Abstract:

The models of forest self thinning are generally nonlinear and dynamic. The artificial neural network has the characteristic of expressing arbitrary nonlinear mapping, which provides theoritic feasibility for modeling forest self-thinning law. Based on the principle and algorithms of the neural network model based modified simplex method (BP-MSM mixed algorithms), this paper analyzed the effect of BP-MSM mixed algorithms and Zhang's model for modeling forest self thinning further. The comparisons in forest self-thinning of Populus tremula var. davidiana forest, Pinus yunnanensis forest and Cunninghamia lanceolata forest illustrated that the simulated effect of BP-MSM mixed algorithms were superior to Zhang's model significently when establishing network structure of 1∶5∶1 by three layers. The results of forest self thinning examples showed the surplus square of BP-MSM mixed algorithms were only 3.89%～27.16% of Zhang's model, which were satisfactory and its precision were higher. This study will enrich the simulating method of forest self-thinning, but the network structure of BP-MSM mixed algorithms is important by choosing the numbers of concealing layer and neural points.

Key words: Self-thinning law, BP-MSM mixed algorithms, Forest self-thinning