Gulf Coast flood losses: Ivan, not Katrina.

Ivan produced 19% more claims here than Katrina.

Katrina is the storm the country reaches for whenever Gulf Coast flooding comes up. In the federal claims record for these two Alabama counties it is second. Hurricane Ivan accounts for 6,699 paid NFIP claims and $351,938,152; Katrina accounts for 5,629 claims and $287,940,155. Ivan leads on both measures — 19% more claims and 22.2% more paid.

That is a local fact, not a national one. Katrina's catastrophic losses landed to the west, in Louisiana and Mississippi. Ivan came ashore on the Alabama coast, and the claims follow the landfall: its dates of loss here run 2004-09-15 to 2004-09-18. Anyone writing about flood risk on this stretch of coast using Katrina as the reference storm is using the wrong one, and the correction is a single API query away.

Events with at least 100 claims. All 27 labeled events are in the CSV. Dates are the observed minimum and maximum date of loss among the claims FEMA itself labeled with that name — not an inferred window.

25 of the 49 years on record had fewer than fifty claims.

Flood loss on this coast is not a steady annual cost with occasional spikes. It is a flat line interrupted by a handful of days. 2004 and 2005 — Ivan and Katrina — hold 13,082 claims between them, which is 37.5% of everything filed across 48 years. Meanwhile 25 separate years produced fewer than fifty claims each.

Ten years shown of 49 (1978 to 2026). Every year is in the CSV.

The 1970s row covers 1978-01-15 onward; the 2020s row is incomplete and runs to 2026-06-18. Dollars are nominal and not comparable across decades — see section 06.

The 2010s are the quiet decade in this record — 1,886 claims across ten years, less than a seventh of the 2000s. Read that as the absence of a landfalling storm in the window, not as a change in exposure. Sally arrived in September 2020 and the 2020s row passed the entire 2010s in a single event.

20 ZIP Codes carried the wrong county label until this run.

The first version of this aggregation grouped claims by ZIP Code alone and stamped each ZIP with a county name. That is a bug, because a reported ZIP is not constrained to lie inside the county the claim was filed under. The two counties are fetched in sequence, Baldwin first, so every ZIP with claims under both counties inherited the Baldwin label — and both counties' claims were merged into that one mislabeled row.

The worst case was ZIP 36582. It has 882 claims filed under Mobile County and 1 filed under Baldwin County. The old aggregation emitted a single row reading 883 claims, county Baldwin — a figure a local reader would have caught on sight.

The fix is to key the aggregation on the (ZIP, county) pair, so a shared ZIP produces one row per county with its own honest count. 20ZIPs do that here, plus the bucket of claims with no ZIP recorded at all. Every other aggregation on this page — the totals, the years, the events, the zones, the causes, the occupancy types — was unaffected, and the script's assertion that each grouping sums back to the fetched claim count passes before and after.

Why this is a ZIP map and not a pin map.

The claims dataset does carry latitude and longitude. Both are rounded to one decimal degree — roughly 11 kilometers, about seven miles — as deliberate privacy protection, because each row is one household's insurance claim on one address. At that resolution a coordinate does not identify a property, a street or a neighborhood. It identifies a rectangle large enough to hold most of Mobile Bay.

Plotting those coordinates as dots would produce a map that looks precise and is not. Points would stack into a lattice at every tenth of a degree, and a reader would reasonably infer that a dot marks a house. So this pipeline does not fetch the coordinate fields at all. The spatial unit is the 5-digit ZIP Code FEMA already reports, drawn on 2020 Census ZCTA boundaries, and the map claims exactly the resolution the source supports.

Take the data.

One CSV holds every aggregate on this page: the overall totals, all 138 ZIP-and-county rows, all 49 years, all 27 FEMA-labeled events, all 37 rated zones, all 12 cause codes and all 14 occupancy codes. It is generated from the same constants this page renders from, so the published numbers and the downloadable numbers cannot drift apart.

Download cds-gulf-coast-flood-losses.csv

Published 2026-08-15 against FEMA data as of 2026-08-03. Corrections welcome — section 05 is what one looks like when it lands.