting, measurement and quality assessment of timber, transportation, invoicing and payment.

2. Background 2.1 Brief description of private forests Private forests are an important resource of national economies in Serbia and Slovenia. Forest cover accounts for 29.1% (2,252,400 ha) of the territory in Serbia, of which 47% (or 1,058,400 ha) are privately owned forests (Banković et al. 2009). Private forests are characterized by small-scale and fragmented forest properties owned by a large number of forest owners (about 900,000). More than 72% of the owners have properties smaller than 1 ha, while the average forest property size reaches 1.27 ha (Glück et al. 2011). Since 2006, some »large« private forest owners (churches and religious communities) have emerged as a result of the restitution process. By the end of 2014, 23,195 ha of forests and forest land were returned to churches and religious communities (Restitution 2014). The process of property restitution to churches, religious communities and physical persons has not been completed yet. In Serbia private forest owners associations (PFOAs) do not have great significance and impact on forest policy. The first PFOAs were established in Serbia with the assistance of FAO projects in 2006 (Nonić et al. 2010). By 2015, 22 PFOAs were established at the local level, as well as the Serbian Federation of Private Forest Owner Associations as the umbrella organization in 2009. However, due to the necessary change in their legal form, only eight PFOAs have continued to be active after 2011. In Slovenia, forest cover accounts for 58.4% of the territory (1,183,433 ha). According to data from the 2010–2020 forest management plans, Slovenian PFOs control a larger share of the country’s forests than in any other country in the region (76% of approximately 1.2 million ha). The property is divided into approximately 314,000 individual plots, owned by roughly half a million owners. Individual properties are mostly small (64% less than 1 ha) and fragmented, while individual owners possess three plots on average (Pezdevšek Malovrh et al. 2010). That situation resulted in an underutilized management of private forests (harvesting rate below potential), which hinders wood mobilization. Although PFOAs started to Croat. j. for. eng. 38(2017)1

Š. Pezdevšek Malovrh et al.

develop in Slovenia at the beginning of the 2000s, membership is still low (less than 1% of owners are members of PFOAs). Thirty local PFOAs were established by 2015 (Leban 2014). In addition to local PFOAs, the Association of Private Forest Owners was established at the national level in 2006. Its main goals are to promote cooperation among owners, support the establishment of new local associations and facilitate links between the public forest administration and private forest owners (Mori et al. 2006).

2.2 Potentials of wood mobilization from private forests The potentials of wood mobilization from private forests in Serbia and Slovenia was analyzed on the basis of data obtained from public forest administrations. The realizable potential was calculated as the difference between the annual growth increment and the volume of wood harvested and was presented as the percentage of utilization (Tables 1 and 2). In the Serbian study area, the overall potential for wood mobilization is 616,689 m3/yr, and the average utilization amounts to 57% of the annual growth increTable 1 Potentials for wood mobilization from private forests in the Serbian study area Forest region

Increment, m3/yr Annual cut, m3/yr

Realization, %

Belgrade

2353

656

28

Kučevo

47,970

50,517

105

Boljevac

98,278

49,815

51

Despotovac

26,595

23,820

90

Kragujevac

30,942

7954

26

Loznica

75,849

22,144

29

Užice

30,776

13,750

45

Prijepolje

37,191

19,295

52

Ivanjica

28,875

11,895

41

Raška

22,022

9038

41

Kraljevo

17,257

9398

54

Kruševac

21,900

18,859

86

Kuršumlija

31,129

18,384

59

Niš

21,194

4307

20

Pirot

20,207

17,682

88

Leskovac

61,804

37,897

61

Vranje

42,347

37,917

90

Total

616,689

353,328

Average: 57

Source: PE »Srbijašume« 2013

129


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Table 2 Potentials for wood mobilization from private forests in Slovenia Regional unit

Increment, m3/yr Annual cut, m3/yr Realisation, %

Tolmin

467,508

238,086

51

Bled

18,778

187,158

101

Kranj

388,995

263,703

68

Ljubljana

618,765

428,648

69

Postojna

238,533

206,704

87

Kočevje

257,199

185,747

72

Novo mesto

428,966

299,006

70

Brežice

311,833

340,363

109

Celje

365,079

228,999

63

Nazarje

271,003

206,396

76

Slovenj Gradec

236,247

214,013

91

Maribor

460,081

324,307

70

Murska Sobota

126,244

114,705

91

Sežana

239,204

121,154

51

Total

4,594,435

3,358,988

Average: 73

Source: Report of Public Forestry Service of Slovenia about forests for the year 2014, 2015

ment (PE »Srbijašume«, 2013). Based on the data for 2014, the total available potential for wood mobilization in Slovenia is 4,594,435 m3/yr, and the average utilization amounts to 73% of the annual growth increment (Report of Public Forestry Service of Slovenia about forests for the years 2014, 2015).

3. Methods 3.1 Survey method Similar representative nationwide surveys were administered to private forest owners in Serbia and Slovenia with some variation in accordance with the country-specific conditions mainly in organization of forestry sector, in order to determine the possibilities for wood mobilization. The questionnaires were developed as a result of literature analyses and previous socio-economic research related to owner attitudes, motivation and behavior related to the management of their forest properties and wood mobilization (Pezdevšek Malovrh 2010, Glück et al. 2011, Pezdevšek Malovrh et al. 2015, Posavec et al. 2015). The survey questioned owners about a range of issues, and sev-

130

eral questions were analyzed in relation to the research aims. These were related to the mobilization of wood resources (PFOs attitudes towards mobilization, obstacles or problems that prevent them to increase the level of mobilization), forest management and sociodemographic characteristics of PFOs. Personal data about PFOs were found in the encrypted relational databases of the Land and Property Register obtained from the Surveying and Mapping authority of the Republic of Slovenia (SMARS 2007) in Slovenia and from the public enterprise for state forest management »Srbijašume« and Republic geodetic authority in Serbia. Therefore, our target population in both countries consisted of individual PFOs. At the time of the research, 330,949 distinct PFOs were listed in the Slovenian Land and Property Register and the following PFOs were excluded from study population as they were not considered as a part of our target population or could not be used in our study due to the missing of relevant data: co-owners, church, commons, companies, owners younger than 15, and those without an address or living abroad. From the final population, PFOs were selected with a simple random sample. The data were obtained through an email survey. In order to maximize response rates and reduce survey error, the Dillman’s Tailored Design Method (TDM) was partly adopted. As recommended by Dillman (2007), the postal and e-mail survey involved a sequence of five contacts, two of which were used in our data collection, including a questionnaire and cover letter with a token incentive and replacement questionnaire 2–4 weeks later. The reply envelopes were enclosed by postal survey to make it easier for the respondent to return the questionnaire. In Serbia, a stratified random sample was selected from 107,790 PFOs. The criteria for PFOs classification to strata included the existence of PFOAs in the past, as well as the geographical distribution (forest territorial units and cadastral municipalities) and size of forest properties. On the basis of previously mentioned criteria, 10 municipalities were selected as territorial units in four forest areas (Severnokučajsko, Timočko, Južnokučajsko and Podrinjsko-kolubarsko forest area). All PFOs within the territorial units were divided into strata according to their property size (up to 0.99 ha; from 1 to 4.99 ha; from 5 to 9.99; from 10 to 19.99 ha; more than 20 ha). 310 PFOs were selected randomly (confidence level of 95% and confidence interval 5%) within each stratum in order to ensure that all groups are equally represented. The data were collected through personal interviews. The survey was distributed via email to 2012 PFOs in Slovenia, while in Serbia 310 PFOs were visited for Croat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

face-to-face interviewees. The total response rate for the survey in Serbia was 80% (248 replies) (Nonić et al. 2013), while in Slovenia it was 30.9% (622 replies). Taking into account the high non-response rate in Slovenia, mainly caused by errors in the national register, the results should be interpreted tentatively. The questionnaire was tested in October 2012 and the survey was carried out in the period from November 2012 to July 2013 in Serbia. The questionnaire test in Slovenia was conducted between February and March 2015 and the survey was conducted from March to May 2015. Due to confidentiality concerns, non-respondents were not followed further, so the differences among those respondents were not estimated. Representativeness of the sample was checked by inspecting spatial distribution of the respondents to test their random distribution across the country.

3.2 Data analysis Data analysis in this study was performed in two stages. The first stage involved secondary data analysis to estimate the potentials of wood mobilization from private forests in Slovenia and Serbia. The second stage involved a summary of collected data through the use of frequency distribution and selected location measures (mean). Data analysis was conducted by the SPSS 20 statistical software package.

4. Research framework As recognized by previous research, efficient PFOs organization at the local level is an important step in solving the problem of wood mobilization from smallscale forest properties (Nonić and Glavonjić 2012, Nonić et al. 2011, Pezdevšek Malovrh 2010, Glück et al. 2011, Weiss et al. 2012). There are a number of PFOs organization models that can be divided into two major groups: a) organization with a focus on management, marketing support and provision of services such as technical and financial support, as well as knowledge and information exchange, and b) organization focused on gaining political support by including PFOs in the political process, with active participation in the creation of policy frameworks for the forestry sector (Weiss et al. 2012). According to Rauch and Gronalt (2005), there are two common distinct supply chain types from smallscale forests: a) PFOs with small forest properties that »handle all forest activities including harvest planning, felling and timber haulage by themselves and sell directly to the industry or trader« or b) PFOs who Croat. j. for. eng. 38(2017)1

Š. Pezdevšek Malovrh et al.

»may be members of a Forest Owner Cooperative, an organization of forest owners that bundles harvested timber and typically sells to the wood processing industry«. Similarly, Mendes et al. (2011) and Glück (2002) consider that resolving the aforementioned issue of wood mobilization can be achieved through the potential of forest cooperatives and associations. Table 3 Basic characteristics of private forest owners Characteristics of private forest owners, %

Serbia

Slovenia

Male

94.8

65.4

Female

5.2

34.6

Age

n.a

Average 57.5

<30 years

2.8

3.9

30–60 years

69.0

54.1

>60 years

28.2

42.0

Gender

Primary occupation Farmer

36.7

Unemployed

10.1

5.7

Pensioner

22.2

49.5

Student

0.4

Employed

19.8

38.5

Other

11.3

5.9

Level of education Primary school

41.5

18.4

Secondary school

48.4

59.2

University education

10.1

22.2

The average distance from the residence to the forest property 5 km

65.3

67.9

6–20 km

27.0

18.9

21–100 km

7.7

13.3

Size of forest property

Average: 7.4

Average: 7.5

<1 ha

8.5

35.9

1–5 ha

52.0

40.1

5–10 ha

20.2

8.6

10–20 ha

11.7

5.9

>20 ha

7.7

9.5

131


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

In our study, Germany and Austria were chosen as case countries for an overview of existing PFOs organizational models, based on the fact that these countries have a long tradition in PFOs cooperation, which resulted in a high level of exploitation of forest resources. Moreover, due to similar forest sector organization models, these experiences can be applied in the analyzed case countries. According to research studies conducted in Germany, there are three different PFOs organizational models depending on wood mobilization and the wood supply chain (HAF 2008). »Model I« is an organization of PFOs who independently perform all forest management tasks. According to this model, an association fulfils the role of a coordinator among the PFOs, forest service and the wood processing industry. »Model II« is an administrative version of the association that performs the role of a coordinator between the forest owner and forest service. »Model III« is an association involved only in the coordination of PFOs, while coordination in other areas is carried out by other, larger associations or specialized marketing companies. In Austria, Rauch and Gronalt (2005) distinguished among four PFOs organizational models, depending on the knowledge related to forest management: 1) the »model of PFOs who are acting independently«, 2) the »Styrian model«, 3) the »individual accounting model« and 4) the »dividend model«. Within the first model, PFOs take the greatest share of responsibility. In the »Styrian model« owners themselves perform the activities of harvesting and transport, while the association sells assortments concludes contracts and sorts the invoices to individual owners. The »individual accounting model« is characterized by joint forest management of several owners, while the income and expenses are calculated independently for each owner. The »dividend model« involves joint forest management of all members, whereby revenues and expenses are associated with a joint account, i.e. there is no individual accounting for each plot.

5. Results 5.1 The basic characteristics of private forest owners The profiles of PFOs are presented in Table 3. The results show that PFOs in Serbia and Slovenia are mostly males (in Serbia 94.8% and in Slovenia 65.4%), aged between 30–60 (69% in Serbia and 54.1% in Slovenia), mainly with high school education (48.4% in Serbia, and 59.2% in Slovenia). The basic occupation

132

of PFOs in Serbia is farming (36.7%), while in Slovenia most of them are pensioners (49.5%). More than 60% of PFOs in both countries live close to their property (at a distance shorter than 5 km). The average size of a property in Serbia and Slovenia is almost identical (7.4 ha in Serbia and 7.5 ha in Slovenia), with a predominant share of small forest properties of up to 5 ha (60.5% in Serbia and 76.0% in Slovenia).

5.2 Private forest owner attitudes towards wood mobilization There were eight statements in the survey that measured the PFOs’ attitudes related to wood mobilization (opportunities and potential solutions leading to an increase in the level of mobilization), whereby due to the better comparability of the results, responses to the offered statements are recoded into three groups (agreement, disagreement and don’t know) (Fig. 1). It appeares that the majority of PFOs in both countries believe that a better logistics and infrastructure (better openness of forest complexes) are potential solutions leading to an increase in the level of mobilization. Moreover, PFOs in Serbia think that interest association of PFOs (42.3%) and market share and marketing (41.5%) have a decisive impact on the increase in the level of wood mobilization from private forests. PFOs stated that education and training (18.5%), greater involvement of employees in public enterprises (11.3%) and the existence of forest extension services (19.8%) do not have an impact on wood mobilization. A 52.3% of respondents in Slovenia think that interest association of PFOs and education and training (36.6%) have an important role in solving the problem of insufficient wood mobilization from private forests. Moreover, they also consider that the use of wood for biomass (36.4%), more intensive participation of the state through grants, loans and fiscal policy instruments (34.8%) and more intensive extension service offered by public forestry service (32.3%) can contribute to the solving of the problem.

5.3 Obstacles to the increase in the level of wood mobilization Approximately a half of PFOs in Serbia (50.4%) think that the level of mobilization is unacceptable with a possibility for improvement, while the situation in Slovenia is the opposite, as 51.5% of PFOs consider the level of mobilization suitable. PFOs in Serbia consider that conservation of biodiversity and protective functions of forests (52.4%), Croat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Š. Pezdevšek Malovrh et al.

Fig. 1 Attitudes of PFOs towards the opportunities and potential solutions leading to an increase in the level of mobilization. Notes: Agreement = strongly agree plus agree; disagreement = disagree plus strongly disagree; Dnk = Don’t know. The original scale and coding was done as strongly agree – 5; agree – 4; I do not know – 3; disagree – 2 and strongly disagree – 1; * Offered response for Serbia; ** offered response for Slovenia (due to different organization of forestry sector) high acquisition costs (22.6%) and unfavorable technical characteristics of the equipment (18.1%) (Table 4) are the main obstacles to an increase in the level of wood mobilization. In addition, 27% of PFOs consider that there are no obstacles related to wood mobilization from private forests. Table 4 The main obstacles to an increase in the level of wood mobilization (multiple answers) Obstacles

Serbia, %

Slovenia, %

Legislation

3.2

13.7

Lack of planning documents*

3.6

6.2

Unfavorable technical characteristics of the equipment

18.1

15.8

Biodiversity conservation and protective functions of forests

52.4

4.9

High acquisition costs

22.6

27.5

Social functions of forests

6.5

6.5

No obstacles

27.0

37.1

* For Slovenia, an offered response was »management plans«

Croat. j. for. eng. 38(2017)1

The largest number of PFOs in Slovenia considered that there were no obstacles related to wood mobilization (37.1%), stating that high acquisition costs (27.5%), unfavorable technical characteristics of the equipment (15.8%) and legislation (13.7%) are the main obstacles.

6. Proposed models of private forest owner organization The cooperation of PFOs is one of the key instruments to increase the level of wood mobilization from private forests as recognized in previous researches (i.e. Rauch and Gronalt 2005, MCPFE, DG AGRI, UNECE/FAO 2010, Becker 2010, Pezdevšek Malovrh 2010, Nonić and Glavonjić 2012, Weiss et al. 2012), whereby the choice of organizational form of cooperation depends on the identified types of PFOs (Schwarzbauer and Stern 2010, Pezdevšek Malovrh et al. 2015, Nonić et al. 2013). Analysis of wood mobilization potentials in Serbia and Slovenia showed that the harvesting intensity in private forests is below the potentials; therefore, there are preconditions to increase the level of wood mobilization. Moreover, PFOs in both countries think that

133


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Fig. 2 Activities within »Model I« there are possibilities for improving the level of wood mobilization and that interest association of PFOs can contribute to the solving of the problem. Based on the above facts and the experience from the selected case 1 countries (Austria and Germany), four models of PFOs organization have been proposed to boost the mobilization of wood from private forests. The choice of a particular model depends on the PFOs’ experience in the forests management, professional and technical capacity of PFOs, offers of service providers (silviculture, harvesting and transportation), as well as the local market of timber assortments. »Model I« is proposed for PFOs who possess skills and knowledge in the field of forestry and are full-time engaged in their forest. According to previous research, these owners have been referred to as »active« owners (Pezdevšek Malovrh et al. 2015), »the owners focused on timber production« (Kline et al. 2000, Boon and Meilby 2007), »businessmen« (Mizaraite and Mizaras 2005), »the owners with full-time work in forestry« (Wiersum et al. 2005) or »economically oriented owners« (Loenstedt 1997, Becker et al. 2000, Bieling 2004, Ingemarson et al. 2007). Timber production as the predominant management orientation is of high importance, because they generate economic revenue (Ní Dhubháin et al. 2007, Pezdevšek Malovrh et al. 2015). The income from the forest is largely involved in the total annual household income. Within this organizational model, PFO organization performs the arrangement of timber sales, measurement and qual-

1

Based on the experiences from case countries, four models of owner organization are proposed in order to more easily present activities and actors within the model, where the proposed models can also be seen as one model with four different business plans.

134

ity assessment of timber, invoicing and payment. With timber sales arrangements, the PFOs ensure contractfixed price of wood throughout the year. PFOs or members of the organization perform timber harvesting and its transportation to the wood processing industry. In addition, PFOs are involved in the measurement and quality assessment of timber. Forest road construction and maintenance is performed by a contractor (entrepreneur) hired by the organization, because of the high initial investments in the purchase of machinery (Fig. 2). The proposed model of PFO organization is similar to cooperatives. Similar results were reported by Rauch and Gronalt (2005) in the framework of the »Styrian model«, in which owners perform harvesting and supply of wood to the wood processing industry, while other activities are carried out by the employees of the association and representatives of the wood-processing industry. According to a research in Germany, in one of the organizational models, the owners within the association carry out all necessary activities, including forest road construction (HAF 2008). »Model II« is similar to »Model I« and is proposed for the same group of PFOs with the difference that the transportation of timber is performed by a contractor due to high costs and long amortization period of the transportation vehicles (Fig. 3). Within »Model II«, activities of the PFOs organization end on a roadside landing or after the measurement, quality assessment and loading of timber. The proposed model of PFOs organization is the same as in Model I (cooperatives). »Model III« is proposed for PFOs who live close to their forest property, whose main source of income is not related to forestry, which determines their valuaCroat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Š. Pezdevšek Malovrh et al.

Fig. 3 Activities within »Model II« tion of production and other forest functions. According to previous research, these owners are referred to as »multi-objective owners« (Kuuluvainen et al. 1996, Karpinnen 1998, Kline et al. 2000, Boon et al. 2004, Mizaraite and Mizaras 2005, Nonić et al. 2013, Pezdevšek Malovrh et al. 2015,) or »multi-functional« owners (Wiersum et al. 2005). Forestry neither affects the total annual household income nor has a small impact on it. They spend little time performing activities in their forests and are, therefore, without experience. A PFOs organization performs timber sales arrangement, measurement and quality assessment of timber, and invoicing and payment to owners. As in the previous models, the arrangement of timber sales ensures

a contract-fixed price of wood throughout the year. In addition, the PFOs organization performs tasks such as harvesting and transportation contracts, forest road construction and maintenance contracts, while these works are carried out by a contractor (Fig. 4). In addition, together with PFOs, the association controls all contracted activities. The »individual accounting model« in Austria and »model II« in Germany have similar characteristics. Within these models, owners do not perform work in their forest and leave these activities to private companies (Rauch and Gronalt 2005). In addition, Lutze (2010) studied the »business manager« model in which employees of the association carry out part of the tasks

Fig. 4 Activities within »Model III« Croat. j. for. eng. 38(2017)1

135


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Fig. 5 Activities within »Model IV« and perform operational management of other activities in the supply chain. »Model IV« is proposed for absent owners, including owners who live far away and have no contact with their forest property, the ones who live abroad or old owners with no descendants willing to manage their forests. These owners are referred to as »new« or »urban« forest owners (Ziegenspeck et al. 2004, Hogl et al. 2005, Schwarzbauer et al. 2010, Weiss et al. 2012), »resigning owners« (Boon et al. 2004) or »passive« (Kline et al. 2000, Pezdevšek Malovrh 2015) forest owners. The owners have no knowledge and experience in forest management and, therefore, are not interested in the management of their forest property. Moreover, none of the forest management objectives are important to them except ownership and keeping the forest in the family. These owners were created as a result of demographic change, the process of restitution or acquisition of forest ownership through the process of state or social property privatization (Nonić et al. 2013). A PFOs organization acts similarly as in the previous model, except that there is no classification of invoices to the owners individually because of joint forest management. Therefore, both invoices and payments are related to the joint account of the organization (Fig. 5). PFOs are not physically present and do not participate in any of the activities. All activities are arranged and performed exclusively by the organization. Therefore, the PFOs organization arranges harvesting and transportation contracts, forest road construction, maintenance contracts and timber sales, performs the measurement and quality assessment of timber, invoicing and payment to the owners. Measurement and grading is performed while loading on

136

a truck-road. There is no need for an individual measurement for each owner, but only of the total cargo to be shipped. The management is carried out in the total area, so that every member of the organization receives some income from forests in accordance with the volume and value of timber or their share in the total forest value. In the »dividend model« Rauch and Gronalt (2005) report similar results, according to which joint forest management is carried out and owners receive funding according to their forest property size, i.e. their share in the total managed forest complex.

7. Conclusions The study explored the potentials of wood mobilization from private forests in Serbia and Slovenia, the characteristics of PFOs, PFOs attitudes related to wood mobilization and the main obstacles. On the basis of the results, it was established that there are potentials for additional wood mobilization. It was also found that PFOs are mainly representatives of the elderly population, farmers or pensioners with a fragmented forest property. The main obstacles to the increase in the current level of wood mobilization in Serbia are the conservation of biodiversity and the protective forest function, as well as high acquisition costs, also stated as the main obstacle in Slovenia. In addition, about one third of owners in Slovenia considered that there were no obstacles to wood mobilization improvement, and it can be concluded that Slovenian owners seem to be uninterested. This may be a consequence of the 2014 ice break in Slovenia and the Croat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

resulting increase in the quantity of wood on the market, which was not substantial in Serbia. Despite the obstacles related to wood mobilization, this paper also presents potential solutions leading to an increased wood mobilization through better logistics and infrastructure, more intensive use of wood for biomass, intense participation of the state (through loans, subsidies and fiscal policy instruments) and PFOs organization. On the basis of the obtained results and experience from the case countries (Austria and Germany), four models of PFOs organization are proposed, as the cooperation of PFOs is one of the key instruments to increase the level of wood mobilization from private forests. The models take into account the characteristics and attitudes of PFOs, as well as the activities in the supply chain, including timber sales arrangements, construction and maintenance of forest roads, harvesting, measurement and quality assessment of timber, transportation and invoicing and payment. Moreover, models can also be proposed to different types of PFOs. They can be included in the production of business plans and their choice depends on their forest management goals, professional and technical capacities of the owner, local service providers’ offer and local market of timber assortments. The proposed models for active owners with fulltime work in forestry provide some security in forest management and business, as the PFOs organization arranges sales and ensures contract-fixed prices of wood throughout the year. In this way, forest owners have secured timber sales and are encouraged to increase wood mobilization. For PFOs whose main source of income is farming or another economic activity, the presented model of organization enables professionalization of a whole range of forestry services. PFOs without sufficient motivation, time or knowledge required for forestry works have the possibility to delegate forest management activities to relevant professional workers, which could result in the improvement of the existing level of wood mobilization from small forest properties. For absent owners, the benefits from PFOs organization would be mainly determined by a certain type of security reflected in the sustainable management of their forest by a professional thereby generating certain revenues. An important strategic measure for these owners would be their inclusion in professional networks and information channels, especially if one bears in mind the growing number of these owners and the tremendous potential of wood mobilization from their forests. Croat. j. for. eng. 38(2017)1

Š. Pezdevšek Malovrh et al.

In the coming period, it is first of all necessary for state institutions to stimulate wood mobilization from private forests through various (regulatory, economic and informational) policy measures, taking into account different types of forest owners.

8. References Banković, S., Medarević, M., Pantić, D., Petrović, N., 2009: The national forest inventory of the Republic of Serbia – the growing stock of the Republic of Serbia. Ministry of agriculture, forestry, and water management – Directorate for Forests, Belgrade, 244 p. Becker, G., Borchers, J., Mutz, R., 2000: Die Motive der Privatwaldbesitzer in NRW (Eigentumsverbunden und nutzungsorientiert/den meisten ist Wald mehr als Holz. Allgemeine Forst und Jagdzeitung 22: 1180–1183. Becker, G., 2010: Prospects for the market supply of wood and other forest products from areas with fragmented forest ownership structures. Case study: Saxony/Germany. AlbertLudwigs-Universität Freiburg, Institute of Forest Utilisation and Work Science, Freiburg in Breisgau, 93 p. Beurskens, L.W.M., Hekkenberg, M., 2011: Renewable energy projections as published in the national renewable energy action plans of the european member states. European Environment Agency, Energy Research Centre of the Netherlands, 244 p. Bieling, C., 2004: Non-industrial private-forest owners: Possibilities for increasing adoption of close to nature forest management. European Journal of Forest Research 123(4): 293–303. Blennow, K., Persson, E., Lindner, M., Faias, S.P., Hanewinkel, M., 2014: Forest owner motivations and attitudes towards supplying biomass for energy in Europe. Biomass and Bioenergy 67: 223–230. Bolkesjo, T.F., Baardsen, S., 2002: Roundwood supply in Norway: Micro-level analysis of self-employed forest owners. Forest Policy and Economics 4(1): 55–64. Boon, T.E., Meilby, H., Thorsen, B.J., 2004: An empirically based typology of private forest owners in Denmark: Improving communication between authorities and owners. Scandinavian Journal of Forest Research 19(4): 45–55. Boon, T.E., Meilby, H., 2007: Describing management attitudes to guide forest policy implementation. Small Scale Forestry 6(1): 79–92. Dillman, D.A., 2007: Mail and internet surveys – the tailored design method. Wiley, New York, 523 p. Directive, 2009: Directive 2009/28/EC of the European Parliament and of the Council of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and 2003/30/EC. Official Journal of the European Union L 140/16, Brussels, 47 p. EC, 2008: Mobilisation and efficient use of wood and wood residues for energy generation. Standing Forestry Commit-

137


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

tee ad hoc Working Group II on mobilisation and efficient use of wood and wood residues for energy generation. European Commission Agriculture and Rural Development, 57 p. EC, 2013: A new EU Forest Strategy: for forests and the forestbased sector, Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions COM(2013) 659 final, Brussels, 17 p. EC, 2014: EU leaders agree 2030 climate and energy goals. http://ec.europa.eu/clima/news/articles/news_2014102401_ en.htm. Accessed 8 December 2015. Glück, P., 2002: Property rights and multipurpose mountain forest management. Forest Policy and Economics 4(2): 125– 134. Glück, P., Avdibegović, M., Čabaravdić, A., Nonić, D., Petrović, N., Posavec, S., Stojanovska, M., 2010: The preconditions for the formation of private forest owners‘ interest associations in the Western Balkan Region. Forest Policy and Economics 12(4): 250–263. Glück, P., Avdibegović, M., Čabaravdić, A., Nonić, D., Petrović, N., Posavec, S., Stojanovska, M., 2011: Private forest owners in the western Balkans – ready for the formation of interest associations. European Forest Institute Research Report 25, EFI, Joensuu, 234 p. HAF, 2008: Schulungsunterlagen zur Mobilisierung der Waldbesitzer, Perspektiven für eine verbesserte Rohholzbereitstellung. Holzabsatzfonds und Pöyry Forest Industry Consulting GmbH, Bonn, 176 p. Halder, P., Paladinić, E., Stevanov, M., Orlović, S., Hokkanen, T.J., Pelkonen, P., 2014: Energy wood production from private forests – nonindustrial private forest owners‘ perceptions and attitudes in Croatia and Serbia. Renewable and Sustainable Energy Review 35: 515–526.

Leban, V., 2014: Efficiency analysis of forest owners associations in Slovenia and Germany. MSc. thesis, University of Ljubljana, Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, Ljubljana, 165 p. Lindstad, B.H., Pistorius, T., Ferrantic, F., Dominguez, G., Gorriz-Mifsude, E., Kurttila, M., Leban, V., Navarro, P., Peters, D.M., Pezdevšek Malovrh, S., Prokofieva, I., Schuck, A., Solberg, B., Viiri, H., Zadnik Stirn, L., Krč, J., 2015: Forest-based bioenergy policies in five European countries: An explorative study of interactions with national and EU policies. Biomass and Bioenergy 80: 102–113. Loenstedt, L., 1997: Non-industrial private forest owner’s decision process: A qualitative study about goals, time perspective, opportunities and alternatives. Scandinavian Journal of Forest Research 12(3): 302–310. Lutze, M., 2010: Forest owner associations as intermediaries between forest owners and the wood based industry. In: Proceedings of IUFRO Conference 3.08 Small Scale Forestry, 6.06.02 Extension, 6.06.01 Technology Transfer, Bled, Slovenia, 389–400. Mantau, U., Saal, U., Prins, K., Steierer, F., Lindner, M., Verkerk, H., Eggers, J., Leek, N., Oldenburger, J., Asikainen, A., Anttila, P. 2010: Real potential for changes in growth and use of EU forests, Final report, EUwood, Hamburg, 160 p. MCPFE, DG AGRI, UNECE/FAO, 2010: Good practice guidance on the sustainable mobilization of wood in Europe. Ministerial Conference on the Protection of Forests in Europe (MCPFE), FOREST EUROPE Liaison Unit Oslo, Norway; European Commission (EC), DG Agriculture and Rural Development, Brussels, Belgium; United Nations Economic Commission for Europe/Food and Agriculture Organization of the United Nations (UNECE/FAO), Timber Section, Geneva, Switzerland. 76 p.

Hogl, K., Pregernig, M., Weiss, G., 2005: What is new about new forest owners? A typology of private forest ownership in Austria. Small-scale Forest Economics, Management and Policy 4(3): 325–342.

Mendes, A., Štefanek, B., Feliciano, D., Mizaraite, D., Nonić, D., Kitchoukov, E., Nybakk, E., Duduman, G., Weiss, G., Nichiforel, L., Stoyanova, M., Mäkinen, P., Alves, R., Milijić, V., Sarvašová, Z., 2011: Institutional innovation in European private forestry: The emergence of forest owners‘ organizations. In: Innovation in Forestry: Territorial and Value Chain Relationships (Weiss, G., Pettenella, D., Ollongvist, P., Slee, B., eds.), CAB International, 68–86.

Huber, W., Schwarzbauer, P., Stern, T., 2013: Analyse der Motive österreichischer Kleinwaldeigentümer als Schlüssel für die Holzmobilisierung. Schweizerische Zeitschrift fur Forstwesen 164(9): 278–284.

Mizaraite, D., Mizaras, S., 2005: The formation of small-scale forestry in countries with economies in transition: Observations from Lithuania. Small-scale Forest Economics, Management and Policy 4(4): 437–450.

Ingemarson, F., Lindhagen, A., Eriksson, L., 2006: A typology of small-scale private forest owners in Sweden. Scandinavian Journal of Forest Research 21(3): 249–259.

Mori, J., Kotnik, I., Lesnik, T., 2006: Možnost sodelovanja Zavoda za gozdove Slovenije, Kmetijsko gozdarske zbornice Slovenije in Zveze lastnikov gozdov Slovenije za razvoj povezovanja lastnikov gozdov (Possible roles of the Slovenian Forest Service, the Chamber of Agriculture and Forestry of Slovenia and the Forest owners association of Slovenia in enhancing forest owners‘ associations and cooperation). Gozdarski vestnik 64(9): 476–502.

Hatemaki, L., Muys, B., Pelkonen, P., Pettenella, D., 2014: Forest bioenergy in Europe: Reassessment needed. Think Forest, European Forest Institute, Joensuu, 7 p.

Karpinnen, H., 1998: Values and objectives of non-industrial private forest owners in Finland. Silvia Fennica 32: 43–59. Kline, J.D., Alig, R.J., Johnson, R.L., 2000: Fostering the production of non-timber services among forest owners with heterogeneous objectives. Forest Science 46(2): 302–311. Kuuluvainen, J., Karppinen, H., Ovaskainen, V., 1996: Landowner objectives and non-industrial private timber supply. Forest Science 42(39): 300–309.

138

Nichiforel, L., Schanz, H., 2009: Property rights distribution and entrepreneurial rent-seeking in Romanian forestry: A perspective of private forest owners. European Journal of Forest Research 130(3): 369–381. Croat. j. for. eng. 38(2017)1


Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140) Ní Dhubháin, Á., Cobanova, R., Karppinen, H., Mizaraite, D., Ritter, E., Slee B., Wall, S., 2007: The values and objectives of private forest owners and their influence on forestry behaviour: The implications for entrepreneurship. Small-scale Forestry 6(4): 347–357. Nonić, D., Bliss, J.C., Milijić, V., Petrović, N., Avdibegović, M., Mataruga, M., 2011: Challenges of organizing private forest owners in Serbia. Small-scale forestry 10(4): 435–455. Nonić, D., Glavonjić, P., 2012: Organizing private forest owners in order to mobilize wood resources: Analysis of association models in Austria, Bavaria and Serbia (Organizovanje privatnih šumovlasnika u cilju mobilizacije drvnih resursa: analiza modela udruživanja u Austriji, Bavarskoj i Srbiji). Šumarstvo 3–4: 131–151. Nonić, D., Milijić, V., Radosavljević, A., 2010: Development of private forestry sector in Serbia and its role in NFS/NFP processes. In: Proceedings of the 11th International Symposium on Legal aspects of European Forest Sustainable development, Zvolen, Slovakia, 90–101. Nonić, D., Ranković, N., Glavonjić, P., Nedeljković, J., 2013: Typology of private forest owners in Serbia (Tipologija vlasnika privatnih šuma u Srbiji). Šumarstvo 3–4: 133–156. Nonić, D., Petrović, N., Medarević, M., Glavonjić, P., Nedeljković, J., Stevanov, M., Orlović, S., Rakonjac, Lj., Đorđević, I., Poduška, Z., Nevenić, R., 2015: Forest Land Ownership Change in Serbia. COST Action FP1201 – FACESMAP Country Report, European Forest Institute Central- East and South-East European Regional Office, Vienna, 64 p. NREAP, 2013: National renewable energy action plan of the Republic of Serbia, Ministry of Energy, Development and Environmental Protection, Belgrade, 159 p. PE »Srbijašume«, 2013: Cross section of private forests based on data from the National Inventory and Temporary annual private forest management plans for 2013. Project: Organization of private forest owners in the Western Balkan countries, with special emphasis on Serbia. Internal documentation of PE »Srbijašume«, Belgrade. Pezdevšek Malovrh, Š., Hodges, D.G., Marić, B., Avdibegović, M., 2011: Private forest owners expectations of interest associations: Comparative analysis between Slovenia and BosniaHerzegovina. Šumarski list 135(9–10): 1–10. Pezdevšek Malovrh, Š., 2010: Influence of institutions and forms of cooperation on private forest management. PhD thesis, University of Ljubljana, Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, Ljubljana. Pezdevšek Malovrh, Š., Nonić, D., Glavonjić, P., Nedeljković, J., Avdibegović, M., Krč, J., 2015: Private forest owner typologies in Slovenia and Serbia: Targeting private forest owner groups for policy implementation. Small-scale Forestry 14(4): 423–440.

Š. Pezdevšek Malovrh et al.

selected South-Eastern European countries. Biomass and Bioenergy 81: 144–153. Rauch, P., Gronalt, M., 2005: Evaluating organisational designs in the forestry wood supply chain to support forest owners’ cooperations. Small-scale Forest Economics, Management and Policy 4(1): 53–68. Restitution, 2014: Overview of the number of requests received in the Directorate for restitution. Agency for Restitution, Republic of Serbia. http://www.restitucija.gov.rs/direkcija-za-restituciju.php. Accessed 28 August 2015. Report of Public Forestry Service of Slovenia about forests for the year 2014, 2015: Zavod za gozdove Slovenije. http://www. zgs.si/slo/zavod/informacije_javnega_znacaja/letna_porocila/ index.html. Accessed 08 February 2015. Schaffner, S., 2008: Waldbesitzertypisierungen und ihre Relevanz für die Holzmobilisierung – Classifications of forest owners and their relevance for timber mobilization. Schweizerische Zeitschrift fur Forstwesen 159(12): 435–440. Schwarzbauer, P., Thoroe, M., Boglio, D., Becker, G., Stern, T., Giry, C., 2010: Prospects for the market supply of wood and other forest products from areas with fragmented forest-ownership structures. Final study report, University of Natural Resources and Applied LifeSciences, Vienna (BOKU), Confederation of European Forest Owners (CEPF), Centre TecnològicForestal De Catalunya (CTFC), Albert-Ludwigs-Universität Freiburg (ALUFR), KompetenzzentrumHolz GmbH (Wood K plus), National Professional Center for Forest Ownership – Institute for ForestryDevelopment (CNPPF-IDF), 217 p. Schwarzbauer, P., Stern T., 2010: Energy vs. material: Economic impacts of a »wood-for-energy scenario« on the forestbased sector in Austria – A simulation approach. Forest Policy and Economics 12(1): 31–38. SMARS, 2007: Relational databases from the Landowner register. Ljubljana. Suda, M., Warkotsch, W., 2002: Mit den forstwirtschaftlichen Zusammenschlüssen ins 21. Jahrhundert. AFZ/Der Wald 1: 6–9. Stern, T., Weiss, G., Bostrom, C., Huber, W., Koch, S., Schwarzbauer, P., 2013: Identifying measures for wood mobilisation from fragmented forest ownerships based on case studies from eight European Regions. Jahrbuch der Österreichischen Gesellschaft für Agrarökonomie 22(1): 19–28. Weiss, G., Gudurić, I., Wolfslehner, B., 2012: Review of forest owners‘ organizations in selected Eastern European countries. Food and Agriculture Organization of the United Nations (FAO), Rome, 57 p.

Pezdevšek Malovrh, Š., Zadnik Stirn, L., Krč, J. 2010: Influence of property and ownership conditions on willingness to cooperate. Šumarski list 134(3–4): 139–149.

Wiersum, K., Elands, B., Hoogstra, M., 2005: Small scale forest ownership across Europe: characteristics and future potential. Small-scale Forest Economics, Management and Policy 4(1): 1–19.

Posavec, S., Avdibegović, M., Bećirović, Dž., Petrović, N., Stojanovska, M., Marčeta, D., Pezdevšek Malovrh, Š., 2015: Private forest owners’ willingness to supply woody biomass in

Ziegenspeck, S., Härdter, U., Schraml, U., 2004: Lifestyles of private forest owners as an indication of social change. Forest Policy and Economics 6(5): 447–458.

Croat. j. for. eng. 38(2017)1

139


Š. Pezdevšek Malovrh et al.

Different Organizational Models of Private Forest Owners as a Possibility to ... (127–140)

Živojinović, I., Weiss, G., Lidestav, G., Feliciano, D., Hujala, T., Dobšinská, Z., Lawrence, A., Nybakk, E., Quiroga, S., Schraml, U., 2015: Forest land ownership change in Europe,

COST Action FP1201 FACESMAP Country Reports-Joint Volume, EFICEEC-EFISEE Research Report, University of Natural Resources and Life Sciences (BOKU), Vienna, 693 p.

Authors‘ addresses: Špela Pezdevšek Malovrh, PhD.* e-mail: spela.pezdevsek.malovrh@bf.uni-lj.si University of Ljubljana Biotechnical Faculty Večna pot 83 1000 Ljubljana SLOVENIA Peter Kumer, MSc. e-mail: peter.kumer@zrc-sazu.si Research Centre of the Slovenian Academy of Sciences and Arts Anton Melik Geographical Institute Gosposka ulica 13 1000 Ljubljana SLOVENIA Predrag Glavonjić, PhD. e-mail: predragglavonjic@yahoo.com Prof. Dragan Nonić, PhD. e-mail: dragan.nonic@sfb.bg.ac.rs Jelena Nedeljković, PhD. e-mail: jelena.nedeljkovic@sfb.bg.ac.rs University of Belgrade Faculty of Forestry Kneza Višeslava 1 11000 Belgrade SERBIA Bratislav Kisin, MSc. e-mail: bratislav.kisin@srbijasume.rs State Enterprise for Forest management »Srbijašume« M. Pupin Boulevard 113 11070 Belgrade SERBIA

Received: February 02, 2016 Accepted: June 30, 2016

140

Mersudin Avdibegović e-mail:mavdibegovic@gmail.com University of Sarajevo Faculty of Forestry Zagrebačka 20 71000 Sarajevo BOSNIA AND HERZEGOVINA * Corresponding author Croat. j. for. eng. 38(2017)1


Original scientific paper

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling in a Populus canadensis Moench Plantation in Beijing, China Zhongke Feng, Fei Yan, Mohammad Rahmat Ullah, Yi Dang Abstract Volume table preparation using the traditional method and a collection model requires the harvest of approximately 200–300 trees of individual species. Although high precision could be achieved using that method, it causes huge damage to the forest. To minimize these losses, in this study, a South NTS-372R total station with a precise angle and distance measurement mode was used to measure 507 trees of Populus canadensis Moench without single tree felling. Moreover, the C# programming language was used in this study and the collected volume data were inserted in the total station. Using this method, a real-time precise measurement of volume could be achieved. After data collection, the optimal binary volume model of Populus canadensis Moench could be obtained through a comparative analysis. It turns out that the Yamamoto model is the optimal binary volume model (also known as two predictor variable model), with 0.9641 as the coefficient of determination (R2) and 0.19 m3 as the standard deviation of estimated value (SEE), which presents a good imitative effect. Moreover, it showed relative stability with the general relative error (TRE) of –0.12% and the mean system error (MSE) of –1.24%. The mean predicted error (MPE) of 1.18% and the mean predicted standard error (MPSE) of 9.25% showed high estimated precision of the average and individual tree volumes. The model has only three parameters, so it is suitable for volume table preparation. Finally, this study will present some new technical methods and means for volume modeling for further application in forestry. Key words: total station, volume model, Populus canadensis

1. Introduction Forest volume is a major forest resource indicator, and its estimation is important for forest surveys around the world. Forest production is mainly guided by significant tree volume growth or decline. Volume estimation by scientific and precise methods would be a reliable basis for forest management and planning (Meng 1996). During a forest volume survey, in most countries, a tree volume table is widely used to improve the work efficiency. The most important and widely used type of tree volume is the trunk volume with bark, which was the research object of this manuscript. Three eleCroat. j. for. eng. 38(2017)1

ments have been considered for developing the regression equation for a volume table, and these are trunk volume with its DBH (diameter at breast height), tree height and stem form. Moreover, three types of volume models with one, two and more than two predictor variables are considered for volume estimation. Consequently, the tree volume tables also could be divided into one-way volume table, binary (standard) volume table and ternary (form class) volume table (Meng 1996, State Forestry Bureau of China 2013). In the case of harvested sample trees, the sectional quadrature method (measurement section method) could be adopted for volume estimation. Generally, 200–300 trees of each species will be randomly drawn

141


Z. Feng et al.

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

(Meng 1996, Lei 2005, State Forestry Bureau of China 2013 and 1978), and their ground diameter, DBH, tree height and volume of wood will be measured to develop the volume model. Based on comparison and optimization, the optimal volume model could be confirmed for volume table preparation. The precise measurement of the sample volume and the confirmation of the optimal volume model play a decisive role in its accuracy and applicability. It has been over a century since the binary volume table was developed in 1846 (Meng 1996, Zeng and Tang 2011). Forest scientists have always been engaged in the improvement of the accuracy of the estimation of volume and in the development of various models. At present, more than ten models are commonly used, and those were invented by foresters such as Meyer, Cason, Takada Co-Yan, Meng Xianyu, etc. (Feng 1997, Jin and Ding 2011, Zhong and Zhong 1999). Among these models, the Yamamoto model and the Parabola model are frequently used in China. In 1804, Cotta from Germany made a volume table for beech (Liu and Zhang 1997). In the 1970s, 35 binary volume tables of conifer species and 21 of broadleaf species in a large area were integrated and developed by the Chinese Ministry of Agriculture and Forest (Meng 1996, Zeng 2014). Making the comparison among 3682 tree samples of 14 species, Meng (1982) analyzed the accuracy of a variety of volume equations, which are usually empirical, multivariate or polynomial regression equations and are based on a large amount of experimental data. China has thousands of forest plantations. If the volume tables of all the species need revision, according to the requirements, then for every revision, approximately 300,000 standing trees need to be logged (Meng 1996, Jiao 2013a). This method of calculating the volume table establishes volume equations with relatively high precision, but it causes huge damage to forest resources. Moreover, this method is time-consuming and inconvenient, which is also inconsistent with the purpose of ecological protection. Based on this concern, Sun et al. (2013) have carried out a study on volume measurement without single tree felling and have used remarkable numbers of advanced instruments to achieve the best results for volume estimation. For that study, volume was measured by using an electronic theodolite on 202 Dahurian larch trees without any tree felling in the Wangyedian Forest Farm, Chifeng, Neimenggu Province. Compared with a traditional volume table, about 85% accuracy was achieved. In another research under a project for the establishment of the volume model in Beijing, the volume of 87 poplars, including Populus

142

tomentosa and Populus canadensis Moench, both are fast-growing poplar species, had been measured precisely without any tree felling (Jiao et al. 2013a and 2013b). Then, all the 87 poplars were logged and their real volumes measured. Compared with the real volumes, the optimal accuracy of the precision measurements without any tree felling could reach up to 95%. In October 2013, using an electronic theodolite and volume calculation software, Jiao et al. (2013a) measured the volume of 400 groups of 107 poplars. The result of its model prediction achieved by the PSOSVM algorithm is good. Thus, the function of advanced instruments for conduction measurements, such as precise angle measurement and distance measurement, could be used in the precision measurement of volume without tree felling. However, the major instruments used in these experiments are electronic theodolites that could only achieve angle and distance measurements. Therefore, there is still need for the assistance of relevant software for the volume calculation. Yan et al. (2012) conducted some experiments in Beijing and proved that the more advanced electronic total station could achieve real-time storage and computing of measurement data. This result promotes the possibility of integrating office work and field work in the precision measurement of volume without tree felling. In this research, the South NTS-372R total station, with precise angle and distance measurement mode and the volume program in the C# programming language inserted in the total station, is used in the realtime precise measurement of the volume of 507 Populus canadensis Moench in Beijing without tree felling. After data collection, the optimal binary volume model of Populus canadensis Moench could be obtained by comparative analysis, and the model provides new technical methods and tools for further research on volume table preparation.

2. Materials and methods 2.1 Description of sample tree species Populus canadensis Moench, with its oval and broad crown, taupe, rough bark, caved vertical slits, tall height as well as its large and shining blade, especially in summer, with dense leaves casting green shade, is suitable to be a roadside tree, a shade tree and a tree in shelter-belt plantations. At the same time, it is also a good choice for afforestation in industrial and mining areas and beside houses, villages, roads and water (Fig. 1). The volume table of Populus canadensis Moench was prepared in the 1970s during its large scale plantation in Beijing. State Forestry Bureau of China Croat. j. for. eng. 38(2017)1


Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Z. Feng et al.

Fig. 2 Distribution of sample collection of Populus canadensis Moench Fig. 1 Populus canadensis Moench (photographed in 2014) (1978) and Luo and Kang (2008) realized that those volume tables could not satisfy the current needs and, therefore, needed to be revised. On this backdrop, importance of this research lies in the volume table preparation of Populus canadensis Moench using scientific methods and advanced instruments with high precision and accuracy. In this research, 507 poplars were selected as sample trees, 39 from Daxing District, 25 from Fangshan District, 29 from Fengtai District, 75 from Changping District, 78 from Miyun District, 29 from Yanqing District, 76 from Huairou District, 27 from Chaoyang District, 71 from Tongzhou District and the remaining 58 from downtown and other regions. The sample distribution is shown in Fig. 2. Besides, the sampling scheme of selected trees was as follows: Þ all the selected trees were separated according to the DBH class by 2 cm. DBH of sample trees in this research ranged from 8.5 cm to 88.5 cm. In each DBH class, according to height-diameter ratio, the trees with high/middle/low ratio were chosen Þ to obtain a standard and precise volume table, the standard Populus canadensis Moench with straight trunks were selected and measured Croat. j. for. eng. 38(2017)1

Þ to carry out of this research, other relevant infor-

mation of sample areas such as place, terrain, soil, vegetation and forest situation were also investigated and recorded.

2.2 Instrument description In this study, NTS-372R Total Station, shown in Fig. 3, was used for volume measurement. It is based on the principle of phase method, shooting an extremely narrow and small industrial laser beam on the target without a prism and making use of the diffuse reflection of the target to return the signal. It is a precise

Fig. 3 South NTS-372R total station

143


Z. Feng et al.

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

instrument, by which some spatial geometric eleThe volume of treetop cone is: ments, such as the three-dimensional coordinates of 2 p hDn–1 the point, its angle and distance, could all be mea (1–3) Vtreetop cone = 12 sured. The Nanfang NTS-372R non-prism Total StaSo, the volume of a standing tree: tion is equipped with a 3R-level visible laser, 2“ angular accuracy, 2+2 ppm ranging accuracy with a prism V = V1 + V2 + ... + Vn = and 3+2 ppm ranging accuracy without a prism. Its measuring time is only 0.5 seconds (Yan et al. 2011). In 1.3p ( D02 + D0 D1.3 + D1.3 2 ) p h2 ( D1.3 2 + D1.3 D2 + D22 = + addition, due to Win CE operating system in this in 12 12 strument, the C# programming language could be 2 inserted in the system to allow for integrating 1.3p ( Doffice + D0 D1.3 + D1.3 2 ) p h2 ( D1.3 2 + D1.3 D2 + D22 ) p h D2 0 = + + ... + n n −1 (1–4) and field works for the precision measurement of vol12 12 12 ume without tree felling. Wherein, the formulas of: 2.3 Principles and methods Dn diameter of any part of the trunk height of any circular truncated cone, are as folh n 2.3.1 Principles of volume measuring without tree lows: felling The schematic diagram of volume measurement a Dn = 2 L ⋅ sin n (1–5) with no tree felling is shown in Fig. 4. The method 2 needs to take the trunk as the datum formed by the cone in the treetop and several subsequent circular (1–6) hn = L[tan(90° –rn) – tan(90° – rn–1)] truncated cones and find the sum of their volume, which is the standing tree volume (Xiong et al. 2007, (1–7) L = S · sin r1 Whitney et al. 2002). The measurement principle and calculation model is: D12 (1–1) V = V1 + V2 + V3 + ...+ Vtreetop S= 2 a sin 12 The formula for the volume of each circular trun2 (1–8) cated cone is as follows: Among them, 2 2 p h( Dtop + Dtop Dbottom + Dbottom ) L is the horizontal distance between the center of Vcircular truncated cone = (1–2) 12 total station telescope and the central point of Where: trunk, an is the horizontal angle when observing diameter of the upper cross-section Dtop the diameter of any part of the trunk Dbottom diameter of the bottom cross-section rn zenith distance of each observation H height of any circular truncated cone on the D0 ground diameter of the sample tree trunk. D1.3 DBH. Each value needs 3 precise measurements by a diameter tape and the average value to put in the program. The rest of Dn and hn can be measured directly by the total station. Then, the precise measurement of the volume without tree felling could be achieved by putting this program into the C# programming language inserted in the total station.

Fig. 4 Schematic diagram of volume measurement

144

2.3.2 Specific measurement steps 1) Based on the experience of taking measurements, the instrument requires setting at the place where the distance is 1–1.5 times longer than the height of the target tree, which can ensure a moderate observation angle of the total station and is conducive to the stability of observation. The instrument should Croat. j. for. eng. 38(2017)1


Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Z. Feng et al.

Fig. 5 Interface where the ground diameter of the tree is entered and the volume is calculated by the program be located at a place with good sight to ensure that more sample trees are observed. When measuring in mountainous areas or slopes, the general principle is to observe from the upper slope to the under slope, which is also convenient for observers to view. In regard to some inclement weather, such as strong winds (over level 4) and heavy rain, the observation should be suspended to ensure the security of the instrument and make the data reliable. 2) With centimetres as the unit of measurement, the diameter tape is utilized in the measurement of ground diameter and the DBH at 1.3 m in height. In terms of the diameter measurement, it needs more than 3 measurements, and the average value is needed for the calculation in the total station (Fig. 5). 3) From the height of 1.3 m to the top of the tree, the instrument needs to aim at the trunk edge of different segments in sequence and to obtain a1, a2, a3,..., the horizontal angles when the total station telescope aims at the left edge and the right edge separately, as well as r1, r2, r3,..., rn the zenith distance when aiming at the right edge. When the total station measures to the top of the tree, the volume could be obtained and stored in the total station just by pressing the over button (Fig. 5). According to the evaluation of tree height, in general cases, the tree could be divided into 5–10 segments; each of them approximately 0.5–3 m. This decision is based on the conclusion of experiments, which not only ensures the accuracy but also keeps the workload manageable. In each of the upward measurements, the total station needs to select the places where the trunk edge could be seen clearly and avoid scars and branches so that the observation precision could be ensured. Croat. j. for. eng. 38(2017)1

2.4 Selection of volume model From these research findings, the commonly used binary volume models are as follows (Meng 1996, Zeng 2004, Zeng 2014): In formulas in Table 1, V represents the standing volume, D is diameter, H is tree height, lg is logarithm, and a1, a2, a3, a4, and a5 are model parameters. GAOFS represents the Germany Academy of Forestry Sciences. Therefore, this research takes the volume of 507 Populus canadensis Moench trees as the dependent variable by using 16 commonly used binary volume models and adopting the Marquarat algorithm and the tree height and diameter as the independent variables to fit the volume model. Marquardt algorithm is the most widely used optimization algorithm. It outperforms simple gradient descent and other conjugate gradient methods in a wide variety of problems (Zeng 2014). As a result, the optimal binary volume model, which has small and stable parameters, is selected in this study for its simplicity and effective structures.

2.5 The evaluation of the model Binary volume models (shown in Tab. 1) are used in this study to develop several regression models. There are many indexes to evaluate models, but under the considerations of all the factors, when evaluating and comparing different volume models, the following six are treated as the basic indexes, namely R² is the coefficient of determination, SEE is the standard deviation of the estimated value (Kozak and Kozak 2003, Parresol 1999, Zabek 2006, Zianis and Mencuccini 2004, Harmel and Smith 2007), TRE is the overall

145


Z. Feng et al.

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Table 1 Commonly used volume models Binary Volume Models

Inventor

Number

V = a0 + a1 D + a2 D 2 + a3 DH + a4 D 2 H + a5 H

Meyer W.H.

1–1

V = a0 + a2 D 2 + a2 D 2 H + a3 H + a4 DH 2

Meng Xianyu

1–2

V = a0 + a2 D 2 + a2 D 2 H + a3 H 2 + a4 DH 2

Näslund M.

1–3

Terasaki Wata

1–4

V = a0 D a1 H a2

Yamamoto

1–5

V = a0 (D+ 1)a2 H a3

Korsun F.

1–6

V = a0 + a1 D 2 H

Spurr S.H.

1-7

V = D 2 ( a0 + a1 H )

Ogeya N.

1–8

Takada Kazuhiko

1–9

V = a0 D a1 H 3− a2

Dwight T.W.

1–10

V = a0 ( DH )a1

Spurr S.H.

1–11

V = a0 + a1 D 2 + a2 D 2 H + a3 H

Stoate T.N.

1–12

V = a0 D 2 H

Spurr S.H.

1–13

Terasaki Wata

1–14

lg V = a0 + a1 lg D + a2 (lg D)2 + a3 lg H + a4 (lg H )2

GAOFS

1–15

V = a0 D 2 H + a1 D 3 H + a2 D 2 lg D

Zhao Kesheng

1–16

V = a0 D a1 e

V=

a2 H −

a3 H

D2 H a0 + a1 D

V = a0 D 2 e

a1 −

a2 H

relative error, MSE is the mean system error, MPE is the mean predicted error, and MPSE is the mean percentage standard error (Zeng and Tang 2011, Zeng et al. 1999, Tang and Li 2002). The expression of these indexes is as follows:

2

R = 1−

SEE =

TRE =

∑ ( yi − y1 )2 ∑(y

∑(y

− y)

− y1 )2

∑ ( y − y ) ∑ y × 100

(2–3)

i

1

MSE =

MPE = ta ×

(2–2)

n− p

(2–1)

i

1

146

i

∑(y

i

− y1 )

y1 n × 100

SEE × n × 100 y

MPSE = ∑ ( yi − y1 ) / y1 / n × 100

(2–4) (2–5) (2–6)

Where: actual observed value yi y1 predicted value of model mean value of the sample y n number of samples p number of model parameters tα t value when the significance level is 0.05. Croat. j. for. eng. 38(2017)1


Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Z. Feng et al.

Table 2 Actual measurement data of the volume of Populus canadensis Moench Tree species Populus canadensis Moench

Sample

507

Variable

Mean

Minimum

Maximum

Standard deviation

Coefficient of variation, %

DBH, cm

38.2

8.5

88.5

12.9

33.77

Tree height, m

23.8

10.3

35.0

4.6

19.33

Volume, m3

1.4022

0.0385

6.1957

0.9999

71.31

In these six indicators, R2 and SEE reflect the goodness of fit of this model. However, these are the most common indicators used to assess regression models. TRE and MSE are also important indicators to show the fitting effect, and the value of these two indicators is normally less than 3% and 5%, respectively. This value is close to zero, which reflects its effectiveness for volume estimation. The MPE reflects the estimated precision of the average volume, and the MPSE reflects the estimated precision of a single tree volume.

Based on the 16 commonly used binary volume models mentioned in 1.4, this research takes the actual measurement data of all the 507 Populus canadensis Moench as the sample data and adopts the Marquarat algorithm to conduct model fitting. During this process, the 1st Opt15PRO software and formulas are used for calculating various statistical indicators (2–1) ~ (2–6). The results are shown in Tables 3 and 4 and Fig. 6. From the indicators in Tables 3 and 4, and using 16 binary volume models, the value of R2 is always over 0.96, which shows that the DBH and tree height have already accounted for 96% of the variation, reflecting the high values of the precision of the average and individual tree volumes. Among them, the 1–1 Mayer model and the 1–15 GAOFS model enjoy the highest correlation. SEE takes m³ as the basic unit and retains two significant figures. Then, except for 1–11 Spurr’s

3. Results and discussion The measurement results of these 507 Populus canadensis Moench, such as DBH, tree height and volume, are imported into the computer. After statistics, the actual measurement data are shown in Table 2.

Table 3 Fitting results of the binary volume model of Populus canadensis Moench Model

Model parameter a0

a1

a2

a3

a4

a5

1–1

–0.161410

0.016783

–0.000177

–0.000285

0.000036

0.00186

1–2

–0.009439

0.000179

0.000024

–0.000626

0.000005

1–3

–0.001743

0.000183

0.000024

0.000098

0.000007

1–4

0.000674

1.947611

0.025715

5.480645

1–5

0.000061

1.941859

0.885561

1–6

0.000050

1.981313

0.889570

1–7

0.047941

0.000033

1–8

0.000130

0.000028

1–9

27719.743170

43.335726

1–10

0.000061

1.941859

2.114439

1–11

0.000013

1.672499

1–12

–0.075889

0.000141

0.000027

0.004805

1–13

0.000033

1–14

3.052688

–7.327658

21.540754

1–15

–3.442538

2.528834

–0.169925

–0.908352

1–16

0.000032

0.000000

0.000063

Croat. j. for. eng. 38(2017)1

147


Z. Feng et al.

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Table 4 Statistical results of the binary volume model of Populus canadensis Moench Model

Statistical indicators R2

SEE/m3

TRE, %

MSE, %

MPE, %

MPSE, %

1–1

0.9643

0.19

0.00

1.33

1.18

11.52

1–2

0.9641

0.19

0.00

–0.38

1.18

9.06

1–3

0.9641

0.19

0.00

–0.53

1.18

9.04

1–4

0.9642

0.19

–0.16

–1.65

1.18

9.53

1–5

0.9641

0.19

–0.12

–1.24

1.18

9.25

1–6

0.9641

0.19

–0.17

–1.68

1.18

9.49

1–7

0.9635

0.19

0.00

–2.42

1.19

10.76

1–8

0.9638

0.19

0.50

0.14

1.18

9.03

1–9

0.9637

0.19

0.24

0.31

1.18

8.96

1–10

0.9641

0.19

–0.12

–1.24

1.18

9.25

1–11

0.9364

0.25

0.42

1.74

1.57

12.32

1–12

0.9640

0.19

0.00

–0.12

1.18

9.58

1–13

0.9635

0.19

1.19

2.59

1.20

9.18

1–14

0.9633

0.19

0.77

2.33

1.20

9.49

1–15

0.9643

0.19

0.00

–0.71

1.18

9.13

1–16

0.9642

0.19

–0.02

–0.56

1.18

8.95

model SEE value of 0.25 m³, the standard deviation of the rest of the models estimated values are all 0.19 m³, which is attributed to the basic information of calculating the predicted error. The total relative error (TRE) and the mean systematic error (MSE) range from –0.17% to 1.19% and from –2.42% to 2.59%, respectively. Both of them are under control within ±3%, proving that they have achieved an ideal model fitting effect (Zeng and Tang 2011). The mean predicted error (MPE) ranges from 1.18% to 1.57%, indicating that the average predicted precision of the volume reaches over 98%. The majority of the mean predicted standard error (MPSE) is lower than 10%, and only the 1–1 Mayer model, the 1–7 and 1–11 Spurr models are a little bit higher, reaching 10.76%, 11.52% and 12.32%, respectively. All the indicators reflect the average level of estimated error of one tree volume. By comparison, the binary volume model of the 1–15 GAOFS model is titled the optimal model. However, in addition to the optimal precision, the optimal model should also have fewer and more stable parameters with a simple and effective structure. Therefore, the commonly used binary volume models that have fewer than 3 parameters are 1–5, 1–6, 1–7, 1–8, 1–9, 1–10, 1–11, 1–13, 1–14 and 1–16. With further statistical analysis, it is found that the total relative error (TRE) and the mean system error (MSE) of the models with 3 parameters,

148

namely the Yamamoto model (1–5) and the Dwight T.W. equation (1–10) are, –0.12% and –1.24% respectively, while the first model bears a simple structure. The volume models with 2 parameters also perform well and meet the regulatory requirements, but their precision is

Fig. 6 Distribution of residuals in Yamamoto model Croat. j. for. eng. 38(2017)1


Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

slightly lower than that of the Yamamoto model (1–5) and the Dwight T.W. equation (1–10). The total relative error (TRE) and the mean system error (MSE) of the 1–13 models with one parameter are 1.19% and 2.59%, respectively, and meet the regulatory requirements. In summary, the 1–5 Yamamoto model, the most commonly used model in China, is suitable for selection as the optimal binary volume model (3–1). Moreover, this model is to be fitted properly considering its convenient use with few parameters. The Yamamoto model is also known as the standard binary volume model in China (Zeng 2014). The standard (Yamamoto) binary volume model:

V = 0.00006107D1.942H0.8856

(3–1)

The distribution of residuals of the 507 sample trees volume was calculated by determining where the actual observed value was the predicted value of the model (Fig. 6). It was concluded that the positive and negative residuals in each DBH class was almost equal, which shows that the residuals from Yamamoto model was of good randomness and normal distribution.

4. Conclusion

Z. Feng et al.

(Project No. 2015ZCQ-LX-01) and the Beijing municipal key laboratory of precision forestry science and technology innovation base 2015 cultivation and development of the special project (No. Z15110000161596). We are grateful to the undergraduate students and staff of the Laboratory of Forest Management and »3S« Technology and Beijing Forestry University.

5. References Feng, S.Z., 1997: Examination and study for volume equation in the production of forestry. Journal of Harbin University of Science and Technology 1: 72–75. Harmel, R.D., Smith, P.K., 2007: Consideration of measurement uncertainty in the evaluation of goodness-of-fit in hydrologic and water quality modelling. Journal of Hydrology 337(3): 326–336. Jiao, Y.Q., Feng, Z.K., Zhao, L.X., Xu, W.H., Cao, Z., 2013a: Research on living tree volume forecast based on PSO embedding SVM. Spectroscopy and Spectral Analysis 34(1): 175–179. Jiao, Y.Q., Zhao, L.X., Deng, O., Xu, W.H., Feng, Z.K., 2013b: Calculation of live tree timber volume based on particle swarm optimization and support vector regression. Transactions of the Chinese Society of Agricultural Engineering 29(20): 160–167.

Based on this study and comparison of volume data, an optimal binary volume model of Populus canadensis Moench has been finally developed. As shown in the results, the Yamamoto model has proper fit, good stability, and higher precision of both the MPE of volume and the MPSE of each volume. In addition, the number of parameters is only 3. Considering all of these indicators, this model is found to be the most suitable for volume table preparation. However, the measurement of data is easy due to the straight trunks with few branches of Populus canadensis Moench. Considering coniferous species, a good sight situation could be obtained when measuring the tree volume and more precise result could be measured. On the other hand, for broadleaf trees, it is still worth further study to determine whether this method could achieve the same positive effect. This research makes use of an advanced instrument which is easy to handle and operate for volume estimation. Most importantly, the application of this method could reduce forest volume losses due to tree felling, as well as provide new techniques for the development of volume models and volume tables. Therefore, this method should be applied and further developed.

Parresol, B.R., 1999: Assessing tree and stand biomass: a review with examples and critical comparisons. For Sci 45(4): 573–593.

Acknowledgments

State Forestry Bureau of China, 1978: Tree volume table (LY 208–77), 13–15 p.

Financial support for this study was provided by Key Technology and Equipment of Precision Forestry

State Forestry Bureau of China, 2013: Technical procedures of volume table (LY/T 2013–2013), 4–7 p.

Croat. j. for. eng. 38(2017)1

Jin, M., Ding, G.J., 2011: Two-dimensional yield-rate tables for merchantable volumes of Pinus massoniana in Guizhou province. Journal of Zhejiang A&F University 4: 576–582. Kozak, A., Kozak, R., 2003: Does cross validation provide additional information in the evaluation of regression models? Can J For Res 33(6): 976–987. Liu, Z.Y., Zhang, Z.Y., 1997: The basic principles of the Bulgarian beech forest classification. Acta Agricultural Universitatis Jiangxiensis 6: 180–182. Lei, J.F., 2005: Forest resources of China. Chinese Forestry, Beijing, 172–173. Luo, X.X., Kang, X.G., 2008: Progress in research on the comprehensive monitoring of forest resources. Journal of Zhejiang Forestry College 25(6): 803–809. Meng, X.Y., 1982: Comparisons of binary equation of wood volume. Journal of Nanjing Forestry University: Natural Sciences Edition 4: 80–87. Meng, X.Y., 1996: Forest measurements. Second edition, China Forestry Publishing House, Beijing, 148–151 p.

149


Z. Feng et al.

Developing a Volume Model Using South NTS-372R Total Station without Tree Felling ... (127–150)

Sun, M.Y., Feng, Z.K., Yan, F., Fan, Z.M., Zheng, J., 2013: Principle and application of automatic accurate measurement of stem volume by using electronic theodolite. Forest Resources Management 05: 80–84. Tang, S.Z., Li, Y., 2002: The Basics of Statistics for biological mathematical model. Science Press Beijing, 15–20 p. Whitney, J.D., Tumbo, S.D., Miller, W.M., Wheaton, T.A., 2002: Comparison between ultrasonic and manual measurements of citrus tree canopies. In Proceedings of the ASAE Annual International Meeting, St. Joseph, MI, USA (Vol. 2831). Xiong, N.N., Wang, J., Luo, X., Feng, Z.K., Han, G.S., 2007: Measuring tree crown volume based on three-dimension laser scanning and mapping system taking Pinus tabulaeformis as a case. Journal of Beijing Forestry University 29(2): 61–65. Yan, F., Guo, W.W., Feng, Z.K., Cao, M.L., Wu, L.L., 2011: Study on Automation Technology and Method of Forest Measurement in Fixed Sample based on Total Station and PDA. Key Engineering Materials, Materials, Mechatronics and Automation 467: 1852–1857. Yan, F., Ullah, M.R., Gong, Y.X., Feng, Z.K., Chowdury, Y., Wu, L.L., 2012: Use of a no prism total station for field mea-

surements in Pinus tabulaeformis Carr. stands in China. Journal of Biosystems Engineering 113(3): 259–265. Zabek, L.M., 2006: Prescott C E. Biomass equations and carbon content of aboveground leafless biomass of hybrid poplar in Coastal British Columbia. Forest Ecology and Management 223(1): 291–302. Zeng, W.S., Luo, Q.B., He, D.B., 1999: Research on weighting regression and modelling. Scientia Silvae Sinicae 35(5): 5–11. Zeng, W.S., 2004: On the Methodology for Establishing One– way Tree Volume Models. Forest resources management 2(1): 21–23. Zeng, W.S., Tang, S.Z., 2011: Goodness Evaluation and Precision Analysis of Tree Biomass Equations. Scientia Silvae Sinicae 47(11): 106–112. Zeng, W.S., 2014: Establishment of Compatible Tree Volume Equation Systems of Chinese Fir. Forest research 27(1): 6–10. Zhong, Y.S., Zhong, Y.Z., 1999: Parameters estimating of volume equations by least method squares of relative error. Journal of Northwest Forestry University 14: 84–87. Zianis, D., Mencuccini, M., 2004: On simplifying allometric analyses of forest biomass. Forest Ecology and Management 187(2): 311–332.

Author’s address: Prof. Zhongke Feng, PhD. e-mail: 15210026573@163.com Fei Yan, PhD. * e-mail: yf-perfect@163.com Beijing Forestry University Forestry College Precision Forestry Key Laboratory of Beijing No.35 Tsinghua East Road Haidian District,Beijing CHINA

Received: March 13, 2016. Accepted: May 16, 2016.

150

Mohammad Rahmat Ullah e-mail: r_ullah101@yahoo.com Yi Dang e-mail: 275027301@qq.com Beijing Forestry University College of Forestry No.35 Tsinghua East Road Haidian District,Beijing CHINA *Corresponding author Croat. j. for. eng. 38(2017)1


ISSN 1845-5719


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.