Details zur Publikation |
| Kategorie | Textpublikation |
| Referenztyp | Preprints |
| DOI | 10.2139/ssrn.4758452 |
| Titel (primär) | Understanding the differences in various satellite remotely sensed soil moisture downscaling methods |
| Autor | Hao, L.; Wei, Z.; Zhao, T.; Zhong, Y.; Peng, J.
|
| Quelle | SSRN |
| Erscheinungsjahr | 2024 |
| Department | RS |
| Sprache | englisch |
| Topic | T5 Future Landscapes |
| Abstract | Passive microwave remote sensing soil moisture products encounter limitations in regional-scale applications due to their coarse spatial resolution. To address this issue, various soil moisture downscaling methods have been proposed. Visible-thermal infrared remote sensing data, with its rich observations and high resolution, has emerged as an ideal data source to support downscaling methods. However, the suitability of these downscaling methods varies significantly across diverse study areas and input data types. Additional research and discussion are needed regarding the computation and combination of land surface parameters, the selection of regression models, and the scope of applicability of different methods. Therefore, this study constructed 40 downscaling methods using 5 global or local regression models combined with 8 downscaling factor combinations. The Disaggregation based on Physical And Theoretical scale Change (DISPATCH) algorithm was selected for comparison due to its semi-empirical characteristics. The accuracy differences of these downscaling algorithms were comprehensively evaluated using in situ and airborne data from three field campaigns (SMAPEx-5, SMAPVEX16, SMELR). The research findings suggest that across the semi-arid regions of Yanco and Shandian River, the majority of combinations demonstrate high accuracy. Conversely, in Iowa, characterized by a semi-humid climate, the accuracy of most combinations experiences a modest reduction. The introduction of brightness temperature can significantly improve the R and ubRMSE of the downscaling results, but it simultaneously smoothes the spatial details of various downscaling methods. Among these methods, the DISPATCH downscaled results exhibit consistent spatial patterns and rich spatial details. When compared to global linear regression, local linear regression proves to be more accurate but is susceptible to the influence of missing values. Machine learning methods yield more stable downscaling results, with the random forest method demonstrating higher R and lower ubRMSE. Overall, through a comprehensive comparison of various methods, this research provides profound insights into soil moisture downscaling, offering valuable guidance for the selection and application of downscaling methods. |
| dauerhafte UFZ-Verlinkung | https://www.ufz.de/index.php?en=20939&ufzPublicationIdentifier=29771 |
| Hao, L., Wei, Z., Zhao, T., Zhong, Y., Peng, J. (2024): Understanding the differences in various satellite remotely sensed soil moisture downscaling methods SSRN 10.2139/ssrn.4758452 |
|