Estimation and evaluation of high-resolution soil moisture from merged model and Earth observation data in the Great Britain

Soil moisture is an important component of the Earth system and plays a key role in land-atmosphere interactions. Remote sensing of soil moisture is of great scientific interest and the scientific community has made significant progress in soil moisture estimation using Earth observations. Currently, several satellite-based coarse spatial resolution soil moisture datasets have been produced and widely used for various applications in climate science, hydrology, ecosystem research and agriculture. Owing to the strong demand for soil moisture data with high spatial resolution for regional applications, much effort has recently been devoted to the generation of high spatial resolution soil moisture data from either high-resolution satellite observations or by downscaling existing coarse-resolution satellite-based soil moisture datasets. In addition, land surface models provide an alternative way to obtain consistent high-resolution soil moisture information when forced with high-resolution inputs. The aim of this study is to create and evaluate high-resolution soil moisture products derived from multiple sources including satellite observations and land surface model simulations. The JULES-CHESS simulated soil moisture and satellite-based soil moisture datasets including SMAP L3E, SMAP L4, SMOS L4, Sentinel 1, ASCAT, and Sentinel 1/SMAP combined products were first validated against observed soil moisture from COSMOS-UK, a network of in-situ cosmic-ray based sensors. Second, an approach based on triple collocation was applied to compare these satellite products in the absence of a known reference dataset. Third, a combined soil moisture product was generated to integrate the better-performing soil moisture estimates based on triple collocation error estimation and a least-squares merging scheme. From further evaluation, it is found that the merged soil moisture integrates the characteristics of model simulation and satellite observations and particularly improves the limited temporal variability of the JULES-CHESS simulation. Therefore, we conclude that the triple collocation merging scheme is a simple and reliable way to combine satellite-based soil moisture products with outputs from the JULES-CHESS simulation for estimating model-data fused high-resolution soil moisture for the British mainland.