Application of Digital Surface Model for Estimating Forest Stand Volume Using Regression Methods

Ground inventory provides most accurate data for management purposes in forestry، but it is costly and time consuming. Up to now، various approaches has been proposed for reducing the amount of ground inventory. In this paper، a new approach had proposed and tested for estimating forest stand volume. The approach is based on this idea that forest volume can be estimated by variation of trees height at sample plots. For this aim 150 circular sample plots with systematic random design were collected. The DSM has been extracted from UltraCamD images with 1 to 10 m size by 1m span. Corresponded to ground samples، standard deviation and range of DSM had extracted. In order to modeling، different regression methods and one layer perceptron were used. The results showed that standard deviation of 5 m DSM was the most appropriate data for modeling. Perceptron and linear regression were better than other modeling methods. At best، Bias and RMSE of forest volume were 2 and 47 percent، respectively.