| The research of forest vegetation plays an important role in the research of forest ecosystems. This paper analyzed the correlation between the remote factors, topographical data and Canopy density and measured biomass taking Shimian County of China's in Sichuan Province as the study area. The Landsat-5 TM data, DEM data and forest resource inventory data are the data sources. Firstly, Landsat 5 TM was geometric corrected by LRM model and SCS model, and collecting the samples which are uniform internally and the area is more than one hectare as the study samples. Secondly, the biomass for each sample was calculated by using the forest resource inventory data, and the biomass GIS database was established according to the forest map. Thirdly, calculation the vegetation index, the ratio of bands, principal component analysis, tasseled cap transform data. Fourthly, overlay analysis is used between the samples biomass and the remote-sensing geological data which are sampling by Center method, window method and mean method, then, the correlation between the Landsat TM and its derived data, topographical data, Canopy density and the biomass were analyzed. Finally, Each optimal sampling factors which is obviously correlated with biomass at 0.05 confident level is as independent variables. Using multivariate statistical analysis and the BP neural network to estimate biomass, then to evaluate the accuracy of the estimation models, And eventually get the estimate model of each specie group. The conclusions of the research are as follows:Comparing three estimation methods, the results are that: when the brich, alnus, pinus yunnanensis, soft broadleves, spruce and quercus are estimated by BP neural network, the precision is highest,which were respectively75.9%, 73.5%, 73.8%, 63.8%, 71.1%, 70.7%;When the tsuga chinensis and abies are estimated by stepwise regression analysis the precision is highest, which were respectively 81.9% and 60.1%. |
PDF Full Text Request