Study On Estimation Of Forest Biomass Based On Remote Sensing Data

In this paper, a model of multi-regression and neural network was established based on TM imagery and data of 232 plots of forest inventory to estimate the biomass in the southern side of Xiaoxing'an mountains.The 20 dependent variables(including circumstance factors , biology factors and remote sensing factors),such as NDVI(normalized differential vegetation index) , RVI , EVI and TM imagery specific value et al,was regressed stepply ,in which 13 dependent variables were seleted to construct the multiple regression equation and R of which is 0.844, and passed the partial correlation coefficient test, spearman rank correlation coefficient test, D-W test. The average accuracy, estimated for independent plots by the model of feed-forward neural network which be established under the enviroment of MATLAB6.1, is 90.61%.A principal component analysis on ensemble estimate plots show : in first principal component the TM2 which indicates green light in 0.52~0.60 um, the TM3 which indicates red light in 0.63~0.69the TM3, TM∑ 1~7 which indicates light absorbd by chlorophyl and terra, rock, grads of soil sort recognise provide a majority of contribute; in second principal component TM(4 X 3/7),TM(4/3) and TM4 that indicate the grad of covery vegetation and reflecte of water provide a majority of contribute.Collected weather data of 1999 from 44 weather station in Heilongjiang province to insert value by kriging and generate grade date in 1 Kmx 1 Km to estimate the net primary production(NPP) of terrestrial vegetation in study area in 1999 by TM imagery based on the CASA model and Maimi model. The result of estimation by Miami model is 5~8×108tC/a; the result of estimation on October NPP in study area by CASA model is 0.4~1.6× 108tC/m.Distill TM imagery value from distributing area of forest in grad and estimate biomass by multi-regression and neural network model, estimate woods carbon storage by the result and transform coefficient; estimate soil carbon storage by forest type and transform coefficient; the sum of two result is the total carbon storage of forest in this area: 61~225t/hm2.