Misperception of Extreme Weather Event Mortality Risk in the United States
Interactive companion to the commentary in GeoHealth, 10, e2025GH001782 (2026) · doi:10.1029/2025GH001782 · Open access (CC BY-NC)
County-level map
Mortality risk perception gap = climate risk perception rank − attributable mortality rank. Positive values (red) indicate under-awareness of mortality burden; negative values (blue) indicate over-awareness; values near zero (gray) indicate alignment. Hover over a county for details.
Summary statistics by event type
Mortality risk perception gap summary and modeled change in the gap as the percentage of adults who think global warming is happening moves between percentiles (Table 1 of the paper; 95% CIs).
| Event | Counties | Gap mean (SD) | Min to max | Δ gap, 20th→80th pct (95% CI) | Δ gap, 10th→90th pct (95% CI) |
|---|
Belief that global warming is happening → smaller gaps
Modeled change in the mortality risk perception gap when a county's "happening" percentage moves from the 20th to 80th, or 10th to 90th, percentile. Negative values = smaller gap (better risk understanding).
Methods, data, and caveats
Risk perception. County-level percentage of adults who think "global warming will harm them personally a moderate amount or a great deal," from the Yale Program on Climate Change Communication (YPCCC) 2024 Climate Opinion Maps. These are model-based (multilevel regression and poststratification) estimates from 31 nationally representative surveys (~35,000 respondents), not direct observations.
Mortality burden. County-level attributable mortality rates from previously published studies: wildfire smoke (Ma et al., 2024, PNAS), heat (Chu, Dubrow et al., 2025, JAMA Network Open), flood (Chu, Warren et al., 2025, Nature Communications), and drought (Hu et al., 2025, preprint). The all-events rate is the sum of the event-specific rates.
Perception gap. Perceptions (%) and mortality rates (per 100,000) were each rank-transformed (1 = highest). Gapi,j = PerceptionRanki − MortalityRanki,j for county i and event j. Counties with missing heat estimates (n = 2,035) were excluded from the heat ranking.
Regression. Generalized linear models of the gap on the county percentage who think global warming is happening, adjusted for median household income and a thin-plate spline of the population-weighted county centroid (age, education, and race/ethnicity covariates were dropped after multicollinearity diagnostics). The all-events model explained 66.6% of deviance (R² = 0.64); event-specific models were similar.
Caveats (from the paper). Opinion estimates are modeled, population-level, and describe general climate opinions, with unpropagated uncertainty and potential non-independence with regression covariates; ranking may magnify small differences. Findings describe broad spatial and conceptual patterns rather than precise county-level estimates. Gaps use excess mortality only, not morbidity, economic, or social losses.
About this dashboard. The map displays the 3,094 unique counties in the underlying county dataset (consistent with the paper's rank scale of 1–3,094; 12 duplicated county rows were removed). Connecticut and six additional counties (Adams, Boulder, Jefferson, and Weld, CO; Gallatin and Park, MT) are not present in the dataset and appear as "No data." Summary statistics in the table are taken verbatim from the published Table 1 (which reports 3,106 included counties). County geometries are U.S. Census cartographic boundaries (us-atlas, Albers USA projection), contiguous U.S. only.
Data availability. County-level extreme weather-related mortality estimates: Yale ArcGIS dashboard. Yale Climate Opinion Maps 2024: YPCCC.