日本地球惑星科学連合2021年大会

講演情報

[E] ポスター発表

セッション記号 A (大気水圏科学) » A-CG 大気海洋・環境科学複合領域・一般

[A-CG36] 衛星による地球環境観測

2021年6月3日(木) 17:15 〜 18:30 Ch.06

コンビーナ:沖 理子(宇宙航空研究開発機構)、本多 嘉明(千葉大学環境リモートセンシング研究センター)、高薮 縁(東京大学 大気海洋研究所)、松永 恒雄(国立環境研究所地球環境研究センター/衛星観測センター)

17:15 〜 18:30

[ACG36-P07] Evaluation of Himawari estimated precipitation product compared with GSMaP and JMA radar analysis

*豊嶋 紘一1、広瀬 民志2、樋口 篤志1 (1.千葉大学 環境リモートセンシング研究センター、2.国立研究開発法人 宇宙航空研究開発機構)

キーワード:衛星降水観測、静止衛星

The Global Satellite Mapping of Precipitation (GSMaP) uses geostationary satellite infrared band data to estimate precipitation to fill the gap in microwave radiometer (MWR) observations. However, the precipitation estimation that depends only on the cloud top temperature (altitude) information shows a remarkable tendency to overestimate precipitation around the anvil cloud in the tropical region, and the accuracy of the infrared (IR) area is lower than that of MWR. Himawari-8 has IR multi-band observations available in the Asian monsoon region. A product that can estimate rainfall with higher accuracy than the conventional method using only a single IR band observation using random forest machine learning using the Global Precipitation Measurement (GPM) Mission Ku band precipitation radar (PR) observations (Hirose et al. 2019) was created. To evaluate this product, we created annual Himawari Rainfall Analysis (HRA) products for 2019 and compared them with GSMaP MVK version 7. The satellite information flag of GSMaP was used to distinguish the observation area of MWR. Since HRA adjusts the hyperparameters of the machine learning model based on the root mean square error (RMSE), it has the problem that the estimation accuracy decreases for extreme rainfall, which is rarely observed. To solve this problem, we introduced a method to correct the HRA observational rainfall intensity histogram based on the MWR observational rainfall intensity histogram and compared the results before and after the correction was applied.
In the product applied with histogram correction, the part that tended to be excessive in the tropical ocean area became smaller, but the excessive amount in the coastal area became conspicuous. A one-year long-term analysis revealed that the histogram correction improved the intensity distribution to a level comparable to that of MWR. However, focusing closely on land and sea, it did not match. It does not have a real rainfall intensity distribution. As a result of the histogram correction, the excessive tendency of the tropical ocean area and the coastal area's excessive tendency were corrected.
In addition, the product is compared with the Japan Meteorological Agency Radar (JMARAD) over the Japan area. Rainfall accumulation is compared for each rainfall event period. The precipitation distributions of frontal precipitation in the mid-latitudes and during the rainy season are similar to those of JMARAD, and the differences in the accumulation values are small. Although the reproducibility of the partial accumulation of localized rainfall areas during the occurrence of a line-shaped rainfall system is low, the precipitation distribution is relatively consistent. We focused on the precipitation accumulation during typhoons approaching and crossing Japan, and the rainfall distribution is consistent between them. However, the rainfall tends to be underestimated in the rainfall associated with the unique topography of Shikoku and the southern coast of the Kii Peninsula. In the case of the Japan Sea snowfall cases, it is not easy to find the characteristics of the snowfall streaks as seen in JMARAD. On the other hand, we find the precipitation accompanied by rapidly developing of isolated cumulonimbus clouds unaffected by radar shielding due to topography in the summer season because of the high temporal resolution of Himawari-8 observation.