What goes into a city-level risk or preparedness comparison?
How should we compare U.S. cities facing extreme weather without confusing exposure, recorded losses, vulnerability, and preparedness? I’ve been working on a nationwide city-level research framework that links government hazard and community records and makes the components behind those comparisons more transparent.
The study scores all 19,518 incorporated U.S. places and ranks the 3,173 with at least 10,000 residents. Census-designated places are left out on purpose: they have no municipal government, so they can’t adopt a mitigation plan or apply for FEMA assistance.
Federal storm records often name a county or a forecast zone, not a city. I implemented six ways of tying each event to the places it touched, from storm paths to place names in the written narratives.
In the end, 591,983 events (63.3%), carrying 49.2% of documented damage, were attributed to incorporated municipalities. Height now shows each city’s documented burden.
FEMA’s National Risk Index estimates expected annual losses for 18 hazards, and the CDC’s Social Vulnerability Index adds who is most exposed to them. Height now shows each city’s baseline risk and vulnerability.
Funded and completed FEMA mitigation projects, plus whether the city had an approved, unexpired mitigation plan on December 31, 2025. Award dollars are left out of every score because they’re missing for most cities.
Models estimate how many impacts each city should expect given its exposure, population and state; a score of 0.5 means things went about as expected. Columns rising through the glass plane did better than expected. Shorter ones did worse.
Houston ranks 153rd on burden, 35th on capacity and 3,127th on outcome performance. New York: 470th, 409th and 2,566th. Dallas: 25th, 932nd and 3,148th. Averaging those into one number would hide exactly what makes each city different.
Height now shows how wide each city’s 90% rank interval is once event sampling and model refitting are included. The median spans 828 places, more than twice the 374 from weighting choices alone. Chicago ranks #410, but could sit anywhere from #345 to #1,336.
Weakly associated, not one score. High-risk cities aren’t systematically better or worse documented, and the two readiness-facing measures carry largely independent information.
That is also why the absence of documented mitigation can’t be read as an absence of preparedness. A blank in one federal record says nothing about the others.
“Averaging them hides the disagreement; publishing them side by side is the finding.” From the manuscript’s discussion.
Deciles, not exact ranks, are the defensible unit of comparison.
Only 1,161 cities (37%) have a weight-only interval narrower than a tenth of the field, and the full-pipeline interval is the one to report.
See Chicago in the city explorer →A significant pre-award loss trend among the 314 cities that received their first funded FEMA mitigation award between 2013 and 2021, compared with cities that never did.
Cities that got mitigation money were already losing more before their first award.
Grants are a response to losses, so later losses can’t be read as mitigation working or failing. The study makes no causal claim, and the pipeline withholds one automatically when the pre-trend test fails.
of NOAA storm records have blank or unreadable damage. They stay missing, never zero.
City readiness studyof FEMA mitigation records carry no city name. A city with no record hasn’t been shown to do no mitigation.
City readiness studyIllinois ground stations in my AerisIQ rainfall-forecast validation, where missing observations stay missing instead of becoming zero rainfall.
AerisIQ“A reader deciding how much to trust a city ranking needs to know what kinds of error the method is prone to, not only that the final run passed.”
Chicago and Houston were missing or ranked unexpectedly. I traced it to hazard-specific normalization and scoring, revised the method, and checked the corrected outputs against the underlying FEMA data.I also reran calculations across 200+ sample cases to find missing cities and unusual rankings.
A name-matching defect that credited 1,354 flood events to a 14,000-person desert town.
A flood-polygon proxy that assigned a $10 billion county-wide Hurricane Harvey flood to the one city its proxy segment crossed.
A rank-direction inversion.
An outcome-scoring artefact that flattered places recording the fewest events.Cities with no attributed event scored a median 0.92 under the earlier specification; after the fix, 0.66.
To support inquiry, comparison and data-quality improvement, not to certify that one city is categorically more resilient than another.
Every damage figure is a documented lower bound: NOAA Storm Events captures about a quarter of national billion-dollar losses in nominal terms (15% for hurricanes, 13% for drought).
FEMA’s 18 National Risk Index hazards, sorted by how reliably NOAA events could be located.
Dark: fully supported (9). Tinted: descriptive only (5). Struck through: unsupported (4).
| Source | Vintage | Role |
|---|---|---|
| NOAA Storm Events | 2010–2025 | Events, damage, deaths, injuries, narratives |
| FEMA National Risk Index | Current tract release | Expected annual loss and exposed population, 18 hazards |
| FEMA OpenFEMA HMA | Through 2025 | Funded and completed mitigation records |
| FEMA plan-status records | Status at 31 Dec 2025 | Approved, unexpired mitigation plan |
| CDC/ATSDR SVI | 2022 | Social vulnerability |
| Census TIGER/Line | 2025 | Municipal boundaries |
| American Community Survey | 2024 five-year | Population for eligibility and scaling |
| BLS CPI-U | Through 2025 | Inflation adjustment to 2025 dollars |
Search any of the 3,173 cities, see its four components side by side, and see how wide its rank interval really is.
First author: Anika Raj. Mentors: Dr. Abhinav Wadhwa and Dr. Ashish Sharma, Discovery Partners Institute, University of Illinois System. Submitting to Natural Hazards (Springer Nature), October 2026.
Climate R&D intern, with Dr. Abhinav Wadhwa.