Investigation of the Trends and Model Evaluations of multiple Hydrological Parameters over Global Arid Regions

Arid and semi-arid lands occupies over 30 percent of the land surfaces of the Earth. However, the dramatic human-made changes to the global arid and semi-arid land cover and their responses to such changes, as well as their impacts on the earth system has not yet been fully explored. It is critical to ascertain the nature of such changes and design potential mitigation. In this study, the remote sensing technologies are applied to investigate the changing trend in the global arid regions in the Köppen-Geiger climate classification: including four arid types: BSh (Hot semi-arid climate), BSk (Cold semi-arid climate), BWh (Hot desert climate) and BWk (Cold desert climate). The changing trend of Evapotranspiration (ET), Groundwater storage, precipitation and surface soil moisture (SSM) are presented during 2010-2019, showing Australia and part of southern Africa were experiencing severe drought in all parameters, while groundwater losses were found in the MENA and Western Asia region. The comparison of FLDAS and GLDAS datasets in SSM and ET, evaluated by the observation datasets from MODIS and NSAS-USDA, shows that FLDAS and GLDAS have strong consistency except desert region: China, Egypt, Saudi Arabia, especially a negative correlation in Egypt. To rebuild the FLDAS and GLDAS's products, automated model selection and tuning are performed among different candidate supervised models. The ensemble models (random decision forest and bootstrap decision forest) outperform others in showing effective performance in estimating ET and SSM values with (R2 > 0.83) and (R2 > 0.86), respectively.