Summer heat has been changing over the long term in response to climate change, raising concerns that patterns of heat-related mortality may also evolve over time. These changes involve both increases in mean temperature levels and greater short-term ...
Summer heat has been changing over the long term in response to climate change, raising concerns that patterns of heat-related mortality may also evolve over time. These changes involve both increases in mean temperature levels and greater short-term temperature variability. However, most previous studies have focused primarily on average heat exposure, while long-term changes in mortality sensitivity to short-term temperature fluctuations have received less attention. In addition, differences in local climate distributions, exposure metrics, and threshold definitions across cities have led to substantial heterogeneity in risk estimates, limiting comparability across studies and their policy relevance. This dissertation examines long-term changes in summer heat–related mortality risk in major Korean cities along two dimensions—average risk levels and sensitivity to short-term temperature variability—and evaluates factors that constrain cross-city comparability under a common threshold framework, as well as the suitability of alternative heat exposure indicators.
In Chapter 1, distributed lag non-linear models (DLNMs) were used to estimate minimum mortality temperature (MMT) and mortality sensitivity to short-term temperature variability (TV total), adjusting for mean heat exposure, across cities and time periods. The results showed that MMT tended to shift over time, although the magnitude and statistical support of these changes varied across cities. In contrast, mortality sensitivity to temperature variability did not exhibit a consistent declining trend, and increasing trends were observed in some cities. In second-stage meta-regression analyses, increases in mean WBGT within cities were strongly associated with higher MMT (+1.52°C per 1 SD increase; 95% CI: 1.12–1.92). Population aging, however, was negatively associated with MMT (−1.60°C per 1 SD increase; 95% CI: −2.60 to −0.60), indicating constraints on upward shifts in the minimum-risk temperature, while simultaneously showing a positive association with TV-related mortality sensitivity. These findings indicate that changes in average heat-related risk and vulnerability to short-term temperature variability do not follow a single adaptation pathway and therefore require joint evaluation.
Chapter 2 compared absolute temperature-based indicators with an anomaly-based WBGT indicator (ΔWBGT), defined as deviations from a short-term baseline. Spatial heterogeneity in MMT across cities was substantially lower for ΔWBGT (standard deviation: 0.26) compared with maximum temperature (1.30°C) and maximum WBGT (0.89). When a single national threshold based on the mean city-specific MMT was applied, ΔWBGT also resulted in the fewest misclassified warning days (late: 401 days; early: 357 days) relative to other indicators. These results suggest that ΔWBGT improves cross-city alignment and interpretability under a common threshold in settings with pronounced climatic heterogeneity.
Overall, this dissertation demonstrates the need to assess long-term heat adaptation by jointly considering changes in average risk levels (MMT) and sensitivity to short-term temperature variability, and shows that social factors may influence these two dimensions in different ways. In addition, the findings suggest that anomaly-based heat indicators can serve as practical alternatives for national-scale heat–health risk assessment and heat warning system operations.