| Based on TM imagery and plots data of forest inventory in 1990 and 1994, the biomass in Maoershan Forest Farm, Shangzhi City was estimated by using interpolation of GIS, multi-regression and neural network in this study.Firstly, the biomass value of each plot was calculated by using the forest inventory data of 1990 and 2004 and one element estimation model. Secondly, according to 192 random samples of plots, the forest biomass of other plots was obtained by using interpolation. There was no significant difference between the two sets of data through Paried-Samples T Test. Therefore, the interpolation could be used to estimate the biomass. Then, principal component analysis was implemented by applying 16 dependent variables(including environmental factors and remote sensing factors). Then, multiple regression equation was established by selecting 6 principal components, the equation coefficient test, equation test and D-W test were passed and R of 0.812 was acquired. The average accuracy was 96.063% by building BP neural network and estimating biomass of plots on MATLAB7.0.The spatial distribution maps of forest biomass of 1990 and 2004 were obtained by interpolation of GIS. The results showed that the biomass of research area increased after 15 years'management through comparison. Biomass, altitude and slope were related by significant correlation through analyzing the relationship between biomass and topographic factors.
|
PDF Full Text Request