Multi-scale satellite-derived soil moisture modeling across space and time
| Author(s) | Llamas, Ricardo M. | |
| Date Accessioned | 2024-01-24T15:06:20Z | |
| Date Available | 2024-01-24T15:06:20Z | |
| Publication Date | 2023 | |
| SWORD Update | 2024-01-22T20:09:37Z | |
| Abstract | Understanding soil moisture processes and their spatial-temporal distribution holds great promise for advancements in water resources management, agriculture, and natural disaster assessment. Soil moisture, a crucial component of available water, influences evaporation evapotranspiration and contributes significantly to the water cycle and terrestrial carbon exchange. ☐ The first study of this work addresses the challenge of obtaining continuous soil moisture measurements over large regions. Geostatistical approaches, namely ordinary kriging (OK), regression kriging (RK), and generalized linear models (GLMs), are explored to model soil moisture, and fill spatial gaps in satellite-derived data. The study supports the use of geostatistical techniques as effective alternatives for gap-filling in satellite-derived soil moisture products. ☐ The second study focuses on downscaling satellite-derived soil moisture using a modular spatial inference framework with terrain parameters. Two modeling methods, Kernel-Weighted K-Nearest Neighbor (KKNN) and Random Forest (RF), are tested for generating fine-resolution soil moisture predictions. The study emphasizes the need for increased monitoring efforts in heterogeneous landscapes and explores the development of modular cyberinfrastructure tools for downscaling. ☐ The third study uses machine learning methods to develop a new fine spatial resolution soil moisture dataset, NASMo-TiAM 250 m. The dataset incorporates satellite-derived soil moisture data, in situ measurements, and static and dynamic ancillary variables to estimate fine-resolution soil moisture. The study showcases promising correlation coefficients and low RMSE values, suggesting the potential of machine learning algorithms for improving soil moisture predictions. ☐ The work’s collective results highlight the need for finer soil moisture information and the potential of remote sensing products to estimate soil moisture distribution. The research contributes alternative approaches to fill spatial gaps and downscale soil moisture estimates from satellite sensors. Moreover, it emphasizes the significance of machine learning algorithms in addressing knowledge gaps and improving soil moisture predictions in challenging regions. As monitoring networks expand and computational capacity improves, satellite soil moisture validation products are expected to see significant improvements, benefiting various ecological studies. Future advancements promise to provide more accurate methods and higher-resolution soil moisture products over extensive areas. | |
| Advisor | Vargas, Rodrigo | |
| Degree | Ph.D. | |
| Department | University of Delaware, Department of Plant and Soil Sciences | |
| Program | University of Delaware, Water Science and Policy Program | |
| DOI | https://doi.org/10.58088/v32f-8802 | |
| Unique Identifier | 1439103978 | |
| URL | https://udspace.udel.edu/handle/19716/33883 | |
| Language | en | |
| Publisher | University of Delaware | |
| URI | https://www.proquest.com/dissertations-theses/multi-scale-satellite-derived-soil-moisture/docview/2917418641/se-2?accountid=10457 | |
| Keywords | Soil moisture | |
| Keywords | Satellite | |
| Keywords | Modeling | |
| Title | Multi-scale satellite-derived soil moisture modeling across space and time | |
| Type | Thesis |