关键词: 地上生物量, 植被指数, 高光谱, 典型草原, 反演模型

Abstract: Grassland biomass is an important parameter of grassland ecosystem, and the complexity of grassland canopy vegetation spectrum makes it a challenge to evaluate grassland growth status in a long term. However, at present, few studies have deeply explored the original spectral information of typical grasslands in Inner Mongolia, and discussed the impact of spectral information of surface objects on the estimation of existing above ground biomass. In this study, ASD Field Spec3 portable spectrometer was used to collect the canopy height spectral data of grassland in Lemaudeng Pasture, Xilin Gul, Inner Mongolia from July 2017 to August 2018, and the reflectance spectral curve of grassland was analyzed to characterize the trend of vegetation change. At the same time, the optimal prediction model is selected by spectral preprocessing method combined with various hyperspectral models. The results showed that: (1) From the perspective of comparing the prediction accuracy of different GLM generalize linear models, the best hyperspectral modeling method is select $\frac{S D_r}{S D_b}$, the best model for the variable is y=-3.795 3x²+60.065x-78.455(x is $\frac{S D_r}{S D_b}$, y is the estimated fresh weight of aboveground biomass), fitted R² =0.662, Predicted R²=0.302. (2) In the analysis between hyperspectral variables and above-ground biomass dry weight, select $\frac{S D_r-S D_y}{S D_r+S D_y}$. The best model as a variable is y=7.744e^3.4349x (x is $\frac{S D_r-S D_b}{S D_r+S D_b}$, y is the estimated dry weight of above-ground biomass), fitted R²=0.559, Predicted R²=0.304. The results of this study have scientific value for hyperspectral prediction modeling of grassland biomass.

Key words: Above-ground biomass, Vegetation index, Hyperspectral, Typical grassland, Inversion model