Flood risk analysis studies
How deep, and how often — with the evidence still attached
Generated live: points sampled against a size function that refines toward the coastline, triangulated, and shaded by depth.
Coastal surge, riverine, pluvial, and the compound cases where more than one arrives at once. Studies are non-regulatory by default and built to support a regulatory submission where one is coming — a reviewer can follow the chain from bathymetry to floodplain without taking anything on faith.
Start from the decision
A design elevation, a floodplain delineation, an evacuation trigger, and a capital plan need different accuracy. The decision sets the method, not the other way round.
Assemble the domain
Bathymetry and topography from the best available sources — global datasets offshore, survey and topobathy nearshore — with the seams documented rather than smoothed over.
Design the mesh, then check it
Resolution allocated by a size function built against shoreline curvature, bathymetric gradient, and wavelength criteria, then quality-controlled on minimum angle, area ratio, and connectivity.
Force it defensibly
Historical events, synthetic storm suites, or joint-probability treatment where surge, rainfall, and discharge coincide. The choice is stated and tested, not inherited from the last project.
Validate against what happened
Skill assessed against gauge records, buoys, and post-event high water marks — per station and in aggregate, using metrics fixed before the results are seen.
Deliver something re-runnable
Depth rasters, return-period surfaces, and the complete model archive: mesh, forcing, control files, and the scripts that produced every figure.
- Depth and elevation rasters on an agreed grid
- Return-period surfaces and design condition tables
- Model archive: mesh, forcing, control files, run scripts
- Technical report with assumptions and limitations stated plainly
- A return period estimated from a short record carries wide uncertainty. A 100-year surface is a statistical statement, not a promise.
- Models are calibrated against the events that happened, which are not necessarily the events that matter most.
- Resolution buys detail, not accuracy. A finer mesh over poor bathymetry produces a more precise wrong answer.