Forest Harvesting Decision Support for Sustainable Development

The principles outlined here apply as much to other types of regional analysis, and not just forestry, although forestry applications show many of the complexities of the general problem. The objective is first to map the forests in some valid fashion, then to predict their growth in the future, and finally to schedule harvesting that both preserves the regeneration rate of the whole forest, preserves biodiversity, and allows sensible access and road construction programs. None of these three have obvious answers. Traditional forest mapping has been performed by skilled air-photo interpreters outlining stand types according to the appropriate classification scheme. There are many difficulties with the identification of these boundaries, and with the verification of the interpretation on the ground. Some recently proposed alternatives include fuzzy boundary methods, and automated individual tree crown identification. Growth rate prediction of established forest is reasonably satisfactory using the yield-curve approach, but there has been little success in predicting the type of forest to expect after cutting thus handicapping long-term sustainable-yield planning. Most decision support systems for planning the forest harvesting have used classical optimization techniques, with the adjustment of various parameters to achieve an acceptable simulation of test data. Experience to date suggests that these systems are difficult to use in an operational context. Moreover, current methods do not usually take adequate account of the spatial distribution of the cut schedule. This means that the results must be drastically modified to accommodate the realworld engineering problems: especially those concerning access and road construction. We are attempting to address these issues in several ways. Firstly, a North American workshop on harvest scheduling methods took place in March 1997, and the practical discussions may help us to reorient future work. Secondly, new graph-theoretic algorithms may help to put the primary emphasis back on the spatial distribution of the cutting program by basing results on the existing or proposed road network. Thirdly, it appears that practical systems should be based more on the idea of a toolkit, rather than a program that provides final answers. This 'flight-simulator' approach both allows input from the forest manager, and provides manager training by indicating the consequences of his proposed actions.