Forest structure optimization using evolutionary programming and landscape ecology metrics

Abstract

A new design framework is introduced for forest structure optimization based on a genetic algorithm landscape encoding and landscape ecology metrics. Landscape ecology is an important interface between the forest management community, which is a traditional user of operations research methods, and the biological conservation community which is relatively new to OR methods and whose goals are increasingly allied with the spatial ecology concepts emerging from landscape ecology. Deforestation and forest fragmentation are increasingly recognized as the underlying drivers of global biodiversity loss, hence forestry management will need to explicitly incorporate spatial ecology objectives. A deforestation model is presented which simulates a landscape progressively fragmenting by the incremental removal of forest patches. Principal components analysis (PCA) of multiple deforestation simulations captures the relative influence of the mean proximity index and the mean nearest neighbour distance, two widely used landscape ecology metrics. An evolutionary programming method based on a genetic encoding of landscape structure is used to optimize forest patch selection by maximizing landscape performance with respect to single and multiple landscape ecology metrics weighted according to the PCA. This optimization approach is envisioned as a key component of a new forestry OR paradigm for designing multi-use landscape systems, incorporating both biodiversity and community needs.