[A-CG36] 衛星による地球環境観測

[ACG36-16] Improving Precipitation Prediction by Data Assimilation of GPM and Other Satellite Observations

*三好 建正¹、小槻 峻司^2,1、寺崎 康児¹、大塚 成徳¹、Wu Ting-Chi¹、富田 浩文¹、Chen Ying-Wen³、金丸 佳矢⁴、佐藤 正樹³、八代 尚⁵、近藤 圭一⁶、岡本 幸三⁶、Kalnay Eugenia⁷、久保田 拓志⁸ (1.理化学研究所、2.千葉大学、3.東京大学、4.情報通信研究機構、5.国立環境研究所、6.気象研究所、7.メリーランド大学、8.宇宙航空研究開発機構)

キーワード：数値天気予報、データ同化、衛星観測、全球降水観測、二周波降水レーダ、降水観測ミッション

This presentation summarizes the recent progress of the project started in 2013 to explore data assimilation methods for GPM and other satellite observations. The details of the project are described in the latter part of this abstract. The achievements are highlighted by successful data assimilation of GPM DPR reflectivity and a new development of an efficient data-driven (DD) method for a satellite simulator a.k.a. an observation operator in data assimilation. By assimilating DPR reflectivity, we estimated a model microphysics parameter corresponding to snowfall terminal velocity and successfully reduced the gap between the model-produced and observed CFAD (Contoured Frequency by Altitude Diagram). The results showed improvements in radiation budgets (OSR and OLR biases) and overall numerical weather prediction skill. As for the DD method for a satellite simulator, we developed a new approach using neural networks to simulate satellite microwave radiances without a need for a bias correction treatment. We applied machine leaning with model forecast data and corresponding actual satellite observations and built a bias aware simulator for satellite radiances. The results showed that the satellite simulator worked properly although slightly worse than the case with a radiative transfer model and bias correction. The early results are encouraging since we do not need a bias correction method to build a generally complex system to assimilate satellite radiance data.

In precipitation science, satellite data have been providing precious, fundamental information, while numerical models have been playing an equally important role. Data assimilation integrates the numerical models and real-world data and brings synergy. We have been working on assimilating the GPM data into the Nonhydrostatic ICosahedral Atmospheric Model (NICAM) using the Local Ensemble Transform Kalman Filter (LETKF). We continue our effort on “Enhancing Precipitation Prediction Algorithm by Data Assimilation of GPM Observations” funded by JAXA, following successful completion of the 3-year project titled “Enhancing Data Assimilation of GPM Observations” from April 2016 to March 2019. The project first started in April 2013 on “Ensemble-based Data Assimilation of TRMM/GPM Precipitation Measurements”, where we developed a global data assimilation system NICAM-LETKF from scratch. This presentation highlights the most recent achievements.