Biomass estimation of desert steppe based on spectral indices along a precipitation gradient

Abstract

Accurate estimation of grassland biomass in desert steppe is crucial for various studies such as grassland dynamic monitoring, scientific management, and environmental degradation. Observational tests were conducted in the Ordos desert steppe in July and August from 2014 to 2016 to measure the grassland biomass and grassland canopy spectral data along a precipitation gradient, to analyze the spectral reflectance characteristics and the correlation between different spectral indices and grassland biomass in the range of 350–1830 nm, and to establish and verify a model for estimating grassland biomass of desert steppe. The results showed that the spectral reflectance of the Ordos desert steppe is similar to the spectral characteristics of typical green vegetation. The spectral reflectance values of control and of lightly, medium-, and heavily degraded plots successively increased in the visible light band at 350–696 nm and decreased in the near-infrared band at 792–1349 nm. The sensitive band of the normalized difference spectrum indices after optimization was 705 and 710. The coefficient of determination of normalized difference spectral indices (705 and 710) and grassland biomass was R2 = 0.735. These values were significantly correlated at the p < 0.01 level. The R2 values of the fitted power function model and the verification power function model were 0.764 and 0.746, respectively. The root mean square error was 41.59 g/m2. The relative error was 28.37%, and the constructed power function model was the best model. Therefore, the optimized spectral indices can be used to effectively estimate the grassland biomass of desert steppe.