A downscaling-and-fusion framework for generating spatio-temporally complete and fine resolution remotely sensed surface soil moisture.

• A downscaling-and-fusion framework was proposed for generating complete 1 km SSM. • BME method considering multi-spatial-scale uncertainties was presented for fusion. • The completeness of fine-scale SSM is directly and significantly improved. • The accuracy and variations of downscaled SSM are well preserved into fused SSM. Surface soil moisture (SSM) plays a crucial role in global energy exchanges and serves as a key requirement for drought monitoring and agriculture production. Microwave remote sensing (RS) SSM products at coarse scale (e.g., 0.25°) can hardly meet the requirements for agricultural and meteorological applications, which necessitate seamless fine-scale SSM data. Current methods for generating complete fine-scale SSM often involve initially filling spatial gaps in SSM ancillary variables (e.g., land surface temperature). However, the uncertainties introduced during the filling process are not adequately and quantitatively considered and utilized in subsequent downscaling process. In this study, we developed a downscaling-and-fusion framework to generate spatiotemporally complete fine-scale SSM (1 km) in the central Tibetan Plateau (TP) of the southwestern China. Firstly, the incomplete downscaled SSM was retrieved by disaggregating the bias-corrected European Space Agency's Climate Change Initiative (ESA CCI, 0.25°) SSM based on apparent thermal inertia (ATI), and then the Bayesian maximum entropy (BME) method considering multi-spatial-scale uncertainties was employed to spatiotemporally merge downscaled and coarse-scale CCI SSM for seamless fine-scale SSM. The completeness of valid merged SSM pixels and pixels unable to retrieved SSM sums up to over 99.9 %. Comparisons with in situ SSM measurements show that merged SSM with more samples have comparable overall performance and slightly better station-level performance than the downscaled SSM with less samples. Station-level statistical metrics averages are superior to those of the overall evaluation for the merged SSM, especially in terms of R (0.592 versus 0.441) and ubRMSE (0.083 versus 0.115 m3/m3). The merged SSM can effectively reflect temporal variability of the in situ SSM measurements, accurately capture the changes of SSM caused by the precipitation events, and preserve the fine-spatial-scale variations of the downscaled SSM. These results demonstrate the success of the proposed downscaling-and-fusion framework in generating spatiotemporally complete fine-scale SSM. [ABSTRACT FROM AUTHOR]