It screens individual infrastructure assets — ports, commercial airports, power plants — against global flood hazard, and translates the hazard into two financial quantities: expected damage as a share of asset value, and the resulting uplift in the loan's probability of default, in basis points.
42,037 assets worldwide: 34,936 power plants (WRI Global Power Plant Database), 3,297 commercial airports with scheduled service (OurAirports), and 3,804 ports (NGA World Port Index). Roads, railways and other spatially distributed infrastructure are deliberately excluded, as a single point cannot represent their exposure.
WRI Aqueduct Floods v2 inundation maps (riverine and coastal), sampled at each asset's coordinates across 689 hazard layers: historical conditions and RCP 4.5 / RCP 8.5 projections for 2030, 2050 and 2080, five global climate models, with and without subsidence, three sea-level-rise percentiles, and return periods from 1-in-2 to 1-in-1,000 years.
Following Eq. 3 of Assab (2025): for each return period, the flood depth at the asset is converted to a damage fraction via a depth-damage function, weighted by the flood's annual probability (1 / return period), summed across return periods and flood types, and scaled by the maximum damageable share of the asset.
Twenty-two: the JRC global and regional industrial curves (Huizinga et al. 2017), sector curves for airports, infrastructure, residential, commerce and industry by region, and the linear specification used in the paper's baseline. The choice is yours in the tool; the default is the JRC global industrial curve.
Econometrically. The paper matches 951 real project-finance loans to Moody's default and recovery data and finds that a one-standard-deviation increase in expected flood damage raises the probability of default by 99 basis points, controlling for size, tenor, sector, country and year. The tool applies that estimate to the asset's expected damage, on top of a baseline PD taken from the same panel (region × sector average).
It is a what-if illustration: it removes damage from floods more frequent than the selected design standard, when the standard is enforced. The paper's key finding on adaptation is reflected separately: enforced design standards reduce default probability by about 4 percent, while policy-only standards do not.
The hazard maps are undefended: they model where water would go without local flood defences. A Dutch port showing 2 m depths at low return periods is exposed by geography but protected by engineering — that is precisely the gap the protection controls let you explore. Treat high modelled damage in defended locations as "dependence on defences", not as expected loss.
No. Baseline PDs are statistical averages by region and sector; damage functions are regional engineering archetypes; locations are single points. FloodGate is a screening and comparison tool — a consistent first pass across thousands of assets, not a substitute for due diligence on one.
Yes — this is a design principle. The entire tool runs from a static dataset (CSV files): the asset register, the per-country depth matrices, and a master results database with expected damage and adjusted PD per asset and scenario. Every figure on screen can be traced to a row in those files and recomputed from the published equations.
The full extraction — every hazard layer over every asset — takes about 90 minutes of compute, so the dataset can be refreshed whenever the underlying sources release updates, and new assets can be added from a simple list of coordinates.
No account — just an email. The free tool asks once for your email address and acceptance of the usage policy, then you're in on that device. Full profiles exist only for the premium tier (your own coordinates, portfolios, saved results) — see pricing.
FloodGate is built on peer-reviewed research by Abderrahim Assab (University of Edinburgh Business School): Assab (2025), Journal of Climate Finance and Assab (2024), JRFM.