Methodology
How FloodLens calculates flood risk for addresses in Louisiana. Last reviewed October 9, 2026.
Four lenses: Home, Drainage, Levee, Surge
A flooded house can be a house problem, a drainage problem, a levee problem, or a storm-surge problem, and they behave differently. The four tabs separate them, and each keeps the same name everywhere in the app.
Home is the property itself: FEMA flood zone, ground elevation, and the risk score with its six components. Below the score, context cards add history and regional background: FEMA National Risk Index ratings, New Orleans 311 reports (Orleans only), NFIP claims history by census tract (Orleans and Jefferson), and recorded flood events (surveyed high-water marks and the parish's disaster declarations). See Recorded flood events and other context below.
Drainage covers the pumps and drainage that serve the neighborhood (see Drainage lens: pumps and drains below). Drains and pumps remove rain that falls inside the levee system. When rain outpaces them, or equipment fails, the result is street and home flooding: more frequent, and usually shallower.
Levee shows whether FEMA maps the location as protected by a levee and where the nearest FEMA-mapped levee line is. Levees and floodwalls keep lake, river, and surge water out. Failures are rare but can be deep and sudden. FEMA's levee data records accreditation for flood-insurance purposes; in this region its owner, design-frequency, and freeboard fields are empty, and it says nothing about a levee's current height, settlement, or inspection results. Those gaps are listed in the card, with dated links to recent reporting that was read directly. Accreditation means reduced flood risk, not zero.
Surge shows the National Hurricane Center's storm-surge maps for the location: for each hurricane category 1 to 5, how deep surge could get above ground. NHC's “Maximum of Maximums” product keeps the highest surge that any storm of a category could produce in a place, across many storm tracks and speeds. It is a planning envelope, not a forecast for any storm. In much of Greater New Orleans NHC marks the ground Leveed Area and shows no depth. That is how NHC mapped it, not a finding that surge cannot reach the area, so read it together with the Levee tab. A location with no surge mapped is not guaranteed safe from other kinds of flooding. Not assessed: any specific storm, what happens if a levee or floodwall is overtopped or fails, rain-driven flooding, how long water stays, and wave or wind damage.
What feeds the score
Only the six components listed under Score Components feed the risk score. Everything else in the app is informational: the Levee and Surge tabs, FEMA National Risk Index ratings, 311 reports, NFIP claims history, and recorded flood events. They are shown next to the score, never combined into it, and have not been tested as predictors of claims. Changing or adding one of them does not change any score.
Coverage — what varies by region
Any Louisiana address can be looked up, but not every component is available everywhere, and the score says so rather than papering over the gap.
Statewide, where the source has data. FEMA flood zone (25%), USGS ground elevation (20%), below-BFE depth (15%), and NOAA Atlas 14 rainfall intensity (15%) come directly from the source APIs — 75% of the model.
Where FEMA's effective map has no flood zone. FEMA's public effective flood-hazard layer returns no zone for some places, for example Slidell and Covington (separately mapped cities in St. Tammany Parish), Thibodaux, and Alexandria, where the legally effective map is a scanned paper map. That is not a lookup failure on our side. In those places FloodLens falls back to FEMA's preliminary flood map, which FEMA publishes for the same locations. A preliminary map is real FEMA data, but it is not in effect and not for insurance determinations, and it can be old: St. Tammany's was issued in April 2008, Rapides' in July 2007, and Lafourche's in October 2024. So the zone is labeled as preliminary with its issue date, and the two score components built on it are marked estimated, never measured; their numbers are computed exactly as for an effective zone. An effective zone always takes priority. Where there is neither (mostly small rural parishes whose only map is a scanned paper map, such as East Carroll, Tensas, Sabine, Red River, Catahoula, Claiborne, and Pointe Coupee), the flood-zone and below-BFE components (40% of the model) are missing, data coverage falls to 60%, below the 70% floor, and FloodLens shows Insufficient Data. We do not substitute a guess. Coverage is by location, not by parish.
Greater New Orleans only. Drainage (15%) is built from pump-station and basin data for the metro area: Orleans from the Sewerage & Water Board's published layers, Jefferson and St. Bernard from a curated dataset, and St. Tammany treated as naturally draining (see the FAQ). It applies only inside a defined footprint around the metro. Subsidence (10%) is interpolated from published InSAR studies and applies only within 25 miles of a documented zone. Outside that footprint both are reported as no data, dropped from the score, and the remaining weights renormalized — the lookup is labeled 75% data coverage. FloodLens does not extrapolate a Kenner marsh subsidence rate to a property a hundred miles away, and does not read "no pump station nearby" as high drainage risk in a parish where no pump data was ever collected.
Elevation outside the delta. The elevation component is normalized over the delta's usable band (−6 to +12 ft NAVD88), where nearly all of the metro's variation sits. Above +12 ft it saturates: high ground scores 0 whether it is 13 ft or 200 ft. Those lookups are flagged estimated rather than measured, because the number is real but no longer discriminates between locations.
Validation. The backtest results below were measured against NFIP claims in the New Orleans metro. They have not been revalidated outside it, so they should not be read as evidence about scoring accuracy in other parishes.
Coverage by feature.
| Feature | Where it applies |
|---|---|
| Risk score: flood zone, elevation, below-BFE, rainfall | Statewide. FEMA's effective flood zone where it exists, otherwise its preliminary zone (labeled, scored as an estimate); Insufficient Data where there is neither |
| Risk score: drainage and subsidence | Greater New Orleans only; otherwise reported as no data |
| Drainage tab pump review | Orleans (SWBNO data); Jefferson and St. Bernard (curated, unverified) |
| Levee tab | Statewide (FEMA data) |
| Surge tab | Gulf coast; inland locations show no surge mapped |
| FEMA National Risk Index | Statewide |
| NFIP claims history | Orleans and Jefferson |
| 311 reports | Orleans only |
| Recorded flood events | Statewide. High-water marks exist only where an event was surveyed; declarations are per parish. |
| Risk heatmap | Greater New Orleans metro grid; blank over open water and where FEMA has no zone of either kind |
Risk Score (0–100)
Each address receives a weighted composite score from 0 (lowest risk) to 100 (highest risk), based on six independently sourced data components. The score is not a probability of flooding — it is a relative risk index for comparing locations within the coverage area.
Score Components
Data Confidence & Coverage
Every component score carries a provenance tier, shown in the results panel:
- Measured — parcel-specific data returned directly by the source (FEMA NFHL zone, USGS elevation, NOAA Atlas 14 rainfall). Rendered as a solid bar.
- Estimated (marked EST) — data is present but methodologically indirect: pump-station distance/capacity proxies, interpolated subsidence rates, zone-implied values, or a flood zone taken from FEMA's preliminary map. Rendered as a hatched bar.
- Missing — the underlying source failed for this lookup. The component is shown as "no data" and is excluded from the overall score; the remaining weights are renormalized. The score ring shows the missing share as a grey gap, and the risk label notes exactly which component is absent.
If less than 70% of the model's weight is backed by available data, FloodLens does not produce an overall score or risk label at all — you'll see "Insufficient Data" along with whichever components could still be retrieved. A missing input is never silently replaced with a neutral default.
Drainage lens: pumps and drains
Greater New Orleans drains by pumping, so a property review is incomplete without the neighborhood's pump stations. Open the Drainage tab to see stations on the map, and every pump-dependent address lookup includes a Pump & drain review. Everything in it is derived from the curated station and basin records; nothing is rated that the records do not support.
Current status: Orleans is sourced; Jefferson and St. Bernard are not. Orleans Parish stations come straight from the Sewerage & Water Board's published station, pump, and basin layers (real coordinates, per-pump capacities, and tributary-area acreage). A check of the older curated records against the 2006 federal IPET report (Volume VI) had found that nearly all of the Orleans capacities and locations were wrong, so they were replaced. Jefferson and St. Bernard records are still curated, and several disagree with IPET. Where a published rating could be read (the IPET report, and the canal-closure pump fact sheet for the 17th Street and London Avenue canals), the station's capacity is shown as rated, with the source linked and the model's listed figure shown beside it if they differ. The remaining stations are labeled unverified. Rated figures date from 2006 or later and may have changed. A pumping rate and rainfall comparison appear only where both the station's capacity and its basin acreage are sourced and the property lies inside that SWBNO tributary basin. Otherwise they are withheld, and the serving basin is never inferred from the nearest station.
The risk score uses the same station data (v7). The drainage component now reads SWBNO's Orleans stations. The backtest was re-run on identical claims and policies: predictive skill was neutral (within-A 3.03 to 3.22; within-X 1.50 to 1.49; every difference within overlapping confidence intervals), and the drainage component moved by about 6 points on average. See the validation page. Jefferson inputs are unchanged.
What it shows. The nearest station and its distance, its capacity in cfs (marked rated, unverified, or withheld), the stations within 3 miles, and the basin's outfall and canal where on record. Where the station's capacity and basin acreage are both sourced and the property sits inside the basin, it also shows a pumping rate: the basin's total station capacity divided by its drainage area, expressed as inches of rain per hour (1 cfs sustained for an hour moves about 1.008 acre-inches), compared with NOAA Atlas 14 rainfall for 1-, 6-, and 24-hour storms at 10- and 100-year return periods. That comparison is a gross one: it ignores infiltration, canal and street storage, and how many pumps are actually running. Today that applies to Orleans only; it is withheld for Jefferson and St. Bernard until their capacities are verified.
Record quality. Each station is marked complete or partial depending on whether its basin has an area, outfall, and canal on record. Where basin area is missing, the pumping rate and rainfall comparison are not shown rather than estimated.
What is not assessed. Operating condition and age, maintenance history, backup power, outage and failure history, and rated-versus-effective capacity are not in the dataset, so FloodLens does not rate them. For Jefferson and St. Bernard, the station set is curated from public documents with approximate coordinates, is not verified against operator records, and is not an exhaustive inventory. In Orleans a property's serving basin is the SWBNO tributary area that contains it (SWBNO's historical map, which predates recent changes); in Jefferson and St. Bernard it is the basin of the nearest station, not a surveyed boundary, which is why no pumping rate is shown there. Proximity to a pump is also not protection: FloodLens's Aug 2017 backtest found areas rated as well-protected by pump proximity were hit hardest.
Recorded flood events and other context
These cards describe the past or the region. None of them is part of the risk score.
- High-water marks. Surveyed flood marks within 2 miles, from the USGS Short-Term Network archive, grouped by event with the nearest mark's distance and elevation. A mark is a water-surface elevation above sea level (NAVD88) at a nearby building, not a depth, and the records carry no height above ground, so FloodLens shows the property's own ground elevation beside it and subtracts nothing. Only some events were surveyed (Hurricane Katrina and the August 2016 flood are dense; Hurricane Ida has only a handful), so no marks nearby does not mean nothing flooded. On the Home map each mark is a dot colored by event; tap a dot for its elevation, distance, and survey quality, or tap an event in the card to highlight its marks. The marks are a snapshot refreshed when new events are surveyed.
- Disaster declarations. The parish's federal major-disaster declarations since 2005 for floods, hurricanes, and coastal storms (plus severe-storm events whose title names flooding or a tropical storm). A declaration covers a whole parish and shows which events drew federal help, not whether a given address was affected.
- NFIP claims history. Claims, biggest events, and a claim rate for the census tract, from insured properties only; see Data Sources.
- FEMA National Risk Index and 311 reports. Regional ratings and residents' drainage reports; see Data Sources.
Risk Heatmap
The map's "Risk Heatmap" toggle shows a continuous, color-blended visualization — not individual markers — on the Home tab. It's generated by pre-scoring a grid of sample points across the Greater New Orleans metro, then letting the map smoothly interpolate color between them. Switching to "Circles" mode shows the underlying sample points individually instead of the blended view. Address lookups cover the state, but the heatmap does not extend beyond the metro. It is blank over open water and wherever FEMA has no flood zone of either kind. Points in areas that have only preliminary maps (much of St. Tammany) are scored from those maps, and the legend says how many.
Sample spacing varies by area, from roughly 0.15 to 0.55 mile. This only affects the visual smoothness of the heatmap layer — it has no effect on the accuracy of an individual address lookup, which is always computed live regardless of where you search.
The heatmap is a snapshot, scored like a live lookup. Its sample points are scored with the same model as a live lookup, including rainfall, which comes from NOAA Atlas 14 values on a 0.05-degree lattice interpolated between lattice points. In spot checks against live lookups, including Orleans addresses, the two agreed to within 0.1 point. It is regenerated whenever the model, the pump-station data, or the subsidence data change, and the map legend shows when it was last generated. It is still a snapshot, so a live lookup is authoritative.
Data Sources
- FEMA NFHL — National Flood Hazard Layer via ArcGIS REST API. Provides flood zone designations, SFHA boundaries, and Base Flood Elevations. Updated periodically by FEMA; may not reflect recent map revisions. Where this effective layer has no zone, FloodLens uses FEMA's Preliminary NFHL (FEMA's PrelimPending service) instead, labeled with the preliminary map's issue date.
- USGS EPQS — Elevation Point Query Service using the 1/3 arc-second (~10m) National Elevation Dataset. Returns ground elevation in feet for any coordinate in the US.
- NOAA Atlas 14 — Precipitation Frequency Data Server. Provides rainfall depth-duration-frequency estimates based on historical weather station records.
- Drainage pump stations — Orleans Parish: the Sewerage & Water Board's public ArcGIS layers (stations, per-pump capacities, tributary-basin acreage), read from a snapshot. Jefferson Parish and St. Bernard Parish: a curated dataset compiled from public records, parish documents, and infrastructure reports, with station locations and pumping capacities, audited against published sources; Jefferson does not publish a pump layer. St. Tammany Parish is excluded from this dataset by design — see the FAQ on why its drainage scoring differs.
- FEMA National Risk Index — Census-tract ratings for inland flooding, coastal flooding, hurricane, and all hazards, relative to the rest of the U.S. and built from historical losses and exposure. Shown as statewide context only; it does not capture pump or levee failure and is not part of the risk score.
- FEMA levee layer — The NFHL Levees layer and each flood zone's levee subtype, used for the Levee tab. Informational only.
- City of New Orleans 311 — Residents' reports of catch basins not draining and street flooding within half a mile, over the last three years (Orleans Parish only). These show where people call, not where water stands, and are not part of the risk score. The city's “Street Flooding” category stopped being used after 2025, so recent flooding may be filed under other categories.
- NFIP Claims — OpenFEMA National Flood Insurance Program claims and policies. Used for model validation, and for the “What has flooded here” card in the Home view (Orleans and Jefferson parishes for now): claims by year, the biggest events, and a claim rate per 1,000 insured policy-years for the census tract. Claims exist only for insured properties, so the counts reflect insurance take-up as well as flooding; this is a record of the past, not a forecast, and it is not part of the risk score.
- NOAA/NHC storm surge maps — The National Hurricane Center's National Storm Surge Risk Maps (“Maximum of Maximums,” hurricane categories 1–5), used for the Surge tab and its map overlay. A planning product that changes as NHC updates it, not a forecast.
- USGS high-water marks — Flood high-water marks from the USGS Short-Term Network archive, used for the recorded flood events card. Only events that were surveyed are present.
- OpenFEMA disaster declarations — Parish-level federal major-disaster declarations for floods, hurricanes, and coastal storms, used for the recorded flood events card.
- Subsidence Zones — Estimated from published InSAR satellite studies of Greater New Orleans ground motion (Jones et al. 2024, Science Advances; Dixon et al. 2006, Nature), CPRA's 2023 Coastal Master Plan, and USGS surficial geology mapping of the Pontchartrain Basin. No public address-level API exists for subsidence; rates are interpolated from documented zone-level patterns (natural levee vs. reclaimed swamp/marsh vs. St. Tammany's Pleistocene terrace), not measured at each address.
Definitions
| Term | Definition |
|---|---|
| SFHA | Special Flood Hazard Area. The land area that FEMA has determined has at least a 1% chance of flooding in any given year (also called the 100-year floodplain). Flood insurance is required for federally backed mortgages in SFHAs. |
| Zone AE | SFHA zone with detailed Base Flood Elevations determined. High-risk area where flood insurance is mandatory. |
| Zone X | Area outside the SFHA. Moderate-to-low risk. Flood insurance is not required but is recommended — roughly 25% of NFIP claims come from Zone X properties. |
| Subsidence | The gradual sinking of land over time. In Greater New Orleans, this is driven mainly by drainage of organic soils and historical groundwater/oil & gas extraction. Measured in millimeters per year (mm/yr); even slow rates compound significantly over a 30-year mortgage. |
| Natural Levee | A ridge of higher, more consolidated ground formed naturally along a river or bayou over centuries of sediment deposition. The oldest and highest neighborhoods in New Orleans (French Quarter, Uptown) were built on these ridges and subside slowly. |
| BFE | Base Flood Elevation. The height floodwater is expected to reach during a 100-year flood. Buildings should be constructed at or above BFE. |
| NAVD88 | North American Vertical Datum of 1988. The reference point for elevation measurements. An elevation of 0 ft NAVD88 is roughly sea level. |
| CFS | Cubic Feet per Second. Measures the pumping capacity of drainage stations. Higher CFS means more water can be moved per second. |
| Leveed Area | NHC's label for ground it maps as behind a levee, where it shows no surge depth. A mapping convention, not a statement that surge cannot reach it. |
| Maximum of Maximums (MOM) | NHC's near-worst-case surge product: for each hurricane category, the highest surge in each spot across many storm tracks, speeds and landfall points. Not a forecast. |
| High-water mark | A surveyed point where floodwater reached after a named event, recorded as an elevation above sea level, not a depth. |
| Disaster declaration | A federal major-disaster designation for a parish after a storm or flood. It covers the whole parish. |
| Preliminary map | A FEMA flood map that has been published for review but has not taken effect. It can differ from the map that eventually becomes effective, is not used for insurance requirements, and can be many years old. |
| NFIP | National Flood Insurance Program. The federal program that provides flood insurance to property owners, managed by FEMA. |
Validation
FloodLens is backtested against 15 years of National Flood Insurance Program claims (2010–2024) in Jefferson and Orleans parishes — 16,526 claims against 2.9 million policy-years of exposure, aggregated to census tracts. The full methodology, results, and limitations are public in the backtest report on GitHub, including the pre-registered pass/fail criteria we set before computing any results.
- Inside FEMA's high-risk zones, the score works. Among properties FEMA rates identically (A/AE zones), top-quintile FloodLens tracts filed NFIP claims at 3.2× the rate of bottom-quintile tracts (95% CI 3.0–3.4, 9,402 claims). This is not repackaged elevation data: ranking the same tracts by elevation alone shows no such separation — the discrimination comes from the combined zone, drainage, and subsidence signal.
- Outside the SFHA (Zone X), we do not yet claim validated differentiation. At census-tract resolution the claim-frequency separation in Zone X was roughly 1.5×, but it did not meet the confirmation threshold we pre-registered, and most of it traces to a single early-window hurricane rather than a stable year-over-year signal. Until finer-grained validation exists, treat Zone X scores as informative but unproven.
- Not everything is validated. The backtest tests the risk score. The Levee and Surge tabs, National Risk Index ratings, 311 reports, NFIP claims history, and recorded flood events are informational and have not been tested as predictors. The October 2026 switch of Orleans pump-station inputs to SWBNO data was backtested on identical claims and policies and was neutral. The backtest is refreshed monthly as OpenFEMA publishes new claims; the figures above are from the original 2010–2024 run.
- We publish what fails. The backtest surfaced a real model defect — the drainage component originally treated pump proximity as strong protection, which the August 2017 pump-failure flood contradicted — and the current model incorporates that correction.