講演情報
[PEM11-P09]Development of a forecast model of the outer radiation belt electrons using the XAI
*西宮 祐太1、三好 由純1、堀 智昭1、Jun Chae-Woo1、三谷 烈史2、篠原 育2、高島 健2、浅村 和史2、東尾 奈々2、齊藤 慎司3、塩田 大幸3 (1.名古屋大学宇宙地球環境研究所、2.宇宙航空研究開発機構、3.情報通信研究機構)
キーワード：
放射線帯、宇宙天気、XAI、機械学習
The radiation belt is a region in the inner magnetosphere where the most energetic electrons in geospace are trapped by the Earth's magnetic field. Large flux variations of energetic electrons are observed in association with magnetic storms, and a prolonged large flux enhancement of the outer belt electrons often leads to satellite anomalies. Forecasting flux variations for energetic electrons is therefore essential in mitigating these risks and is one of the most important aspects of space weather. We have developed a forecast model of the outer belt electron flux variation using a recurrent neural network (RNN) with long short-term memory (LSTM). As inputs for the developed model, we used electron flux observed by the HEP and XEP onboard the Arase satellite, along with solar wind parameters. Moreover, we have also incorporated eXplainable Artificial Intelligence (XAI) into our model to investigate the relative contributions of the input parameters that contribute to electron flux variations. The diagnosis using the XAI technique indicates that both solar wind speed and the time-integrated southward IMF contribute to flux enhancement, while an increase in solar wind density increase contributes to the loss of electron flux.