A Silvicultoral Decision Support System for Loblolly Pine Plantations

Abstract

Different heuristic techniques were used to solve the stand management problem. The techniques, together with a growth and yield model and an expert system, form the methodological components of a decision support system developed for deciding forest management activities throughout a stand rotation. Random search (RS), simulated annealing (SA) and tabu-search (TS), were used to solve the optimization problem. A Growth and Yield model for Loblolly pine in the Carolina piedmont was used to predict forest growth under different thinning and/or fertilization regimes. Silvicultural activities like competing vegetation control, soil preparation, and early fertilization were included as early growth treatments for different soil and site conditions. Response to establishment treatment was predicted using a knowledge system based on expected gain and growth response associated with several soil types. Sensitivity analysis was undertaken to fine tune solutions from SA and TS. Recommended parameters for SA were: alpha = 0.99, max-iteration = 100000 and min-temperature = 0.01. The combination of two diversification techniques: least-used move and swap move were found to interact for improving solutions using TS. Tabu-tenure was found optimal at 50 for a neighborhood size of 156. Tabu-Search proved to be the best technique of all.