^{1}

^{*}

^{2}

^{3}

^{1}

^{1}

^{3}

The importance of continuous cover forestry (CCF) is increasing, yet there is lack of data and understanding about many aspects of this management, including the operational costs. Our objectives were to retrieve available harvesting cost models from published studies on selection cutting in Norway-spruce-dominated stands in Scandinavian countries and to evaluate them against real case studies. First, we retrieved three recently published harvesting cost models which provided explicit cost functions. Models 1 and 2, based on rotation forestry (RF) data and adapted for CCF, had separate sub-models for cutting and hauling costs. Model 3 was based on CCF data and produced total harvesting costs, including the cutting and hauling costs combined. Second, we measured cutting costs for 29 harvesting operations on stands with different stages of CCF structure. We then compared the observations with the simulations of Models 1 and 2 cutting cost sub-models for those cases. Third, we expanded the dataset, including a further 34 harvesting operations in stands with more advanced CCF structures (without measured costs). We then simulated the total harvesting costs for all three models in this dataset to investigate their general behaviour. On average, Models 1 and 2 cutting cost sub-models had relatively good and consistent predictions compared with the observed values. However, they differed in total costs due to different estimates for the hauling cost sub-models. Model 3 had predictions comparable to Models 1 and 2 in the more advanced stages of CCF, but much higher in the less advanced. This study provides important data regarding cutting costs in CCF and demonstrates the feasibility of using existing harvesting cost models.

Continuous cover forestry (CCF) is a form of forest management that avoids the use of clearcutting and thus maintains a continuity of woodland conditions across the site (

In the 2000s, loss of forest nature diversity and the climate crisis raised the need to improve the ecological sustainability and carbon sequestration capacity of forests (

However, various authors (

The forest work environment, defined by characteristics such as stand structure, terrain topography, and soil bearing capacity, affects harvesting activities (

^{3} ha^{-1}, while the total removal and average volume of harvested trees were 110-231 m^{3} ha^{-1 }and 251-410 dm^{3} respectively. Similarly,

Net Present Value (NPV) analysis, a system to measure the present value of the future cash flow the forest will produce, is the most common way to compare the profitability of different forest management methods. The method involves the calculation of long-term income and expenditure, discounting them from the present, and calculating the difference. If positive, the given silviculture system is profitable (

Since most of the articles addressing the profitability of CCF apply roadside prices with harvesting costs subtracted, it would be interesting to compare the underlying harvesting cost models. Our research questions are: (i) What harvesting cost models are available for CCF in the Nordic countries? (ii) Do they correctly predict the costs under a new independent validation? (iii) Do they produce similar outcomes when an identical thinning profile is applied?

We searched the Google Scholar® online repository for papers on profitability in CCF showing total harvesting cost models, including both cutting and hauling. We considered the following criteria: (i) an identical distribution of harvested trees can be fed into all models; (ii) the harvesting cost models result in total costs (expressed in euros); and (iii) recently published (in the last 15 years). Among the potential candidates (

Model 1 (

where _{1} is equal to 1.15, ^{-1}), _{t} is the volume of tree ^{3}), _{saw} is the total volume of sawlogs and _{pulp} is the total volume of pulpwood removed (both in m^{3 }ha^{-1}). This model was initially based on the empirical cost functions for RF stands from

Model 2 (

where _{1} is equal to 1.15, ^{-1}), and_{t} the volume of tree ^{3}). Similarly to Model 1, this model was initially based on the empirical cost functions for RF stands from

Model 3 was fitted by

where ^{3} ha^{-1}), and ^{3}). During model fitting, the logging cost and hauling cost rates were considered to be 70 € and 50 € per hour. Model 3 did not include fixed costs, while models 1 and 2 expected to include them as an additional term to reflect,

This dataset consists of 5 spruce-dominated stands, each with 4-7 rectangular plots (size either 800 or 1000 m^{2}) for a total of 29 plots (

Field trials were carried out during 2021-2022; two stands during winter harvesting in 2021; two stands during summer harvesting in 2021 and 2022; and one stand during winter harvesting in 2022. Three different single-grip harvester brands (John Deere 1170G, Ponsse Beaver, Komatsu 901 901XC) and four machine operators were used in the test cuttings. However, all the harvesters were medium-size types and intended for both thinnings and clearcuttings. Most operators had several years’ professional experience of CCF selection cutting, while one was extensively trained before the experiment. The strip road was placed and marked out approximately in the middle of the shorter sides of the plots (20 m), so that the harvester’s crane reach could reach across the whole plot. Before the trials, the characteristics of the plots’ growing stock were measured. After thinning, all remaining trees were mapped and measured by species and diameter at breast height (dbh), and the damage of trees and soil due to logging was measured. In the plots, after selection cutting, the basal area of remaining trees was in accordance with the forest management recommendations applied in Finland (

Four plots were already quite irregular in their structures and were selectively harvested similarly to an ongoing CCF (henceforth “advanced CCF”), thus indicating favourable stand characteristics for CCF management. The majority (14 plots) had more irregular structures than the average RF stand in Finland but still needed some level of transformation (“medium CCF”) to be fully managed according to CCF principles. The rest (11 plots) had a quite regular structure and underwent their first selective thinning for moving towards an irregular structure (“beginning CCF”). In each stand, there may have been plots of different structures, and average values for each category are presented in

The time and motion study (^{3} removed in each plot. Details on removals and cutting time and costs for each category are shown in

This dataset consists of 20 spruce-dominated stands with a total area of 1-2 ha each, all in a more advanced CCF stage than CCF-BASIS (^{2}) in each stand. The stands belonged to a long-term CCF permanent sample plot experiment involving single-tree selection starting in the 1980s. Most of the stands (16) belonged to the

Fourteen plots were subject to a selection cutting during the winter of 1996/1997 and then all of them during the winter of 2011/2012, for a total of 34 measurements. In both 1996 and 2011, the harvesting was carried out to enhance uneven-aged structures in the stands and to reset their basal areas to the original values of the 1980s. The emphasis of the removal were the larger diameter classes (dbh > 30 cm). However, some larger trees had to be retained in many stands to achieve the target basal area. On the other hand, a small number of mid-sized trees (20 < dbh ≤ 30 cm) was removed where there was a major surplus. Five stands were not harvested due to their operationally unjustifiable harvestable volumes. In 2011, the cutting was similar in type and execution, but removed volumes were larger, and all stands had an operationally justifiable harvestable volume due to the longer period of growth between harvests (15 years).

Advance tree and strip road marking, manual felling, debranching and cut-to-length culling with chainsaw, and forwarder hauling were employed during the harvesting. The trees to fell were selected and very well marked in advance, and other trees removed or lost in harvesting were considered harvest loss. The stand characteristics, including the details of removals, are shown in

All the analyses were carried out in R (

Second, we applied all three complete models to the combined CCF-BASIS and ERIKA datasets and estimated the total costs (cutting and hauling) to further analyse their behaviour against an extended dataset, albeit without observed values. In this case, ERIKA plots were assigned to the advanced CCF stage of the CCF-BASIS dataset. We only carried out visual analyses of the simulation results.

The simulations of both models fitted well with the observed costs, resulting in a root mean square error (RMSE) of 1.05 € and 1.07 € for models 1 and 2 respectively (

The estimate for the total harvesting costs (cutting and hauling) of the three models had different trends against total removals (^{-3} were removed from the ERIKA dataset, where total removals and average tree volume were especially low (for the former, 17.5 and 26.04 m^{3} and for the latter, 176 and 274 dm^{3}). On the other hand, Model 3 showed slightly increasing costs for increasing total removals, as cutting and hauling were modelled simultaneously, the underlying reason being the capacity limit of the workload in haulage with increasing removal.

On average, for all data, Model 2 estimates were significantly lower than for Model 1, whose estimates in turn was significantly lower than for Model 3 (p-value < 0.01 after Welch two sample

In this study, we assessed the harvesting costs for CCF in boreal settings, which is crucial for determining the feasibility of this silvicultural approach in the current debate. Given the increasing interest in CCF, it is important to provide stakeholders with the best tools for decision making. Our first objective was to retrieve published harvesting costs models literature that could be applied in simulation studies. We found only three models, all with limitations. Models 1 and 2 were fitted on RF data and adjusted to CCF with some broad assumptions. They were both also used in subsequent studies about the economic assessment and optimisation of CCF management (

Our second objective was to evaluate the predictions of the retrieved models during an independent validation. This was limited to the cutting cost part of Models 1 and 2, using only the CCF-BASIS dataset (29 harvesting operations). The simulated values of the two models were not significantly different, and both well fitted the observed values (the root mean square error was around 1 € m^{-3} in both cases). The two models reached very similar results, even starting from very different hourly cost assumptions for cutting: 86 € for Model 1 (published in 2010) and 120 € for Model 2 (published in 2017), respectively similar to and much higher than the 85 € paid in the observed operations. Estimates were more correct for advanced continuous cover forestry stages than for the beginning and medium stages, although there were only 4 replicates for the former and 25 for the latter. Cutting costs for the advanced stage included a 15% increase compared to rotation forestry, as suggested by the model developers. We found this correction unnecessary for the beginning and medium stages, where the operating conditions and removals were closer to the rotation forestry stands used for the models’ calibration.

Regarding the third objective, we analysed the total cost prediction of all three models, verifying their predictions using identical thinning profiles. We expanded the dataset by including 34 more harvesting operations from the ERIKA dataset, a long-term experiment involving selectively harvested stands, for a total of 63 operations. Models 1 and 2 showed significant differences in the total costs which were not observed after the analysis of the cutting costs only. Model 1 estimates were on average 1.77 € m^{-3} higher than Model 2. On average, Model 3 predicted values were higher than those of both Models 1 and 2, with a difference of 2.5 and 3.2 € m^{-3}, respectively, considering all the data. The differences decreased moving towards the more advanced continuous cover forestry stage, where the total costs were not significantly different for Models 1 and 3 (9.60 and 9.35 € m^{-3}, respectively) but still significantly lower for Model 2 (7.62 € m^{-3}). The data used for the calibration of Model 3, which are real case studies of selection harvesting in CCF operations, showed an average of 9.44 € m^{-3}. We found only one comparison in the literature (^{-3} for values of the average volume of felled trees of 0.3-0.5 m^{3} (comparable to our advanced CCF stage), similar to Model 3 but higher than values obtained using both Models 1 and 2. It must be noted that for total costs, the hourly cost assumption for cutting was lowest for Model 3 (70 €), compared to the previously mentioned 86 € and 120 € for Models 1 and 2, respectively. On the other hand, hauling cost assumptions were less different among the three models: 60 € for Models 1 and 2, and 50 € for Model 3. Furthermore, it should be stressed that Model 3 was the only model based on actual CCF logging conditions, while Models 1 and 2 were initially modelled according to logging conditions in rotation forestry.

Since there were differences between the models regarding total costs, this may have a consequence for stand-level optimisation, as even small changes in endogenous variable values could have a profound effect on optimal solutions (

Our analyses did not include fixed costs. Models 1 and 2 used values of 300 € and 100-500 € respectively, which could remarkably affect the low-intensity operations described in this study. Given an average of 75-90 m^{3} total removals per ha observed in the advanced continuous cover forestry stage, these fixed cost values could have an impact on the harvested cubic metre, ranging from 1-7 €. It is evident that an adequate estimate of such costs is needed.

Regarding the cutting costs, Models 1 and 2 did not significantly differ in their simulations, and they reliably estimated the observed data in all stages, although we did not use the suggested 15% increase in cutting costs for the first phases of the transformation to continuous cover forestry (here, the beginning and medium stages).

Regarding the total harvesting costs, we found different behaviours for the various stages. For the beginning and medium stages, Models 1 and 2 seemed to provide more adequate harvesting costs than Model 3, which was not calibrated using such conditions. For the initial stages of CCF, we suggest the former models are used, although Model 2 provides consistently lower costs that Model 1. In the more advanced CCF stages, Models 1 and 3 gave similar results, considerably higher than Model 2. The few verification data available seem to suggest the use of Model 1 or Model 3. However, it is necessary to validate these results against a larger number of cases of advanced continuous cover forestry operations or to prepare new models.

This study was supported by Luke’s project CCFBASIS “Technological and conceptual basis for Continuous Cover Forestry” (grant no. 41007-00182900).

Location of the study sites included in the different datasets used.

Size distribution of total standing trees pre-harvesting (grey) and of the felled trees (red), averaged for each category, per hectare. Both CCF-BASIS (beginning, medium, and advanced) and ERIKA datasets.

Simulated cutting costs for CCF-BASIS plots using Model 1 (

Residuals of the simulated cutting costs for CCF-BASIS plots using Model 1 (

Simulated cutting costs for CCF-BASIS plots using Model 1 (

Total harvesting costs for each plot of both CCF-BASIS and ERIKA. Simulated values according to Model 1 (

Summary statistics for the plots and harvesting operations in each category. (BA): basal area. Values are mean ± standard deviation. Productivity is calculated using an increased cutting time of 15 minutes per hour.

Statistics | CCF-BASIS(Advanved) | CCF-BASIS(Medium) | CCF-BASIS(Beginning) | ERIKA |
---|---|---|---|---|

No. of harvesting operations | 4 | 14 | 11 | 34 |

Stand BA pre-harvest (m^{2} ha^{-1}) |
24.1 ± 2.0 | 23.9 ± 10.9 | 33.0 ± 4.2 | 21.5 ± 5.0 |

Stand BA removed (m^{2} ha^{-1}) |
8.9 ± 2.9 | 17.8 ± 4.0 | 18.7 ± 7.2 | 8.5 ± 3.6 |

Volume removed (m^{3} ha^{-1}) |
75.0 ± 29.4 | 179.4 ± 63.5 | 207.0 ± 80.6 | 91.9 ± 45.4 |

Average volume of removed trees (m^{3}) |
0.31 ± 0.14 | 0.50 ± 0.30 | 0.63 ± 0.21 | 0.44 ± 0.32 |

Productivity, with delays (m^{3} h^{-1}) |
20.9 ± 5.1 | 45.5 ± 10.8 | 32.8 ± 11.5 | - |

Cutting costs (€ m^{-3}) |
5.4 ± 1.6 | 3.9 ± 1.4 | 2.1 ± 0.3 | - |

Results of the models’ independent validation for each category, both for the cutting costs only and the total cost analysis. Values are mean and standard deviation, in € m^{-3}. For the total costs, “Advanced” include the respective CCF-BASIS plots plus the ERIKA dataset.

Category | Model | Beginning | Medium | Advanced |
---|---|---|---|---|

Cutting costs only | Model 1 | 3.2 ± 0.3 | 3.9 ± 1.1 | 5.0 ± 1.0 |

Model 2 | 2.8 ± 0.5 | 3.7 ± 1.8 | 5.3 ± 1.4 | |

Total costs | Model 1 | 6.4 ± 0.4 | 7.2 ± 1.3 | 9.6 ± 2.9 |

Model 2 | 4.9 ± 0.6 | 5.8 ± 1.8 | 7.6 ± 3.4 | |

Model 3 | 10.7 ± 2.6 | 11.1 ± 1.8 | 9.5 ± 2.3 |