Seasonal Predictability of Summer Hazards Over the Korean Peninsula and Implications for Hazard Response

Abstract

Summer heat and monsoon rainfall in Korea are closely linked to the phase of quasi-stationary Rossby waves along the meandering upper-level jet. This study evaluates the seasonal predictability of individual summer hazards and examines response strategies for hazards that remain beyond the reach of seasonal forecasting. The upper-level ridge phase is diagnosed using hot-minus-cool composites and Takaya-Nakamura wave-activity flux. A pre-season probabilistic forecast is then developed from slow spring boundary forcing, with cross-validation designed to exclude information from the verification year. Four hazard indices—heat wave days, summer rainfall, SPI drought, and soil moisture—are each decomposed into a predictable component and a synoptic-scale residual. Validation with ERA5 (1979-2024) shows that significant skill is confined to the upper-level circulation and seasonal-mean temperature (Western Pacific Subtropical High r = 0.76; mean temperature r = 0.43). The four hazard indices exhibit very weak correlations with the predictable component (|r| < 0.05) and are strongly correlated only with the synoptic-scale residual. Although the number of heatwave days is strongly correlated with seasonal-mean temperature (0.79), the seasonal predictability of the mean temperature does not translate into predictive skill for heatwave days. All four hazard indices ae significantly associated with the same synoptic-scale residual, suggesting that heat and dryness can co-exist in years that are not predictable in advance. Seasonal forecasting is therefore best used within the limits of a probabilistic tilt of the seasonal mean, while the four hazard indices should be addressed through short-range forecasting and real-time monitoring. Response systems should be reviewed to accommodate the range of variability observed in the record.