Jump to Main Content
Modelling above-ground biomass based on vegetation indexes: a modified approach for biomass estimation in semi-arid grasslands
- Wang, Guoqiang, Liu, Shuman, Liu, Tingxi, Fu, Zhiyuan, Yu, Jingshan, Xue, Baolin
- International journal of remote sensing 2019 v.40 no.10 pp. 3835-3854
- aboveground biomass, arid lands, climate change, ecosystems, grasslands, normalized difference vegetation index, reflectance, remote sensing, satellites, soil, spatial data, statistical models, uncertainty, vegetation cover, China
- Remote sensing could be the most effective means for scaling up grassland above-ground biomass (AGB) from the sample scale to the regional scale. Remote sensing approaches using statistical models based on vegetation indexes (VIs) are frequently used because of their simplicity and reliability. And many researchers have already proven the method is scientific, feasible, and can bring relatively better effects in practice. However, the only deficiency of the method has been criticized because of the uncertainties introduced by saturation of spectral reflectance at high-density vegetation levels and the soil surface at low-density vegetation levels. Therefore, in this study, we aimed to improve grassland AGB estimates by using modified VIs (MVIs) to minimize the influence of the soil background. The field study was conducted in the Chen Barag Banner, the Ewenkizu Banner, and the Xin Barag Left Banner in the Hulun Buir Grassland, Inner Mongolia, northern China. Field plots were photographed and AGB samples were collected during field sampling. Remote sensing data were obtained from MOD09A1 (TERRA satellite). Four MVIs were first calculated based on the corresponding VI: the Ratio Vegetation Index (RVI), the Normalized Differential Vegetation Index (NDVI), the Difference Vegetation Index (DVI), and the Modified Soil-Adjusted Vegetation Index (MSAVI), by improving estimates of vegetation cover (VC). Then, MVIs, i.e., MRVI, MDVI, MNDVI, and MMSAVI, were regressed with the sample-scale AGB using an exponential function, a linear function, a logarithmic function, and a power function. When the accuracy of the models was tested by comparing root mean square error (RMSE), relative error (RE), and coefficient of determination (R²), the results demonstrated that MVI-AGB models performed better than the VI-AGB models. The logarithmic MNDVI-AGB model was the best of the regression functions. This model gave the best estimates of AGB from remote sensing data, compared with the values measured in field analyses. Our proposed method provides a new way to estimate regional grassland AGB and will be useful to analyze ecosystem responses under climate change.