A regional model-based algorithm to simulate root zone soil moisture with multi-source datasets

Spatiotemporal characteristics of root zone soil moisture (RZSM) play a key role in agricultural water management. Reanalysis and satellite products, e.g., ERA5, GLDAS, GLEAM and SMAP provide global-scale RZSM information, but the accuracy of these datasets usually falls short of practical needs in regional agricultural studies. We established a regional model-based algorithm for estimating RZSM, incorporating soil moisture (SM) measurements, as well as climate, topography, soil, vegetation and land use to account for the heterogeneity of the area. Taking the Yellow River basin, which spans multiple wet-dry regions, as the research object to validate the regional model and estimate RZSM information. The results showed that the superposition of predictor variables could progressively improve the performance of the regional model by setting different model input cases. The regional simulations outperformed those by nearby-station model, which was based on the parameter transplantation method, with a 21.2 % increase in the Pearson correlation coefficient (CC) and a 27.6 % decrease in the mean absolute error (MAE). The assimilated RZSM obtained using the regional model, driven by surface SM from reanalysis and satellite products, was closer to the measurements compared to the RZSM datasets provided by ERA5, GLDAS, GLEAM and SMAP, with CC improved by 38.0 %, 8.9 %, 1.5 %, and 3.1 %, and MAE decreased by 19.7 %, 5.5 %, 15.1 %, and 13.2 %, respectively. The ERA5-assimilated dataset exhibited the most significant improvement in accuracy in the study area, particularly in the semi-humid region. Overall, such a simulation scheme is proven to be effective in obtaining high-precision RZSM information at the regional scale and holds great potential for applications in agricultural decision-making and management.