Misperception of Extreme Weather Event Mortality Risk in the United States Authors: Mitchell Manware, Lingzhi Chu, Ying Hu, Jennifer Marlon, Anthony Leiserowitz, Kai Chen Published: 2026 ABSTRACT AbstractCommunicating the risks posed by extreme weather events remains a challenge for researchers and policy makers. This study evaluates the gap between public perceptions on climate risk and scientific estimates of mortality burden from climate‐related extreme weather events. Results show each mortality risk perception gap follows a similar spatial pattern to the observed attributable mortality rate. The perception that “global warming is happening” was significantly associated with smaller risk perception gaps for all extreme weather events. Improving population‐level understanding of climate change is critical to reducing health risks associated with extreme weather events and to motivate climate‐oriented actions. Introduction Recent epidemiological studies have quantified mortality rates associated with various extreme weather events in the US. Specifically, county‐level excess mortality rates have been estimated for wildfire smoke (Ma et al., 2024), temperature (heat and cold) (Chu, Dubrow et al., 2025), floods (Chu, Warren et al., 2025), and droughts (Hu et al., 2025) in the contiguous US, improving the evidence available for public health and policy decisions. Despite these scientific advances, effectively conveying the health risks posed by extreme weather remains a challenge for researchers and policy makers (Howe et al., 2015). Insufficient or ineffective communication on the risk of morbidity or mortality arising from climate‐related hazards can lead to low risk perceptions among the public, driving inappropriate adaptation measures and potentially increasing vulnerability (Howe et al., 2015). According to the Yale Program on Climate Change Communication's (YPCCC) 2024 Climate Opinion Maps, only 47% of American adults believe that “global warming will harm them personally,” which varies across counties (Howe et al., 2015). This study aims to evaluate the gap between public perceptions on climate risk and scientific estimates of mortality burden from climate‐related extreme weather events. Methods We obtained risk perception data from the YPCCC 2024 Climate Opinion Maps (Howe et al., 2015; Marlon et al., 2025). YPCCC methodology uses multilevel regression and poststratification to estimate the percentage of adults who hold a certain climate‐related opinion in each US county (Howe et al., 2015; Marlon et al., 2025). The models are based on 31 nationally representative surveys, collectively including 35,000 respondents, and demographic and political covariates, and thus are not direct observations (Howe et al., 2015; Marlon et al., 2025). The county‐specifc percentage of adults who think that “global warming will harm them personally a moderate amount or a great deal” variable was used to quantify climate risk perception (hereafter referred “climate risk perception”) (Howe et al., 2015). Attributable mortality rates from wildfire smoke (Ma et al., 2024), heat (Chu, Dubrow et al., 2025), flood (Chu, Warren et al., 2025), and drought (Hu et al., 2025) were obtained from previous studies. The total extreme weather event‐attributable mortality rate was calculated by summing these individual rate estimates. To compare climate risk perception, a percentage, to actual climate‐related burden, a rate, the county‐level data were rank transformed. For risk perception, 1 represented the highest percentage and 3,094 represented the lowest. Similarly for extreme weather event attributable mortality rates, 1 represented the highest rate per 100,000 people and 3,094 represented the lowest. For heat‐related burden rank, counties with missing data (n = 2,035) were not counted in the ranking (Table 1). We quantified county‐specific gap between climate risk perception and actual climate‐related burden (hereafter “mortality risk perception gap”) by subtracting the attributable mortality rank from the risk perception rank for each extreme weather event. Table 1Mortality Risk Perception Gap Summary Statistics by Extreme Weather Event TypeNumber of included countiesMortality risk perception gapChanges in the mortality risk perception gapMean (SD)Minimum to maximum“Happening” term changing from 20th–80th percentile (95% CI)“Happening” term changing from 10th–90th percentile (95% CI)All Extreme Weather Events3,106−6.8 (1,287.5)−3,035–2,950−1,213.5 (−1,305.4, −1,121.6)−1,580.9 (−1,696.8, −1,465.0)Wildfire Smoke3,106−10.2 (1,464.7)−3,089–3,052−1,294.3 (−1,341.4, −1,247.3)−1,838.2 (−1,897.6, −1,778.7)Heat1,071−0.3 (469.6)−1,049–1,030−410.3 (−457.0, −363.7)−603.9 (−660.0, −547.8)Flood3,106−9.9 (1,288.5)−3,068–3,045−988.8 (−1,054.4, −923.1)−1,398.9 (−1,484.3, −1,313.4)Drought3,106−8.1 (1,152.5)−3,037–3,002−1,027.5 (−1,098.1, −957.0)−1,401.9 (−1,491.1, −1,312.6) Note. The “Happening” Term 20th–80th and 10th–90th percentile values represent the difference in mortality risk perception gap value between counties with a high or low percentage of adults who think that global warming is happening, estimated from the fitted generalized linear regression model. Negative value indicates the mortality risk perception gap decreases as percentage of adults who think that climate change is happening increases. SD is “standard deviation.” CI is “confidence interval.” Mortality Risk Perception Gap Summary Statistics by Extreme Weather Event Type Note. The “Happening” Term 20th–80th and 10th–90th percentile values represent the difference in mortality risk perception gap value between counties with a high or low percentage of adults who think that global warming is happening, estimated from the fitted generalized linear regression model. Negative value indicates the mortality risk perception gap decreases as percentage of adults who think that climate change is happening increases. SD is “standard deviation.” CI is “confidence interval.” To assess the relationship between the mortality risk perception gap and public understanding that climate change is happening, we then fitted generalized linear regression of the mortality risk perception gap on county‐level percentages of understanding that global warming is happening (“happening”) (Howe et al., 2015). Models adjusted for median age (“age”) (U.S. Census Bureau, 2022c), median household income in the past 12 months (“income”) (U.S. Census Bureau, 2022d), and the percentage of non‐non‐Hispanic White residents within the county (“demographic”) (U.S. Census Bureau, 2022b), with a 12‐year average calculated for 2009 to 2020 using annual estimates, and location using a thin plate spline for population‐weighted county centroid. The percentage of population with less than a completed bachelor's degree (“education”) (U.S. Census Bureau, 2022a) was also included as a covariate, but a 9‐year average for 2012 to 2020 was used due to unavailable data prior to 2012. Covariate data were accessed using the tidycensus package for R (Walker & Herman, 2025). Some demographic covariates included in our regression models overlap with the YPCCC predictors used to generate the opinion estimates, introducing potential non‐independence. Results The county‐level percentages of adults who think that global warming will harm them personally in 2024 ranged from 28.5% to 60.5% with a mean of 37.7% (standard deviation = 5.1%). Mortality risk perception gaps for each of the extreme weather events vary, with each spanning positive and negative values and small and large magnitudes (Table 1). Positive values indicated an under‐awareness of health burden related to each extreme weather event. Notably, each risk perception gap follows a similar spatial pattern to the observed attributable mortality rate (Figure 1). We observed larger under‐awareness of climate risk relative to health burden from wildfire smoke and flood in the midwestern and northern US, and for drought in the southern and western regions. Extreme weather‐related mortality rates and mortality risk perception gaps for United States counties. Left column depicts county‐level mortality rate per 100,000 people summed over all extreme weather events and by event type. Right column depicts county‐level mortality risk perception gap for all extreme weather events as a whole and by event type. Positive values (orange‐red shading) indicate “under awareness” of mortality risk, while negative values (blue shading) indicate “over awareness.” Values near zero (yellow shading) indicate county‐level mortality risk rank and risk perception rank are similar. Diagnostic tests for models with all covariates showed multicollinearity. Individual terms were dropped until model performance was maximized without multicollinearity between covariate terms, with final models retaining terms for spatial smooth, “happening”, and “income.” The all extreme weather events model explained 66.6% of total deviance (R 2 = 0.64), indicating covariate terms are strong fit to the mortality risk perception gap. Event‐specific models also reported similar deviance explained and R2. The percentage of adults who think that global warming is happening was significantly associated with mortality risk perception gaps (p < 0.001) and appeared nonlinear in all models (Figure 2; Figure 3). Association between percentage of population who think that global warming is happening and the mortality risk perception gap. X‐axis depicts the percentage of population who thinks that global warming is happening. Y‐axis depicts the changes in the mortality risk perception gap, holding other covariates constant. Negative slope indicates that counties with a higher percentage of population who think that global warming is happening have a lower mortality risk perception gap for each of the extreme weather events. Change in extreme weather event‐specific mortality risk perception gaps associated with percentile change in the percentage of population who think that global warming is happening. Y‐axis depicts difference in mortality risk perception gap between 20th‐80th or 10th‐90th percentiles of population who think that global warming is happening, respectively. Negative values indicate that as percentage of population who think that global warming is happening increases, the mortality risk perception gap decreases for all extreme weather events. Discussion We found that mortality risks from extreme weather events, in total and in isolation, are not well understood in most counties in the contiguous US. The percentage of a county's population who think “global warming is happening” was significantly associated with smaller risk perception gaps, indicating a better understanding of personal mortality risk for all climate‐related extreme weather events. However, the risk perception data used in this analysis are modeled. population‐level, and describe general climate opinions, which contain unpropagated uncertainty and potential non‐independence. We also note that ranking both opinion estimates and mortality burdens may magnify small differences and increase sensitivity to model uncertainty. Because we did not propagate the full posterior uncertainty of the YCOM estimates, effect sizes and significance values should be interpreted with appropriate caution. Our findings should thus be viewed as describing broad spatial and conceptual patterns rather than providing precise county‐level estimates. Finally, risk perception gaps were calculated using excess mortality rates, which do not consider morbidity, economic, or social losses associated with extreme weather events. With these considerations in mind, the results indicate that improving population‐level understanding of climate change is critical to reducing health risks associated with extreme weather events and to motivate climate‐oriented actions. Conflict of Interest The authors declare no conflicts of interest relevant to this study. Acknowledgments None. Availability Statement The county‐level extreme weather‐related mortality estimates are publicly available at: https://yalemaps.maps.arcgis.com/apps/dashboards/25149f6ab8014cb1bac7eb87934cff6b. The Yale Climate Opinion Maps 2024 estimates are publicly available at https://climatecommunication.yale.edu/visualizations‐data/ycom‐us/. References Associated Data Data Availability Statement The county‐level extreme weather‐related mortality estimates are publicly available at: https://yalemaps.maps.arcgis.com/apps/dashboards/25149f6ab8014cb1bac7eb87934cff6b. The Yale Climate Opinion Maps 2024 estimates are publicly available at https://climatecommunication.yale.edu/visualizations‐data/ycom‐us/.