Deltares coastal flood depth RP100 (2018 baseline)

Source: Deltares Global Flood Maps v2021.06 (CC BY 4.0), GTSM v3.0 forced by ERA5 over MERIT-DEM Year: 2018 sea level Category: Physical risks · Direct Physical Loss engine inputs Coverage: Global Format: Raster grid (92.8 m) Used in risk analysis: Yes — gates Climate hazards (Coastal flood mitigation services) at ≥ 1 m

What it shows

How deep the water stands, in metres, when a once-in-a-century storm surge floods the coast at today's sea level. "Once in a century" is the return period — the average interval between two surges at least that severe.

The figure is the flood itself, not what it costs, and it deliberately counts nothing that would hold the water back: no wave setup on top of the surge, no dike, dune or seawall, no sinking ground. A defended coast therefore reads as if it were undefended.

How it is built

Deltares couples GTSM, a global tide-and-surge model driven by ERA5 weather, to MERIT-DEM, an elevation model corrected for the height of trees and buildings. Darwin imports seven return periods, from the 2-year to the 250-year surge, at the two sea levels Deltares publishes, 2018 and 2050. This page documents the 100-year band at 2018, the one the app reports.

Two more sea levels are derived, not published. Since 2026-09-14 the loss engine also reads a 2035 and a 2080 sea level, so every horizon the app offers has its own coastal depth. Both are built by re-flooding a published return-period set at the IPCC AR6 regional median sea-level rise: 2035 from the 2018 set (one band, since the scenarios are centimetres apart by then, capped at the rise Deltares applied for 2050), 2080 from the 2050 set, split between an optimistic (SSP1-2.6) and a pessimistic (SSP3-7.0) band. On the one derivation that can be checked — 2050 rebuilt from 2018 — the vertical part is exact and the newly flooded ground is recovered at 90–96 %, with no spread beyond the 250-year rung. Their 100-year bands are visible in the layer picker, named derived; each flood loss row records whether its sea level is published or derived. Where a future depth reads below the 2018 one — coasts where the land rises faster than the sea — the engine holds the 2018 depth rather than price a horizon below today.

How to read it

The bands read as how deep the water is, not how rare the event is — the rarity is already fixed at one in a hundred years. Below 0.5 m is very low, 0.5–1 m low, 1–2 m moderate, 2–4 m high, above 4 m very high.

Class thresholds

A site is flagged on Climate hazards when the 100-year depth reaches 1 m — roughly waist-deep water, the bottom of the moderate band.

How this threshold was calibrated

It is unchanged from the layer it replaces: same metres, same return period, same No Data convention. The 1 m cut is a Darwin in-house band boundary, not a published standard.

For audit: the two grids measure 95.6 m and 92.8 m per pixel, so the swap loses no spatial detail.

Why this layer

It replaces the combined Deltares + WRI Aqueduct blend as the raster behind the flag, so the screening view reads what the Direct Physical Loss engine integrates. That blend was archived on 2026-09-14 and left this catalogue with it, which makes this layer the only coastal depth Darwin shows.

The riverine arm caught up on 2026-09-09. A flood figure sums this coastal arm and a riverine one, and until then the river side was sampled on a ~2 km resample while this side read 92.8 m — a twentyfold gap inside a single number. The riverine arm now reads GIRI flood depth RP100 (baseline, 90 m), GIRI's native grid, matching this set to the pixel. Nothing changes on this page; it is the other arm that moved.

The blend it replaces was never an ensemble of two models: both products come out of the same Deltares GTSM surge chain and differ mainly in their elevation model — MERIT-DEM here, an SRTM-based one that reads too high on flat coasts there. Averaging the cleaner model with the coarser one buys extent, not accuracy — and that is now measured: the blend is wider than this layer (×1.13 to ×3.22 depending on the coast), but 62–91 % of the ground it adds lies under the 1 m trigger, and on the area both cover it reads 0.27 to 1.28 m shallower everywhere but the Bay of Bengal. The blend also publishes the 100-year surge at 2018 only, where an annualised loss needs the whole return-period set.