Presentation Information
[AAS15-P07]A Global Heatwave Prediction System for Regional Hotspots based on Dynamic-Statistic Hybrid Approach★Invited Papers
*GYU-RI LEE1, KI-YOUNG KIM1, MIN-SEOK KIM1, JEE-HOON JEONG1 (1.Sejong University)
Keywords:
Heatwaves,Atmospheric Teleconnection,Hybrid Dynamical-Statistical Approach,Seasonal Heatwave Prediction
Driven by accelerating global warming, extreme heatwaves have increased not only in frequency and cumulative intensity but also through a pronounced tail-widening of temperature distributions, where record-breaking extremes intensify faster than regional mean warming. This study proposes a hybrid global heatwave prediction framework that integrates large-scale atmospheric circulation information from dynamical models with statistically constrained regional predictors.
Using ERA5 reanalysis, we identify global heatwave hotspots exhibiting robust upward trends in extreme heat occurrence and intensity. For each hotspot, key predictors are selected based on their physical linkage to regional heatwaves, including large-scale atmospheric teleconnection patterns, sea surface temperature anomalies, and land-atmosphere feedback indicators. These predictors are combined within a regression-based framework, supplemented by machine-learning models in regions where nonlinear processes dominate, to estimate regional heatwave probability and risk conditioned on future circulation states.
The proposed system is designed to flexibly incorporate region-specific circulation modes—such as zonal and meridional wave patterns over East Asia or Eurasian jet variability—allowing consistent application across diverse climate regimes. By explicitly linking dynamical circulation predictability with regional extreme-temperature responses, this hybrid approach improves the prediction of record-shattering heatwaves that are often underestimated by standalone dynamical models, offering a pathway toward more actionable, region-tailored climate risk assessment and adaptation strategies.
Using ERA5 reanalysis, we identify global heatwave hotspots exhibiting robust upward trends in extreme heat occurrence and intensity. For each hotspot, key predictors are selected based on their physical linkage to regional heatwaves, including large-scale atmospheric teleconnection patterns, sea surface temperature anomalies, and land-atmosphere feedback indicators. These predictors are combined within a regression-based framework, supplemented by machine-learning models in regions where nonlinear processes dominate, to estimate regional heatwave probability and risk conditioned on future circulation states.
The proposed system is designed to flexibly incorporate region-specific circulation modes—such as zonal and meridional wave patterns over East Asia or Eurasian jet variability—allowing consistent application across diverse climate regimes. By explicitly linking dynamical circulation predictability with regional extreme-temperature responses, this hybrid approach improves the prediction of record-shattering heatwaves that are often underestimated by standalone dynamical models, offering a pathway toward more actionable, region-tailored climate risk assessment and adaptation strategies.