Unveiling teleconnection drivers for heatwave prediction in South Korea using explainable artificial intelligence

Increasing heatwave intensity and mortality demand timely and accurate heatwave prediction. The present study focused on teleconnection, the influence of distant land and ocean variability on local weather events, to drive long-term heatwave predictions. The complexity of teleconnection poses challenges for physical-based prediction models. In this study, we employed a machine learning model and explainable artificial intelligence to identify the teleconnection drivers for heatwaves in South Korea. Drivers were selected based on their statistical significance with annual heatwave frequency (|R| > 0.3, p < 0.05). Our analysis revealed that two snow depth (SD) variabilities—a decrease in the Gobi Desert and increase in the Tianshan Mountains—are the most important and predictable teleconnection drivers. These drivers exhibit a high correlation with summer climate teleconnection triggers along with emerging regional variabilities to emerge as potential drivers.

Yeonsu Lee, Dongjin Cho, Jungho Im, Cheolhee Yoo, Joonele Lee, Yoo-Geun Ham, Myong-In Lee