Estimation of alpine grassland above-ground biomass and its response to climate on the Qinghai-Tibet Plateau during 2001 to 2019

The Qinghai-Tibet Plateau (QTP) grassland is a critical part of the carbon pool of terrestrial ecosystems and provides important animal husbandry resources. To embody the stability of grassland ecology and the response of grassland above-ground biomass (AGB) to climate change on the QTP, four estimations models (partial least squares regression (PLSR), random forest (RF), Back-Propagation neural network (BPNN) and deep belief network (DBN)), which are based on statistics, machine learning and deep learning regression methods, respectively, were established to estimate the alpine meadow and alpine steppe AGB from 2001–2019. The results showed that: (1) the RF model performs well on the AGB estimation with the highest accuracy (R2 =0.84, RMSE=8.51gC/m2, MAE=6.46gC/m2) and stability (R2 =0.76, RMSE=9.24gC/m2, MAE =8.30gC/m2); (2) in spatial pattern, the AGB decreased from southeast to northwest on the QTP, and present a significant increasing with a rate at 0.19gC/m2a and 0.06gC/m2a on alpine meadow and alpine steppe during 2001–2019, respectively; (3) in the past 19 years, the AGB variations on the QTP showed a relatively stability with the average CV of 0.1; (4) the sustainability of the AGB show a weak anti-persistence (Hurst = 0.43), and it indicates that the future trend of the AGB is opposite to the current trend and the AGB is likely to be decreased in the future; (5) temperature and precipitation have a positive promotion on the most alpine meadow and alpine steppe vegetation growth on the QTP, while warming and wetting have negative promotion effects on the northwest of the plateau. These research results can provide useful reference materials for the study of the impact of climate change on alpine meadow and alpine steppe on the QTP.