Short answer

Coastal flooding and sea level rise viewers are screening tools that intersect a digital elevation model (DEM) with a raised water surface and display the cells that fall below it. NOAA's Sea Level Rise Viewer and FEMA's flood mapping serve very different purposes: the first is a communication and first-pass exposure tool built on a "modified bathtub" inundation model, the second is the regulatory basis for the US National Flood Insurance Program. The single most important thing to get right is the vertical datum: the threshold (often Mean Higher High Water) and the DEM (often NAVD88) must be reconciled, or your inundation extent will be wrong by exactly the datum offset.

If you are screening a site, you can extract a defensible answer in minutes. If someone is going to design a wall, price a policy, or permit a building from your map, the viewer is a starting point, not the deliverable.

How a sea level rise viewer is built

Strip away the web interface and almost every public coastal inundation viewer reduces to the same pipeline:

  1. Start from a high-resolution DEM. NOAA's viewer is built on lidar-derived bare-earth DEMs, usually at 1–3 m horizontal resolution with reported vertical accuracy on the order of ±15–25 cm RMSE in open terrain (worse under vegetation and in marsh).
  2. Choose a water level. This is a still-water plane: a baseline tidal datum plus a sea level rise increment (typically 0–10 ft / 0–3 m in 1 ft steps in NOAA's tool).
  3. Subtract. Every DEM cell whose elevation is below the water plane is a candidate for flooding.
  4. Apply hydrologic connectivity. A naive subtraction floods every low pixel, including a parking-lot depression two kilometres inland with no path to the sea. NOAA runs a connectivity rule (an 8-connected region grow from the coastline) so that only cells with a continuous below-threshold path to open water are mapped as "inundated." Disconnected lows are mapped separately as "low-lying areas."
  5. Estimate confidence. Because both the DEM and the datum surface carry vertical error, NOAA propagates that uncertainty and publishes a confidence layer flagging where the inundation boundary is statistically uncertain (roughly an 80% confidence classification).

This is the modified bathtub or equilibrium approach. It answers one specific question: which land sits below a given static water level and is connected to the ocean? It deliberately does not answer "what happens during a storm."

What the bathtub model leaves out

Understanding the model's blind spots is what separates screening from overconfidence.

  • No dynamics. A bathtub plane has no waves, no wave set-up, no run-up, and no time. Real flooding during a king tide or storm is driven by surge plus wave action on top of the still-water level, which can add a metre or more of effective elevation at the shoreline.
  • No drainage or infrastructure. Levees, tide gates, pumps, and culverts are usually invisible to a bare-earth DEM, so the model can both miss protected areas and flood through features that actually convey or block water. A 1 m DEM cell straddling a 0.6 m floodwall will average it out of existence.
  • No erosion or morphological change. The shoreline is treated as static. Over decadal sea level rise horizons, beaches and marshes migrate, which the equilibrium model cannot represent.
  • DEM error dominates near the boundary. Because the inundation edge is the contour where elevation equals the threshold, a 20 cm DEM bias translates directly into a horizontal shift of the flood line. On a beach with a 1% slope, 20 cm of vertical error is 20 m of horizontal error.

This is why NOAA's viewer carries a confidence layer and explicit "for screening and communication, not for navigation or site-specific decisions" language. It is honest about being a planar approximation.

The datum problem, concretely

The most common technical mistake in this domain is mixing vertical references. Coastal water-level thresholds are usually expressed relative to a tidal datum — most often Mean Higher High Water (MHHW) — while elevation data is delivered relative to an orthometric datum (NAVD88 in the US) or an ellipsoid (e.g. GRS80/WGS84 from raw GNSS or some photogrammetry).

The offset between NAVD88 and MHHW is location-specific and can be tens of centimetres to over a metre. NOAA's VDatum tool exists precisely to convert between these surfaces. A worked example for a US site:

# Reproject a NAVD88 lidar DEM and inspect it before any thresholding
gdalinfo coastal_dem_navd88.tif | grep -i -E "vertical|datum|units"

# Apply a fixed datum offset (NAVD88 -> MHHW) from VDatum, e.g. -0.42 m here,
# to express the DEM in the same surface as an MHHW-referenced threshold:
gdal_calc.py -A coastal_dem_navd88.tif --outfile=dem_mhhw.tif \
  --calc="A-0.42" --NoData=-9999

# Now a 1.0 m MHHW sea level rise plane is simply "where dem_mhhw < 1.0":
gdal_calc.py -A dem_mhhw.tif --outfile=inundation_1m.tif \
  --calc="(A<1.0)*(A!=-9999)" --NoData=0 --type=Byte

The offset (-0.42 above) is illustrative; you must pull the real value for your tide station and area from VDatum. The point is that the subtraction is trivial — the datum reconciliation is where the engineering judgment lives.

A defensible screening workflow

For a corridor or site exposure screen you can do this entirely in QGIS or GDAL:

  1. Acquire the best available bare-earth DEM (USGS 3DEP lidar in the US; Copernicus DEM GLO-30 elsewhere as a coarse fallback) and record its CRS and vertical datum explicitly.
  2. Reconcile datums so the DEM and your chosen water level share one vertical reference (VDatum for US tidal datums).
  3. Threshold the DEM at the relevant water levels (e.g. MHHW + 0.3, 0.6, 1.0, 2.0 m).
  4. Enforce connectivity so you map only ocean-connected inundation. In QGIS use Raster analysis ▸ Sieve to drop disconnected speckle, or build a connectivity mask from the coastline polygon with r.grow (GRASS) / region-grow logic. Classify disconnected lows separately.
  5. Carry uncertainty. Buffer the inundation boundary by a horizontal distance derived from DEM vertical RMSE divided by local slope, and present that as a "boundary uncertainty" zone rather than a crisp line.
  6. Cross-check against an independent product: the relevant NOAA SLR Viewer scenario or, for storm-driven flooding, the FEMA Special Flood Hazard Area (Zone AE/VE) on the effective FIRM.
  7. Label the limits. State the model (bathtub/equilibrium), the datum, the DEM source and date, and that the output excludes waves, surge, and drainage.

Validation and QA

  • Datum sanity check: sample the DEM at a published benchmark or tide-gauge location and confirm the elevation matches the expected datum to within DEM RMSE. A systematic offset means a datum mistake.
  • Connectivity check: confirm no large "flooded" polygon is isolated from the coast. If one is, your connectivity rule failed or there is a real conveyance feature worth investigating.
  • Independent overlay: the screened extent should be broadly consistent with FEMA SFHA and NOAA viewer outputs at comparable water levels; large divergences usually point to datum, DEM resolution, or model-scope differences (static vs surge), all of which you should be able to explain.
  • Slope-aware error: flag flat marsh and beach where small vertical errors produce large horizontal swings, and avoid quoting precise areas there.

Common pitfalls and why they happen

  • Quoting a single flood line as fact. The boundary is a contour of an uncertain surface; it deserves an uncertainty band, not a hairline.
  • Using a DSM instead of a DEM. A surface model includes tree canopy and rooftops, so it will under-predict flooding by treating canopy height as ground.
  • Treating the NOAA viewer as regulatory. It is explicitly a screening and outreach tool. US insurance, permitting, and the floodplain regulatory line come from the FEMA effective FIRM and Flood Insurance Study, which incorporate hydraulic modeling the bathtub approach omits.
  • Ignoring local subsidence. Relative sea level rise includes vertical land motion; in subsiding deltas the local rate can far exceed the global mean, and a global increment understates exposure.

Bathyl perspective

We treat coastal viewers as the first filter in an exposure conversation, not the last word. The deliverable that earns trust states its datum, its DEM vintage and resolution, the model class, and a visible uncertainty band — then points clearly toward FEMA products and site-specific hydraulic modeling when a decision carries real consequence.

Related reading

Sources