Performance of a whole tree mechanised timber harvesting system when clear-felling a 32-year-old Pinus taeda L. stand

Background: Work studies are fundamental for the development and assessment of timber harvesting systems aimed at rationalising and improving forest management activities. Methods: This study evaluated the operational performance of a mechanised whole-tree harvesting system in 32-yearold Pinus taeda L. stands producing multiple timber products. A time and motion study at the cycle element level was conducted to evaluate the operational performance of each component of the harvesting system. Equations were developed to estimate the productivity of tree extraction activity with a wheeled skidder and log loading with a mechanical loader. Results: Tree felling with an excavator-based harvester had the highest mean productivity (135 m3 per productive machine hour), followed by tree extraction with a wheeled skidder (117 m3 per productive machine hour), while manually processing larger logs with a chainsaw had the lowest productivity (25.7 m3 per productive machine hour). Operator, extraction distance and mean log volume had a significant effect on the performance of different activities and were included in productivity models. Conclusions: Operational performance of equipment was variable and dependent on the effect of the operator, extraction distance and log volume. Thus, the use of models to estimate productivity considering such factors, coupled with reduced delays to increase utilisation of equipment, will contribute to the better management and planning of forest harvesting operations under the evaluated conditions. New Zealand Journal of Forestry Science De Oliveira Pitz et al. New Zealand Journal of Forestry Science (2021) 51:12 https://doi.org/10.33494/nzjfs512021x96x


Introduction
The management of planted pine forests is a consolidated activity in Brazil, both by vertically integrated companies and independent producers. However, the forest management strategies adopted by vertically integrated companies differ from those adopted by independent producers, who typically aim to diversify forest production to market logs for different industrial segments. When the objective of forest production is Keywords: forest operations and techniques; work study; forest mechanisation. compared to thinning operations. One of the key factors driving this is better access and mobility for machinery traffic due to the relatively small number of trees per unit area remaining at the final cutting. Because these trees have larger dimensions and volumes, this results in lower specific time consumption and higher productivity of harvesting equipment (Ghaffariyan et al. 2012;Strandgard et al. 2013;Walsh & Strandgard 2014).
However, the large size of the trees can lead to increased safety risks, which implies the use of specific techniques when performing cutting and extraction activities. In addition, a wide variety of log assortments are produced, which increases the complexity of operational aspects of pre-extraction, stacking and organisation of timber.
In Brazil, these forest operations are carried out predominantly using cut-to-length (CTL) or whole-tree (WT) harvesting systems (Seixas & Oliveira Júnior 2001). In most cases the typical WT harvesting systems consist of a feller-buncher, skidders and processors, where only one machine perform all the tree bucking and processing (Rocha et al. 2009;Lopes et al. 2017;Diniz et al. 2018a;Rodrigues et al. 2019). However, there a very few studies analysing this system in Brazilian pine plantations managed on longer rotations, (Pereira et al. 2015;Souza et al. 2018), especially when machinery configurations differ from the typical WT system.
The evaluation of timber harvesting systems is essential for correcting and changing the production process to rationalise and optimise resources (Magagnotti & Spinelli 2012;Ackerman et al. 2014;Szewczyk et al. 2017). It is also an indispensable instrument for comparing different methods or equipment (Spinelli et al. 2014;Marčeta & Košir 2016;Pajkoš et al. 2018).
Our study aimed to: (i) evaluate the operational performance of a mechanised whole tree harvesting system in the final cutting of Pinus taeda stands; (ii) verify the effect of operational factors on specific time consumption and productivity; and (iii) model the relationship between productivity and operational factors to provide information to improve management of these activities.

Study site and stand characteristics
The study was conducted in a commercial forest stand in Capão Alto, Santa Catarina State, Brazil. The terrain slope was level to gentle according to Forestry Commission UK (1996) (Level=0°-6°, Gentle= 6°-11°, Moderate=11°-18°, Steep=18°-27°, Very Steep=>27°) and the climate is classified as Cfb according to Köppen-Geiger with no defined dry season, and mild summers (Peel et al. 2007). The annual mean temperature ranges from 14 to 16°C and the annual precipitation is between 1600 to 1900 mm (Alvares et al. 2013). The forest stand consisted of Pinus taeda and its purpose was to produce wood for multiple uses, so it was subjected to four thinning interventions. Our study was performed when the stand was undergoing the final felling, at the age of 32 years, with a stand density of 357 trees/ha, mean diameter at breast height of 45 cm, mean total height of 31 m and a mean individual tree volume of 2.46 m 3 .

Harvesting operations
We evaluated a mechanised "whole tree" harvesting system configured to produce different demands of log assortments for different destinations. The system consisted of an excavator-based harvester (CAT FM 320D) coupled to a 7000XT Logmax head which felled the trees and a wheeled grapple skidder (John Deere 748H), which extracted the trees from the cutting area to the roadside landing, with an extraction distance ranging from 30 to 310 m.
At the roadside landing area, the trees were bucked and processed in three stages by different equipment. The first logs cut from the trees (large logs) were destined for sawmills and lamination plants, and had volumes ranging from 0.232 to 0.870 m 3 log -1 , smallend diameters ranging from 35 to 70 cm and, often had an irregular shape at the base. These were manually processed using a chainsaw (Stihl MS 361) due to the limitations of other cutting equipment.
The intermediate volume logs (medium logs), destined for sawmills with volumes between 0.157 to 0.227 m 3 log -1 and diameter at the smaller end ranging from 25 to 35 cm, were processed using a mechanical slasher coupled to a Caterpillar 320B. The lower volume logs (small logs), destined for pulp and mechanical processing with volumes between 0.087 to 0.132 m 3 log -1 and small-end diameters ranging from 8 to 25 cm, were processed by the same excavator-based harvester that was used for tree felling, but at a later point in time.
The logs were stacked into product piles and organised in seven different log assortments according to small-end diameter and presence/absence of knots with lengths ranging from 1.90 to 3.10 m. After a period of between two to five days, the logs were loaded onto transport vehicles with a mechanical crawler loader (Caterpillar 320B). The work schedule and utilisation of each piece of equipment within the harvesting system depended on commercial production needs and operational work restrictions. Wood residues were not taken back into the stand.

Performance evaluation
The operational performance of the activities was assessed by time and motion study at the cycle element level following the modelling approach (Magagnotti & Spinelli 2012). The work cycle of each piece of equipment was divided into elements (Table 1) and then the time consumption was measured by the individual time clocking technique using a centesimal chronometer and specific forms. The number of trees felled, extracted or bucked at each working cycle was recorded. The volume produced at each working cycle (in cubic meters of solid wood over bark) was determined by multiplying the number of trees (or logs) by the mean individual tree (or log) volume for the stand. Data on the volumes of individual trees and logs for the stand were obtained from the forestry company's inventory records. Bunking the logs in the trailer or semi-trailer of secondary transport vehicle Data referring to the operational performance factors were also measured for each working cycle. The operator (Op) was considered a fixed-effect factor and different operators were only evaluated for the wheeled skidder and the mechanical loader. The slope (in degrees) was assessed with a TruPulse 360 Laser Rangefinder. The extraction distance (ED, in meters) for a wheeled skidder, which corresponded to the distance between where the skidder stopped to load trees and then stopped to unload trees, was measured with the same device also used to assess the slope. The mean log volume (LV, in m 3 log -1 ) for the mechanical loader was calculated by dividing the total loaded volume in a cycle by the number of logs loaded in the same cycle.
The specific time consumption (s m -3 ) was calculated as the ratio between the time consumed for each element and the production in the respective work cycle. The productivity per productive machine hour without any delays (P PMH , m 3 PMH 0 -1 ) was calculated as the ratio between the production in the work cycle and the total time consumed in the respective cycle (excluding delays). Delay times were recorded and classified according to the IUFRO time model (Björheden et al. 1995) so that the availability (AR) and utilisation rate (UR) could be calculated according to Ackerman et al. (2014).

Data analysis
Specific time consumption and P PMH data were analysed by descriptive statistics and expressed as box and whisker plots. The estimation error for the P PMH variable was determined at the 95% level of probability significance, according to Szewczyk et al. (2017). The effect of influential factors for some activities was analysed using ANOVA. Prior to analysis, the data were subjected to a Kolmogorov-Smirnov normality test at 5% significance level and, in the case of non-normality, were mathematically transformed to achieve normality.
For tree extraction activity with the wheeled skidder and log loading with the mechanical loader, multiple linear regression equations were fitted using a stepwise approach to test the effect of different independent variables on P PMH . Goodness of fit for the models was evaluated by the adjusted coefficient of determination and absolute and relative standard errors of estimates.

Mean values, estimation errors, and ratios of performance measures
Among the activities and equipment evaluated, the highest estimation error (Ԑ) was found for processing medium logs with the mechanical slasher (Ԑ = 9.80%), followed by tree extraction with a wheeled skidder (Ԑ = 8.88%) ( Table 2). The activity of processing large logs with the manual chainsaw had the highest AR but the highest mean T cycle , lowest mean P PMH and UR ( Table  2). The highest operational performance was observed in the activity of tree felling with the excavator-based harvester (T cycle = 32.4 s m -3 and P PMH = 135 m 3 PMH 0 -1 ), although this had the lowest AR (66.2%), and for the tree extraction with the wheeled skidder (T cycle = 45.8 s m -3 and P PMH = 117 m 3 PMH 0 -1 ), which had the highest UR (61.0%).

Equipment
Function/ activity  Mean values (± standard deviation) for total time taken per work cycle, volume per cycle, productivity per productive machine hour, estimation error, availability and utilisation rate for each piece of equipment and function/activity of the harvesting system.

Time consumption and effect of factors on performance
For tree felling with an excavator-based harvester, BF was the element that consumed most time within the work cycle (Figure 1a). There was also a significant effect of ground slope on the time consumed in this element (Table 3). For tree extraction with a wheeled skidder (Figure 1b), TL and TE were the elements that consumed most time in the work cycle; they varied significantly between machine operators as indicated by ANOVA (Table 3). Significant differences between machine operators were also observed on all other variables related to operational performance, except UM (Table  4). On average, Operator 2 took more time and extracted 10.2% less volume per work cycle than Operator 1, resulting in a 27.6% mean productivity difference (mean P PMH of 146 m 3 PMH 0 -1 for Operator 1 compared with 82.8 m 3 PMH 0 -1 for Operator 2). Ground slope had a significant effect on the TL, UM and V cycle elements, which was due to the increased difficulty of working on steeper slopes. However, there was no significant effect of slope on T cycle and P PMH . Extraction distance had a significant effect on all variables assessed for the skidder operation, except V cycle . It was also the single explanatory variable in the models for wheeled loader productivity (Table 4). Even though there was a significant difference in performance between the two operators, longer extraction distances resulted in more time being consumed which consequently reduced productivity (Figure 2a).
Among the tree processing activities, more time per work cycle was consumed using a manual chainsaw (Figure 1c) compared with a mechanical slasher and excavator-based harvester (Figure 1d and 1e,  respectively). Most of the time consumed in the manual chainsaw work cycle occurred at the TB element due to the large size of the logs and, as already mentioned, because the activity was performed with manual equipment. AO was the element that consumed most of the work cycle time for processing medium logs with the mechanical slasher (Figure 1d). In the case of processing small logs with the excavator-based harvester, most of the time consumed in the work cycle was with the PR element.
There was a significant difference in the work cycle times between the two operators ( Figure 1f, Table 3), but there was also a significant effect of mean log volume. In general, the BAT element was responsible for most of the time taken during the work cycle (Figure 1f), which was due to the need to optimise the space occupied by the load on trucks.

Discussion
Analysis of the operational performance of harvesting system equipment Estimation errors (Table 2) were due to variability of the operational performance of activities that, in turn, varied depending on interactions with factors such as the mean volume per tree, type of log assortment produced, extraction distance, slope, operator and among others.
However, values of Ԑ did not exceed 10% for any of the activities.
Under conditions of lower mean tree individual volume, Pereira et al. (2015) reported slightly higher values of P PMH for tree felling with a tracked feller buncher than those observed in the present study and lower values for extraction activity with wheeled skidder. The operational performance for manually processing large logs with a chainsaw (Table 2) was similar to that found by Leite et al. (2014), although the latter study was conducted in eucalyptus plantations with lower mean individual-tree volume.  The P PMH of processing medium size logs with the mechanical slasher was higher than the value reported by Conrad IV and Dahlen (2019) and the P PMH of processing small logs with the excavator-based harvester was also higher compared with that reported by Ghaffariyan et al. (2012) andScorupski et al. (2017), however, the operational conditions of these studies were different. The values of P PMH for log loading with a mechanical loader reported by Ghaffariyan et al. (2012) were higher compared with present study, although in conditions of higher mean log volume.
It should be noted that in most published studies, one piece of equipment performs all the tree bucking and processing operations in the WT harvesting system, which differs from the system studied here. This characteristic leads to a higher probability of occurrence of production bottlenecks, requiring attention in operational management to avoid this. In the current study, although the harvester had the highest P PMH , several delays occurred mainly due to corrective maintenance of the harvester head, resulting in the lowest AR value and, consequently, a relatively low UR. The wheeled skidder AR value was relatively high, and its UR was the highest, with delays due to auxiliary activities. In moments of "excess production time", equipment performed other functions (support or production at another stage). Therefore, better mechanical maintenance practices and use of the equipment according to the limits of technical capacity, can improve the excavator-based harvester availability and increase the overall system production.
Low operating performance was expected for manual processing with a chainsaw due to this being the only non-mechanised activity within the harvesting system. The need to wait for the trees on the roadside landing area to be organised at the end of each extraction cycle of the wheeled skidder caused most of the delays. Hence, greater attention to operational management is required so that this activity does not become the bottleneck of the production system, especially because it is more susceptible to adverse weather conditions and, thus, subject to the risk of accidents and low UR (Shrestha et al. 2005;Silayo & Migunga 2014;Fulvio et al. 2017), as observed in present study.
The AR of the mechanical slasher and excavatorbased harvester processing medium and small logs, respectively, was relatively high. The delays due to rework and organisation logs in product stacks at roadside for subsequent loading reduced considerably the equipment UR. The mechanical loader had the lowest UR, which was due to delays caused by waiting for transport vehicles or displacement between log piles or roadside landing areas.
It is important to highlight that the ratios reported may not reflect the real proportion of availability and utilisation due to this study being short-term. Longterm studies are recommended for more accurately determining the usage ratios, as well as for estimating delays (Spinelli & Visser 2008;Magagnotti & Spinelli 2012).

Factors affecting performance and modelling of productivity
Terrain slope had a significant effect on some elements of the tree felling work cycle with the excavator-based harvester and extraction with the wheeled skidder (Table 3). It is expected that the increase in terrain slope increases the degree of work difficulty and, consequently, the operational safety risks. However, this factor had no significant effect on T cycle and P PMH , which possibly occurred because the maximum inclination observed in this study was only 9 degrees and, therefore, did not impose any major restrictions on equipment mobility. In clearcutting of a Pinus plantation with a lower mean individual tree volume, Diniz et al. (2018b) reported that the wheeler skidder performance only tended to be negatively affected when the slope was above 26 degrees. On lesser slopes, the operator was able to compensate for increased cycle times on steeper areas by working more quickly on the flatter areas, thus avoiding any productivity reduction. The operator had a significant effect on most of the operational performance variables of the wheeled skidder and the mechanical loader (Table 3) and, therefore, regression equations were fitted to individual operators ( Table  4). For both types of equipment, the P PMH was greater for the more experienced operator (Op. 1 in Figure  2a,b), which suggests that it is important to invest in people development and training in order to improve performance in forest operations.
The predictability and effect of extraction distance on operational performance of the wheeled skidder is widely reported in the scientific literature for various equipment and operational conditions (Behjou et al. 2008;Rocha et al. 2009;Ghaffariyan et al. 2012;Walsh & Strandgard 2014;Strandgard et al. 2017). In the case of the mechanical loader, log volume had a significant effect on most of the operational performance variables (Table 3) and, thus, was included as predictive factor for estimating the P PMH in all regression equations (Table  4). There was a tendency to increase the P PMH as the log volume increases (Figure 2b), similar to observations made by Diniz et al. (2018c) for other operational conditions.

Conclusions
The operational performance of the equipment in the harvesting system studied was variable and dependent on the effect of the operator, extraction distance and log volume. For this reason and because it has more equipment and a greater number of processing stages than most of the whole-tree systems that have been studies, there is greater likelihood of production bottlenecks, requiring attention in operational management to avoid or minimise this.
The use of models to estimate productivity considering such mentioned factors and reduced delays to increase availability and utilisation of equipment will contribute to the better management and planning of forest harvesting operations under the evaluated conditions.