Estimating grassland aboveground biomass on the Tibetan Plateau using a random forest algorithm

Abstract

Effective and accurate monitoring of grassland aboveground biomass (AGB) is necessary for improving our understanding of regional carbon cycle and pastoral agricultural management. In this study, we developed a suitable AGB estimation model for the Tibetan alpine grasslands based on the random forest algorithm, using 256 AGB observation data, remote sensing vegetation indices, meteorological data, and topographical data. We estimated the grassland AGB on the Tibetan Plateau during 2000–2014, analyzed its spatiotemporal changes, and further explored the response of AGB to the variation in climatic factors. The results indicated that (1) the RF model performed well in the AGB estimation, which can explain 86% of the variation of the observation data. (2) The grassland AGB decreased from the southeast to the northwest in this region, with an average value of 77.12 gm−2. (3) In the whole study area, the grassland AGB showed significantly positive correlation with temperature and precipitation. The correlation between grassland AGB and MAP was 0.54 (P < 0.05), much higher than that of MAT (R = 0.38, P < 0.05). (4) The inter-annual variation of AGB on the Tibetan Plateau was significantly and positively correlated with temperature (R2 = 0.45, P < 0.05). This study demonstrated that RF model can help improve our understanding of the spatiotemporal dynamics of the grassland AGB and the effects of climate variation.