10:45 〜 11:00
[ACG41-13] Ensemble data assimilation of AMSR2 soil water content into Integrated Land Simulator
キーワード：アンサンブルデータ同化、AMSR2、土壌水分量、陸面データ同化システム
Land surface models (LSMs) are essential tools for hydrological and meteorological predictions. Integrating these models with data assimilation systems is a promising approach to enhance their predictive capabilities. This study presents the development of a land data assimilation (LDA) system based on the Integrated Land Simulator (ILS) coupled with the Local Ensemble Transform Kalman Filter (LETKF). To stabilize data assimilation cycles, we employed a perturbed forcing technique, a common and effective covariance inflation method for LDA. The LDA was evaluated through experiments assimilating satellite-derived soil water content (SWC) observations from Advanced Microwave Scanning Radiometer 2 (AMSR2, Fujii et al. 2009). The resulting analysis states from SWC assimilation were compared against in-situ observations from flux tower sites provided by the FLUXNET2015 (Pastorello et al. 2020) and PLUMBER2 (Ukkola et al. 2022) datasets.
Our experiments revealed that biases between the AMSR2 retrievals and the ILS hampered the direct assimilation of SWC. To mitigate the influence of these biases, we implemented Cumulative Distribution Function (CDF) matching approach to the AMSR2 data prior to data assimilation. The CDF matching successfully prevented the bias-induced degradation in the assimilation cycles, resulting in more accurate SWC estimation compared to open-loop simulations and AMSR2 observations. At the conference, we will present a comprehensive evaluation of the SWC assimilation's impact on a broader range of land surface states, including the latest results.
Our experiments revealed that biases between the AMSR2 retrievals and the ILS hampered the direct assimilation of SWC. To mitigate the influence of these biases, we implemented Cumulative Distribution Function (CDF) matching approach to the AMSR2 data prior to data assimilation. The CDF matching successfully prevented the bias-induced degradation in the assimilation cycles, resulting in more accurate SWC estimation compared to open-loop simulations and AMSR2 observations. At the conference, we will present a comprehensive evaluation of the SWC assimilation's impact on a broader range of land surface states, including the latest results.