FEMA Says Your Lot Won’t Flood. An AI Model Disagrees. That Gap Could Cost You $120,000.

Last September, Zillow started showing First Street Foundation’s climate risk scores on every for-sale listing in the country. Five categories: flood, wildfire, wind, heat, air quality. Each rated on a 1–10 scale, with projections out to 2050.

Most buyers scrolled past it.

They shouldn’t have. First Street’s AI flood model covers 142 million US properties and identifies 6 million facing substantial flood risk that FEMA’s maps don’t show. Not a minor discrepancy. Not a rounding difference. Six million properties where the federal map says “you’re fine” and the climate-adjusted model says “you probably aren’t.”

If you’re buying a lot or building a house, that disagreement has a dollar value. I ran the math.

Why FEMA Maps Lie by Omission

FEMA flood maps are not predictions. They are records. Built from historical flood data, stream gauge measurements, and topographic surveys, they answer one question: where has water gone before? What they do not model is where water will go as rainfall patterns shift, sea levels rise, and storms intensify.

First Street’s model is different. It layers historical hydrology with climate projections (multiple SSP scenarios from IPCC AR6), topography at sub-meter resolution, and local infrastructure data like stormwater capacity and impervious surface coverage. It runs Monte Carlo simulations for each property, generating probability distributions for flood depth at 1-year, 5-year, and 30-year horizons.

FEMA knows this is a problem. Risk Rating 2.0, rolled out in 2023, updated the National Flood Insurance Program’s pricing to incorporate replacement cost, flood type, and distance to water. It was a real improvement. But Risk Rating 2.0 still doesn’t project forward. It prices today’s risk more accurately. It doesn’t price 2045’s risk at all.

A 30-Year Mortgage on a Lot That Shouldn’t Be “Low Risk”

Say you’re building in a zone FEMA classifies as X (minimal flood hazard). No flood insurance required. Your lender doesn’t mandate it. You skip it. Most people do.

Now say First Street rates that same property a Flood Factor 7 out of 10 by 2040. A 1% annual flood probability, rising. What does that disagreement actually cost over a 30-year mortgage?

Inputs: NFIP average premiums from FEMA’s Risk Rating 2.0 data show 37% of single-family policies cost $0–$1,000/year and 32% cost $1,000–$2,000/year in low-risk zones. Re-rated premiums for high-risk zones run $2,500–$5,000/year, consistent with Insurify’s analysis showing flood insurance could cost 64% more without NFIP subsidies. Average residential flood damage of $50,000 comes from FEMA’s historical NFIP claims data. Note: the expected-loss row smooths what is actually a lumpy distribution. Most years you lose nothing. One year you lose $50,000+. Expected value captures the average cost, not the experience.

That $63,000–$120,000 gap is invisible at closing. It doesn’t show up in your GFE, your appraisal, or your home inspection. It only materializes when water enters your house or when the NFIP reprices your zone.

New Construction Is Not Immune

ClimateCheck’s analysis with Zelman Associates found that new homes sit at lower overall flood risk than existing stock, thanks to modern codes and better siting. But 63% of single-family building permits now land in regions with extreme heat risk, up from 40% in 1990. Builders chase cheap land. Climate risk maps rarely factor into production builders’ site selection. If you’re choosing your own lot, you have a choice most buyers don’t. Use it.

What You Can Actually Do

1. Check floodfactor.com before you sign anything. Free. Property-level. Takes 30 seconds. If First Street rates your lot above a 5 and FEMA says minimal risk, that discrepancy deserves investigation, not dismissal.

2. Cross-reference with ClimateCheck.com. Covers flood, heat, fire, drought, and storm risk. Broader than First Street’s flood-only model. Useful for understanding cumulative exposure.

3. Pull the FEMA map at msc.fema.gov. Compare it to First Street. If they agree, good. If they disagree substantially, dig into why. Is it a creek that FEMA hasn’t resurveyed since 2003? A development upstream that increased runoff? Rainfall projections FEMA doesn’t model?

4. Get a flood insurance quote before closing on land. Even in Zone X. Some properties now cost $5,000+/year. Better to know before you own the lot than after.

5. If you’re building in any elevated-risk area, think about what water touches first. A raised foundation adds $15,000–$30,000 over slab-on-grade, but pays for itself with one avoided flood event averaging $50,000 in damage. Elevate mechanicals (HVAC, water heater, electrical panel) above projected flood depth. Sealed concrete and closed-cell insulation below the flood line are cheap insurance the building code doesn’t require but physics recommends.

6. Check local freeboard requirements. Some cities now require 1–3 feet above FEMA’s base flood elevation. That’s a signal. When your local building department voluntarily exceeds federal minimums, they’re telling you something about FEMA’s maps.

First Street Is Not Gospel Either

A strong counterargument deserves a fair hearing. First Street’s model relies on climate projections that carry real uncertainty. SSP scenarios from IPCC AR6 span a wide range of futures, from aggressive decarbonization (SSP1-2.6) to fossil-fuel-intensive growth (SSP5-8.5). Depending on which pathway materializes, the model’s 2050 flood projections could overstate risk by a factor of two or more.

And the model is young. It hasn’t been validated against 30 years of actual flood events because it hasn’t existed for 30 years. FEMA’s maps, backward-looking as they are, rest on decades of observed flood data. There’s a real argument that over-predicting flood risk inflates insurance costs and prices people out of neighborhoods that may never see water. When Insurify’s data shows homeowners are already forgoing coverage due to cost, adding another layer of risk assessment could accelerate that problem.

Fair points, all of them. But the asymmetry matters: if FEMA under-predicts and you skip insurance, you absorb the loss. If First Street over-predicts and you buy insurance, you’re out $900–$1,500 a year on a policy you didn’t need. One of those mistakes is recoverable. One of them isn’t.

What This Analysis Didn’t Prove

First Street’s methodology is proprietary. Independent peer review exists but remains limited. I could not verify the specific calibration parameters behind their Monte Carlo flood depth simulations, and neither can you.

My 30-year cost calculation uses current NFIP premium structures. Those premiums will change. Risk Rating 2.0 is actively closing the gap between FEMA pricing and actual risk, which means the “re-rating shock” in my model may be smaller by the time it materializes, or the premiums may already be rising by the time you read this.

Climate models themselves disagree on regional precipitation changes. Two GCMs projecting 2050 rainfall for Houston can differ by 30%. First Street aggregates across models, but aggregation smooths disagreement rather than resolving it.

New construction already benefits from higher building codes. Homes built to IRC 2021 in flood-prone areas often include elevated foundations and flood vents by default. For these homes, the damage estimate of $50,000 per event is likely too high; a raised foundation with elevated mechanicals might sustain $10,000–$15,000 in a moderate flood. My cost model doesn’t differentiate. Similarly, I didn’t address the private flood insurance market, which is growing and sometimes cheaper than NFIP for low-probability, high-consequence properties. If that describes you, shop private carriers alongside NFIP. Insurify covers the basics of how those markets differ.