How large should forest harvest compartments be: insights from a simulation study within boreal black spruce forest

Abstract From forest value chain optimization perspectives, existing long-term forest management plans are non-optimal because they pay little attention to location and spatial issues that are critical to timber supply costing. In this study, we designed a simulation experiment to measure the impact of harvest compartment size (with variable spatial details) on the uncertainty in standing volume estimations and examined options to protect timber supply under varying compartment size types. Data for the study came from a commercial forest, thus, represent real cases of harvested areas, transportation, loading and hauling, road construction and maintenance costs. In order to evaluate the impact of the size of forest compartments on the variability in standing volume, we created seven forest compartment size types including 61 km 2 , 12, 6, 3.0, 2.0, 1.5, and 1.0 km 2 using the Spatially Explicit Landscape Event Simulator (SELES). Merchantable wood volumes were estimated for each forest stand within a compartment using a non-parametric nearest neighbour method and a bootstrap procedure was used to estimate the coefficient of variation (CV) associated with the mean standing volume of each compartment size type. We developed four models that offer variable protection against potential disruptions (variability) in timber supply related to compartment size type variability. We implemented each of these plans with Monte Carlo simulations using the IBM ILOG optimization programming language and solved them using ILOG CPLEX optimizer. Results show that, reducing the harvest compartment size from 61 km 2 to 1.5 km 2 led to an increase of the CV of standing volume from 6% to 11%. With the compartment size of 1.2 km 2 , our simulation experiment failed to converge even after 12 hours of running. Since the minimum tolerable CV of 11%, is equivalent to the CV of the compartment size of 1.5 km 2 , we recommend maintaining a minimum compartment size of 1.5 km 2 . Methods that modify the even-flow constraint such as the periodic model, or the adaptive model tested in this paper are preferable as opposed to an approach where the targeted periodic timber revenue is reduced by a fixed percentage across the planning horizon.