Abstract
Soil moisture is an important measure of the exchange of water and energy between the land and the atmosphere. Passive microwave remote sensors onboard Earth observation satellites have served as the most promising tool for quantifying the water content stored in the surface soil layer at a quasi-global scale. In recent decades, the utilization of satellite-based soil moisture retrievals has benefited a variety of applications, including the detection of extreme climate events, water resource and irrigation management, and numerical weather predictions. However, the available passive microwave soil moisture datasets are still unable to entirely satisfy the needs of climatological studies and applications due to the insufficient retrieval quality over certain land surface conditions, the coarse spatiotemporal resolution, the absence of information associated with nonuniform vertical moisture gradients, and the unavailability of a consistent long-term satellite-based soil moisture data record. Given these limitations, this dissertation aims to enhance the quality and utility of current passive microwave-based surface soil moisture data by attempting to resolve the above-mentioned issues.
This dissertation first outlines the analyses for soil moisture retrievals that are impacted by water bodies (i.e., lakes) and soil organic matter to provide clues for refining the operational algorithm of deriving soil moisture from observed brightness temperatures at L-band (1.41 GHz). Specifically, this dissertation identified the lake mix-layer temperature from ERA5 Land and the dielectric mixing model of Mironov 2019 as the preferred options to mitigate water contamination and the effects of SOM in the passive microwave remote sensing retrieval of soil moisture. Their utilization will greatly improve the accuracy of the next-generation L-band soil moisture dataset.
Subsequently, the thesis describes the study that fills temporal gaps in Soil Moisture Active Passive (SMAP) data over the conterminous United States (CONUS) by incorporating a satellite-based precipitation product and a data-driven loss function approach. Validation of this SMAP-based 12-hourly soil moisture product not only exhibited great accuracy but also successfully captured most soil moisture peaks caused by heavy rainfall. The proposed loss-function approach could quantitatively characterize local-scale hydrologic losses near the land surface and can be used for back-extension and forecasts of soil moisture estimates through the incorporation of precipitation measurements.
After that, a global-scale comparison of three advanced satellite-based products was conducted to identify their relative strengths in capturing temporal variability of regional-scale soil moisture. As a result, a global complementarity of the areas was observed where each satellite-based soil moisture product showed its respective advantage in capturing soil moisture variations. Such an evaluation can serve as a guideline for data users to select proper soil moisture datasets for their research and applications. In the appendix (p.141-164), the formulations of a layered radiative transfer model have been presented, which inversions can be used to infer vertically heterogeneous moisture profiles from passive microwave observations.
In summary, this dissertation is dedicated to improving the scientific quality and utility of state-of-the-art SMAP soil moisture retrievals by addressing several identified drawbacks. The outcomes of this dissertation hold great promise for changing how radiometer observations are interpreted in the future while the newly yielded soil moisture data with temporal continuity and higher accuracy will help researchers quantitatively understand the linkages between water balance components and deepen the understanding of the terrestrial-atmosphere interactions in the context of climate change.
