Ecological indicators, such as soil moisture content and gross primary production (GPP), play important roles in the management of ecosystems and water resources and within climate change research. However, studies on monitoring and simulation of ecological indicators of arid zones remain limited. This study simulated soil moisture (SM), latent heat (LH), and gross primary productivity (GPP) for the arid Yarkant River Basin in China using the High-Resolution Land Data Assimilation System and Deep learning. Simulation of hydrology over a large spatial scale is difficult due to a lack of observed data and the difficulty in parameterizing models to represent complicated ecological mechanisms. A comparison of Noah Multi-Parameterization (Noah-MP) simulations to multiple datasets at an annual scale obtained correlation coefficients exceeding 0.8. The model was able to replicate the broad temporal dynamics of GPP and LH over the Yarkant River Basin. Point- and regional-scale assessments across the Yarkant River Basin obtained the optimal model performance. The temporal trends in simulated SM anomalies were consistent with observations across the different sub-basins. The simulations of the Noah-MP model driven by Global Land Data Assimilation System (GLDAS) and ERA5 data provided accurate representations of most hydrological variables, except for in the upper reaches of Yarkant River Basin, likely due to the extensive freeze-thaw activity in this transitional region. The results of this study can fill hydrological data gaps in arid zones areas.
Read full abstract