A Regional Forest Tree Layer Biomass Estimation Method Based on Clustering Analysis

Abstract

As an area always contains varies of tree spices or forest types, therefore, when using biomass estimation model based on single tree or forest stand to estimate regional biomass, the modeling workload is big, and the existing models do not adequately reflect the factors that influence the biomass. Aiming at the problems above, this paper proposes a regional forest tree layer biomass estimation method based on clustering analysis, using the forest resources survey data of the study area as the research object, using principal component analysis to extract characteristic factors from 17 indexes, using the improved K-means algorithm to clustering the forest subcompartment, and using support vector regression algorithm to separately build the biomass estimation model based on clusters. The results show that 8 principal components can reflect over 80% information of the original data; the subcompartment of the study area can be divided into 6 classes, the coefficients of each model are ranging from 0.7 to 0.92, the average relative error absolute values of each model are ranging from 11.173% to 23.583%, this method has got a satisfactory accuracy, which can provide a new way for regional biomass estimation.