Hyperspectral Remote Sensing Estimation Models of Aboveground Biomass in Gannan Rangelands

Abstract

Accurate estimates of grass biomass can provide valuable information about the productivity and functioning of rangelands and grassland ecological resource utilizing more reasonable. In order to improve the biomass quantitative analysis with hyperspectral remote sensing data, a field experiment was carried out in Gannan rangelands, Gansu province. To achieve this objective, fresh grass aboveground biomass and hyperspectral canopy reflectance were collected at four types pasture in august 2007. On the base of the analysis of spectral characteristic of four grasslands and correlation between original spectral, hyperspectral feature variables and aboveground biomass of four rest grazing grasslands, the experiment data were classified two groups. One group was used as the training sample to build the regression of models with the one-sample linear method, the nonlinear method and stepwise analysis method, another group was used to the testing sample to predict the precision of regression models. Results show that the regression of quadratic model using RVI provide a better univariate regression involving hyperspectral indices for grass aboveground fresh biomass estimation compared other models in Gannan rangelands, the estimation standard deviation was 0.178 (kg/m2), In conclusion, the results of this paper indicate that the grassland biomass can be estimated at the canopy level using the hyperspectral reflectance.