[A-CG46] エミュレータの開発と応用

[ACG46-P04] Exploring the applicability of deep learning regional climate model emulator for Compound Event adaptation during the west Africa summer monsoon passage: Initial approach

*Precious Eromosele Ebiendele¹、Koji Dairaku¹ (1.University of Tsukuba)

キーワード：Deep Learning, Downscaling, RCM-EMULATOR, Compound Event

Providing high resolution climate information is needed for climate adaptation strategies over West Africa. A novel approach which combines both dynamical and statistical downscaling can offer significant added value in quantifying the current increasing Compound drought and heatwave events severity during the west africa monsoon season. In our study, we established a preliminary benchmark for using a deep learning RCM-emulator which is trained using large-scale monsoon flow sensitivity inputs to simulate West African summer monsoon precipitation. Our findings show that our unique deep learning RCM-emulator can learn the optimal GCM to RCM downscaling function. Furthermore, the RCM-emulator offers a significant computational advantage over an RCM simulation. Conclusively, in terms of computing efficiency and gain, we believe that deep learning RCM-emulators can be can be applied to produce high resolution regional scale climate projections of several compound event drivers.