研究課題/領域番号 |
18KK0117
|
研究種目 |
国際共同研究加速基金(国際共同研究強化(B))
|
配分区分 | 基金 |
審査区分 |
中区分22:土木工学およびその関連分野
|
研究機関 | 東京大学 |
研究代表者 |
金 炯俊 東京大学, 生産技術研究所, 特任准教授 (70635218)
|
研究分担者 |
渡部 哲史 京都大学, 防災研究所, 特定准教授 (20633845)
内海 信幸 京都先端科学大学, ナガモリアクチュエータ研究所, 助教 (60594752)
|
研究期間 (年度) |
2018-10-09 – 2024-03-31
|
研究課題ステータス |
交付 (2022年度)
|
配分額 *注記 |
17,810千円 (直接経費: 13,700千円、間接経費: 4,110千円)
2022年度: 4,290千円 (直接経費: 3,300千円、間接経費: 990千円)
2021年度: 3,510千円 (直接経費: 2,700千円、間接経費: 810千円)
2020年度: 3,510千円 (直接経費: 2,700千円、間接経費: 810千円)
2019年度: 3,510千円 (直接経費: 2,700千円、間接経費: 810千円)
2018年度: 2,990千円 (直接経費: 2,300千円、間接経費: 690千円)
|
キーワード | GPM / Multi-task learning / Precipitation / lake surface area / remote sensing / water big data / データ駆動型モデリング / 水文季節予報 / 衛星観測 / テレコネクション / 人工知能 / 衛星高度計 / 河川の水位 / 海水面温度 / 陸域水貯留 / ニューラルネットワーク / 長期リードタイム予測 / data-driven modeling / seasonal prediction / satellite remote sensing |
研究実績の概要 |
Timely prediction of flood and drought greatly minimize the related losses. It requires precise precipitation estimates as a predictor of terrestrial processes across spatiotemporal scales. In this fiscal year, we proposed a novel precipitation retrieval framework in which regression and classification tasks are simultaneously trained using multi-task learning approach. Satellite-based precipitation estimations provide frequent, large-scale measurements. Recently, deep learning has shown significant potential for improving estimation accuracy. In this project, we designed a novel network architecture and loss function to maximize the benefits of multi-task learning. The proposed multi-task (i.e., two-task) model successfully achieved a better performance than the conventional single-task model possibly due to efficient knowledge transfer between tasks. The product intercomparison showed that our product outperformed existing products in rain rate retrieval and also yielded better skills in the rain/no-rain retrieval task.
|
現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
In this fiscal year, we developed a machine learning based satellite precipitation estimation framework and to investigate how different components which consist of the framework, such as preprocessing, network structure, and loss function, contribute to and interplay for the estimation skill. The followings are implementations: 1) to hybridize physics guide concept to the inference framework based on physical system of precipitation and 2) to enhance typical structure of multi-task learning (MTL) to increase the performance. As we planed to exploit various earth observing satellite missions for a data-driven simulation framework, getting a better quality of precipitation data is indispensible. Therefore, the progress in the fiscal year would be a very important part of the entire research in propsed project, and we evaluate the current status is on the plan and it has proceeded smoothly.
|
今後の研究の推進方策 |
This is the final year which has been extended. Becuase of the pandemic situation, we could not have active exchanges among international counterparts during the project years. In next fiscal year, we will have an international workshop on applications of satellite remote sensing for global water cycle, and actively participate international meetings and conference.
|