The flood risk America has not mapped

America’s Flood Maps
Leave Millions of People Out
— Until Now

FEMA’s detailed flood studies cover only about one-third of U.S. rivers. A national deep-learning model estimates that 11 million more people and 4.1 million more buildings lie within 100-year flood extents the federal maps do not show.

0
U.S. rivers covered by
detailed FEMA studies
0
Additional people inside
model-estimated flood extents
0
Additional buildings inside
model-estimated flood extents
0
Of 917 metro areas
inadequately mapped
01 See Where the Official Map Ends — and the Model Continues

Flood maps help determine where America builds, where insurance is required and how communities prepare. But many rivers have never been studied in detail. Compare FEMA’s published 100-year flood extents with the national model across the contiguous United States. Amber is FEMA; cyan is the model-estimated extent.

SCROLL TO LOAD MAP · OR OPEN FULLSCREEN →
02 Completing the Map Changes the National Picture of Risk

FEMA maps place 16.06 million people inside 100-year flood zones. Adding the model-estimated extent raises that figure to 27.09 million — a 69 percent increase. The combined map also adds 4.12 million buildings beyond those inside FEMA’s published extent.

FIG. 01 — EXPOSURE INSIDE 100-YR FLOOD ZONES
FIG. 02 — MAPPING ADEQUACY, 917 US METRO & MICRO AREAS

The largest gaps leave hundreds of thousands of people outside the official extent

The scale and severity of the gap are different questions. Miami and New York contain the largest numbers of additionally exposed people, while most of the combined exposure in Los Angeles — and nearly all of it in McAllen and Houma — falls beyond FEMA’s published flood extent.

FIG. 03 — LARGEST MODEL-ESTIMATED POPULATION GAPS BY METRO AREA

Which metropolitan areas are the most undermapped?

This ranking measures the share of combined exposed residents who fall outside FEMA’s published extent. It includes metropolitan areas with at least 10,000 people in the combined FEMA and model-estimated flood extent, preventing very small exposed populations from dominating the ranking.

FIG. 04 — METROPOLITAN AREAS, MOST TO LEAST UNDERMAPPED

The model also had to work beyond the places it learned

Researchers tested the national model against FEMA maps it had not seen during training in Texas, Michigan and New Jersey. Its performance varied by terrain, but it generalized more consistently than models trained for a single region.

Validation metrics by region: precision, recall and mIoU for in-sample local models and the out-of-sample national model.
Region Local model, in-sample National model, out-of-sample mIoU gain
PrecRecmIoUPrecRecmIoUvs local
03 The Missing Risk Looks Different in Every City

St. Louis, Grand Rapids and Charlotte illustrate three versions of the same national problem: strong coverage along some rivers, thin coverage elsewhere and large differences in who appears exposed. Select a city to turn the card over and reveal its map.

Notice — Read Before Using These Maps
  • These maps show fluvial and coastal 100-year flood extents. They largely exclude pluvial (rainfall-driven) flooding, because the FEMA maps the model learned from exclude it.
  • They are a research product, not a replacement for regulatory FEMA flood zones. For insurance, lending, or permitting decisions, consult the FEMA Map Service Center.
  • Out-of-sample accuracy varies by terrain (mIoU 0.48–0.54 across validation states); treat any single location as an indication, not a determination.