Merging Soil Moisture Multi-model Products Based on Dynamic Bayesian Model Averaging

Soil moisture (SM) plays a critical role in hydrological processes, land-atmospheric interaction processes, and bio-ecological processes. SM can be obtained from the model produced products such as the global analysis, reanalysis products (e.g. GFS, ERA5 and ERA-Interim form ECMWF) and the land surface data assimilation systems (LDAS), such as Global Land Data Assimilation System (GLDAS). Recently, China Meteorology Administration (CMA) Land Data Assimilation System (CLDAS) version 2, which assimilates several data of regional automatic surface observations and satellite observations, has the finest spatial (0.0625° × 0.0625°) and temporal (hourly) resolutions over China. However, SM data obtained by model products exhibit large uncertainties compared with ground observations. To improve the accuracy of SM estimation from model products, the dynamic Bayesian model averaging (DBMA) is applied to merge eight SM model products during 2010-2017 in eight climate regions over China. SM data are estimated using dense distributed ground observations as a reference standard and then compared with the regional model product CLDAS. Results show that DBMA provides better estimates of hourly SM than single global model products and comparable estimates with the high-resolution CLDAS in most regions. This study reveals the superiority of the DBMA method over the traditional BMA method in merging soil moisture model products and the merged SM products can be used in hydrometeorological predictions and verification in the future.