DescriptionSCAT backscatter and slope contain valuable information about vegetation water dynamics (Steele-Dunne et al., 2012). Previous studies developed a machine-learning based observation operator (Shan et al., 2022). The physically plausible estimates thus link the ASCAT observables to the land surface model (LSM) ISBA-A-gs. In this study, we embedded the observation operator into Simplified Extended Kalman filter (SEKF), and assimilated the ASCAT backscatter and slope into ISBA-A-gs. The data assimilation (DA) system was applied on ISMN stations in Europe from 2017 to 2019. analysis of DA diagnostics (e.g., normalized innovations) shows the importance of correctly representing realistic model and observation error.
The DA system was also evaluated against in-situ soil moisture observations from ISMN and satellite-based LAI observations from 1km v2 Copernicus Global Land Service project (CGLS) product products. Different experiments show the differences about only assimilating backscatter or slope, or two together. In terms of unbiased RMSE, only assimilating backscatter leads to a worse estimate of LAI than open loop (OL) experiments, while only assimilating slope helps to improve LAI. Results show that assimilating slope also helps improve soil moisture in deeper layers, especially over stations of Broadleaf forests. On stations of agricultural crops, assimilating slope creates a drier estimate of root zone soil moisture, improving the performance of deeper soil moisture. Future work is needed to explore the optimal values of observation uncertainty in the DA system.
|Period||12 Dec 2022|
|Event title||AGU Fall Meeting 2022|
|Location||Chicago, United States|
|Degree of Recognition||International|