Estimation Of Forest Biomass Model Based On OLI Remote Sensing Image Data In Xining City

With the influence of forest biomass for global climate change and global sustainable development is increasing, forest biomass monitoring research attracts more and more attention. Fast, accurate estimation of forest biomass is of great significance to global change.Based on Landsat-8 OLI remote sensing image data combined with field survey data during the same period, the research aimed to build a forest biomass model of North and South mountain for estimating forest biomass in Xining Qinghai Province. The different bands spectral reflectance of each plot was obtained, and the vegetation indices, PCA1, bright, green, wet, slope, aspect and derived factors by linear or nonlinear transformation were computed by using ENVI and ArcGIS software.In order to select appropriate factors as variables for the forest biomass model, the correlations between those factors and forest biomass were calculated. Then the multiple regression model was built by using stepwise regression analysis.Comprehensive analysis of the equation of significance(Sig), adjusting values of the correlation coefficient(adjust R2) and model independent test index(F), chose the most suitable for forest biomass estimation model of Xining North and South Mountain. The main results showed that:①Only OLI-5 had a higher correlation in the selected six original bands, showing extremely significant positive correlation(P<0.01). DVI showed significant correlation(P<0.05). After the linear and the nonlinear variation of the variables,(OLI-5)2, eDVI were extremely significant positive correlation(P<0.01). That could see the correlation of OLI-5 and DVI were increased after nonlinear transformed.②Took OLI-5, DVI, and derived factors OLI-52 and eDVI as established biomass estimation model of independent variables, then used stepwise regression analysis, and the forest biomass estimation model has been built. The mode is B=119.495+3.704E-24eDVI+0.026(OLI-5)2-3.478(OLI-5), with its multiple correlation coefficient was 0. 554(P<0. 01).③The curve equation established by derived variable is better than those factors without linear or nonlinear transformation. The curve equation established by a number of variables is better than the curve equation established by a variable.④The absolute error between the predicted and measured values obtained by inversion of the value is 1.600t/hm2. The predicted average relative error was 13.509%. The ratio of the inversion error and the measured value is in the range of 0.101~0.366. That meant the model could be used to estimate forest biomass in Xining, Qinghai Province.⑤The results showed that the average forest biomass density of Xining City was 5.227t/hm2, and the total forest biomass of Xining City was 998991.768 t.