An improved workflow to identify an optimal cable road layout for a large management unit

Cable-based technologies are crucial for the design of harvesting systems in steep terrain. The identification of a cable road layout for a management unit aims at concurrently minimizing harvesting costs and environmental impacts. Snow processes are particularly considered in an Alpine environment when minimizing for the cable roads running in the same direction as the slope line. The problem of selecting the optimal cable road layout can be formulated and solved to optimality as a multi-criteria set cover location problem as proposed by Bont (2012). The approach represents the forest subject to harvest as a raster which must be entirely covered by cable roads. Moreover, helicopter logging is assigned at sites where it outperforms cable roads. Despite the use of the high-performing solver Gurobi in the current implementation, the long time spent to calculate cable road layout alternatives for various objective weights hampers the practical application on a larger scale. Here, we present an improved workflow for the approach which aims at reducing (1) the number of computed alternatives to characterize a Pareto set and (2) computational time via parallelization of the optimization task for subunits constituting a large management unit. We have expanded the workflow with the non-inferior set estimation method (NISE) to plausibly reduce the number of alternatives required to create an approximate Pareto set. That workflow was then modified to parallelize the computation of sub-units on a computing cluster. The resulting workflow was tested for a 300ha management unit in a Subalpine forest located in Klosters (Switzerland). The management unit was partitioned into 5 subunits and the forest subject to harvest was represented as a 15x15m raster. This resulted into subunits consisting of 1’515 to 3’066 forest parcels and 11’185 to 20’935 cable road alternatives, respectively. Four to five alternatives were computed for each subunit to approximate the Pareto sets. The computational times required to solve for the subunits in parallel ranged from 10 to 230 minutes whereas the highest value sets threshold for the overall computational time. -- Bont, Leo Gallus. Spatially explicit optimization of forest harvest and transportation system layout under steep slope conditions. ETH Zurich (2012). http://dx.doi.org/10.3929/ethz-a-00755802