Uncertainty quantification and spatial downscaling for passive microwave remote sensing of soil moisture Jingyao Zheng, Tianjie Zhao
State Key Laboratory of Remote Sensing and Digital Earth, Aerospace Information Research Institute, Chinese Academy of Sciences
Abstract
High-resolution soil moisture (SM) data are essential for hydrological and climate applications but limited by the coarse resolution of passive microwave SM products. This study conducted the first systematic evaluation of 24 global SM products using ground networks. SMAP passive microwave SM demonstrated superior accuracy (ubRMSD < 0.04m^{3}/m^{3}), identifying it as the optimal input for downscaling. Key error sources include vegetation parameter uncertainty (causing dry bias) and limitations in existing freeze/thaw products (leading to over-filtering of valid SM retrievals). Validation method analysis revealed Triple Collocation (TCA) tends to underestimate errors, while Categorical TCA (CTC) generally ranks products correctly but is vulnerable when products covary. We rigorously assessed combinations of downscaling factors and methods. The NSDSI-2 factor (based on soil reflectance) combined with a Taylor series expansion method yielded the highest accuracy downscaled SM. Furthermore, we demonstrated that the evapotranspiration-based DISPATCH method enhances SM representativeness in humid zones specifically under conditions of sufficient evaporative demand (e.g., summer/dry pixels), explaining its limitations in energy-limited regions.Addressing the critical data gap, we produced a novel 1 km resolution downscaled SM product for the TP (2017-2020) using the optimal NSDSI-2 + Taylor approach. Extensive validation across five diverse TP networks confirmed this new product outperforms the original SMAP and other publicly available downscaled products in spatial/spatiotemporal metrics, despite showing limited improvement in densely vegetated areas.
