A Vegetation-Temperature-Radiation-Composite Method for Downscaling Soil Moisture

To address limitations in regional-scale applications of passive microwave remote sensing soil moisture (SM) products, various downscaling methods have been proposed to enhance the spatial resolution of SM. Nonetheless, these downscaled methods, which are based on optical and thermal infrared data, face challenges in vegetated areas and are prone to inaccuracies due to cloud cover affecting ancillary data. In response to these challenges, we introduced a novel approach termed the vegetation-temperature-radiation-composite (VTRC) method. This method recognizes the robust interaction between vegetation and SM and also appreciates the high sensitivity of surface temperature variation to surface net solar radiation (SSR) in relation to SM. Besides, the VTRC method combines ERA5 data with its high temporal resolution and MODIS data to counteract data loss due to cloud coverage. Applied to downscale the ESA CCI SM products from 0.25 degrees to 0.01 degrees spatial resolution, the VTRC method is validated over Castilla y Leon (Spain) and Anhui province (China). The VTRC method demonstrates a significant improvement compared to the feature space-based downscaled method and the original ESA CCI products, achieving higher correlation coefficient ( $R$ ) of 0.56 and 0.66 $\text{m}<^>{3}/\text{m}<^>{3}$ in humid and semi-arid regions with ground observations, respectively. Furthermore, it maintains a consistent unbiased root mean square error (ub RMSE) of 0.05 $\text{m}<^>{3}/\text{m}<^>{3}$ in both regions. Additionally, the inclusion of vegetation information notably improves the accuracy in comparison to solely using land surface temperature (LST) and SSR. Significantly, evaluations across varying vegetation cover surfaces revealed enhanced accuracy, especially in regions with abundant vegetation.