10:00 〜 10:20
[D1-04] A machine learning-based Improved soil moisture estimation: Synergistic use of satellite observations and land surface models over CONUS
キーワード：soil moisture, machine learning, satellite, land surface model
This study aimed to enhance daily soil moisture (SM) data for the contiguous United States (CONUS) by combining three commonly used SM data sources: in-situ measurements, satellite observations, and land surface models (LSMs). Machine learning (ML) techniques were employed, along with simple averaging ensemble approaches, to improve the quality of the data. Three different schemes were tested for each ML model, using satellite-derived variables, LSM-derived variables, or a combination of both as independent variables. The proposed approach was evaluated using various ways with in-situ SM datasets. The evaluation demonstrated that the ML-based ensemble method consistently outperformed existing SM products, particularly in regions with complex topography and dense vegetation. Moreover, our results showed that the ML ensemble approach performed better than other SM datasets, resulting in a higher correlation coefficient and smaller random and systematical error Additionally, we confirmed that the ML-based ensemble exhibited better spatiotemporal quality compared to other SM products.