Inter-comparison of satellite-retrieved and Global Land Data Assimilation System-simulated soil moisture datasets for global drought analysis

摘要

The multi-satellite-retrieved (ESA CCI SM) and the Global Land Data Assimilation System-Noah-simulated (GLDAS-Noah) surface soil moisture (SM) datasets are compared for global drought analysis over a multi-decadal time period (1991-2015). Global drought events and their duration, frequency and severity are assessed on a grid basis with soil moisture anomaly percentage index (SMAPI). The results show that the ESA CCI SM and the GLDAS-Noah based SMAPI values are significantly (p 0.05) correlated over most (83%) of the study region, of seasonally dependent. Both datasets show similar global patterns in drought duration, drought frequency and drought severity. The droughts present generally longer duration, higher frequency and more severity in arid and semi-arid regions than humid and sub-humid regions. The ESA CCI SM droughts are relatively higher in frequency and more intense in severity than the GLDAS-Noah SM droughts in many regions of the globe, while the two datasets show considerable differences in drought duration over arid, semi-arid and highly vegetated regions. For long-term trend detection, both datasets show high consistency in spatial pattern of SMAPI, with major significant drying trends in arid and semi-arid regions. Part (similar to 20%) trends are confirmed by the Global Precipitation Climatology Centre (GPCC) precipitation dataset using the Standard Precipitation Index (SPI). The two SM datasets exhibit large disparity in trending drought duration, drought frequency and drought severity. Despite that, both show major significant increasing trends in arid and semi-arid regions. Both soil moisture datasets are capable of identifying extreme drought events reported in southern China, North America, Europe and southern Africa. The ESA CCI SM dataset is more effective in determining the severity and spatial pattern of drought excluding densely vegetated regions, while the GLDAS-Noah dataset is more powerful in detecting drought occurrence, even over