Sensitivity of municipal heatwave vulnerability assessment to population density: Urban-rural spatial shifts in South Korea

As heatwaves become an increasingly recurrent hazard, accurate assessment of local heatwave vulnerability has become increasingly important. This study examined the sensitivity of the heatwave vulnerability index to the inclusion of population density in terms of entropy weights, relative vulnerability rankings, and spatial clustering across 251 municipalities in South Korea in 2024. Candidate indicators were compiled within the exposure-sensitivity-adaptive capacity framework, and Spearman rank correlation analysis was used to examine their empirical associations with absolute counts of heat-related illness and inter-indicator redundancy. Sensitivity to the definition of the health outcome was also assessed using the crude rate of heat-related illness per 100,000 population. Comparison of HVI1, which included population density, with HVI2, in which population density was excluded and the remaining entropy weights were proportionally renormalized, showed that population density received the highest weight in HVI1(0.44293). Among municipalities classified into Jenks classes 8–10, 49 of 61 municipalities (80.33%) under HVI1 were located in special and metropolitan cities, compared with 32 of 70 municipalities (45.71%) under HVI2. LISA analysis also showed changes in High-High clusters, with some small and medium sized cities and rural municipalities included under HVI2. Because no independent external validation was performed, these findings should be interpreted as evidence of the sensitivity of vulnerability assessment results to index specification rather than as evidence that HVI2 is relatively more accurate. Accordingly, when heatwave vulnerability assessments are used to support policy, index-sensitivity checks and follow-up local assessments that account for urban and rural contexts should be considered. Practical implications Population-based information, such as population density and absolute population size, has long played an important role in assessing spatial variations in climate vulnerability and setting policy priorities for heatwave adaptation ( Cortekar et al., 2016 ). Although this approach may support the efficient allocation of limited resources, assigning high weights to population-based indicators such as population density can alter relative vulnerability assessments between high-density urban areas and low-density rural areas depending on the index specification ( Liang and Kosatsky, 2020 ). Using South Korea in 2024 as a case study, this study found that the inclusion or exclusion of population density can alter the weighting structure of the heatwave vulnerability index, relative municipal rankings, and the geographic composition of spatial clusters. These findings provide a diagnostic basis for examining the sensitivity of vulnerability assessments to urbanization proxy indicators such as population density. From a policy and decision-making perspective, sensitivity analysis of population-density inclusion can serve as complementary diagnostic information for assessing how population-based indicators affect relative vulnerability rankings and spatial priorities and for identifying municipalities that may warrant further local-level assessment ( Vaughan and Dessai, 2014 ). In particular, low-density areas that attain relatively high rankings or are classified within high-vulnerability clusters under alternative index specifications may be considered for subsequent field-based assessment. However, because this study did not directly evaluate the effectiveness or equity of resource allocation based on HVI 2 , actual allocation of adaptation resources should also consider independent evidence on health outcomes, healthcare accessibility, local policy need, and implementation feasibility. Meanwhile, integrating new climate information into operational practice can involve a range of administrative challenges, making it important to consider barriers that may hinder its effective use. Such barriers include institutional inertia associated with conventional cost-benefit-oriented budget allocation and the limited availability of microclimate data in rural areas ( Soares et al., 2018 ). This case study demonstrates that municipalities requiring further local assessment can be identified through index-sensitivity checks based on existing public data, without necessarily establishing new monitoring networks or generating new datasets. Such index-sensitivity checks may therefore provide a practical preliminary diagnostic tool in settings where high-resolution microclimate data are limited.