Where storms overlap Election Day voting
A frozen afternoon snapshot across all 435 U.S. House districts and the 35 states with a U.S. Senate contest. Weather is sampled over populated areas and weighted by 2024 Election Day participation. Partisanship describes historical votes in affected places.
This index measures exposure. A score of 20 does not mean 20% of voters will miss voting. Counts refer to the historical electorate, including people who may already have voted today. Election Day participation proxies and unavailable party splits are labeled.
House district rankings
| Rank | District | Exposure index | Weather severity | ED share | Method coverage | ED count basis |
|---|
Affected places
Shows up to 40 areas. The CSV contains every mapped district piece. Areas are ranked by allocated historical Election Day votes in locations reaching severity 30 or higher. These are 2024 counts and population-based allocations, not a count of people yet to vote. “All modes” is separate partisan context when Election Day party counts are missing.
| Place | Historical ED weight | Weight in severity 30+ | Weather severity | ED R minus D | All-mode R minus D | Weather alerts |
|---|
How the index works and what it cannot establish
At each population sample, peak severity is the largest rain, gust, frozen precipitation, or NWS alert score observed in the current snapshot or forecast before the district's latest poll closing. Peak scores are analyst-defined, bounded from 0 to 100, and uncalibrated. They are not official NWS severity levels.
Exposure index = sum of (historical Election Day weight × local peak severity) / historical all-method presidential votes. A district where 20% of votes were cast on Election Day receives one quarter the exposure score of an otherwise identical district with 80% Election Day voting.
- Exact source Election Day counts are used where available. Elsewhere, a separate EAC EAVS proxy multiplies presidential votes by the administrative Election Day ballot share. Administrative ballots and presidential votes are different denominators. Inconsistent or missing EAVS rows remain unknown.
- Candidate-by-mode counts are never imputed. The affected Election Day partisan margin covers only locations with actual party counts. Affected all-method margins describe a different electorate.
- County or town totals are allocated using 2020 resident population inside exact 2026 block-to-district assignments. Population is a proxy for voter location. The map does not identify individual voters or their travel routes. Alaska uses statewide vote weights, so local partisan variation is unavailable there.
- Ranking requires at least 95% usable vote-method coverage, 95% weather population coverage, and 95% source-vote geography coverage. Unknown districts still have weather severity. Partial-data bounds are included in the CSVs.
- NOAA HRRR provides 3 km hourly weather for the contiguous U.S. from the 18Z run. Alaska and Hawaii use the 12Z GFS at 0.25 degrees. Radar precipitation is MRMS at 19:54Z. Negative radar values remain missing.
- Forecast rain is the preceding hour's accumulation. The 20Z forecast is not counted as future 20Z to 21Z rain. Forecast intervals are clipped to the latest district poll closing, including partial hours. Areas within a district can close earlier. The mean-duration measure weights open time uniformly, not voter arrivals.
- NWS warning polygons and advisory zones are dated and filtered to the relevant window. A short alert overlap can raise the whole hourly weather bin. Forecast-zone polygons can cover more land than the eventual impacts. No morning weather, morning turnout, actual remaining electorate, travel disruption, outages, or polling-place closures are modeled.
- Rain-rate and gust thresholds, plus alert floors, are available in run_metadata.json. Rain sensitivity uses 75% and 125% of the liquid-rain component. It is a parameter check, not a forecast confidence interval.
Google WeatherNext 3 denied bucket access and Global MetNet returned an inaccessible-asset response. No access, billing, or terms were changed.
Sources: NOAA HRRR, NOAA MRMS, NWS alerts, EAC EAVS, Census district assignments, district closing times. Local voting-source details and hashes are preserved in the reproducibility bundle.