Street-level storm-surge flood probability maps from cyclone track ensembles, updated every forecast cycle
G08B 21/10 Alarms responsive to calamitous events, e.g. tornados or earthquakes
G09B 29/00 Maps; plans; charts; diagrams
For a coastal town, each member of a cyclone track ensemble (typically 50) is turned into a simple storm-surge estimate along the coast, which is then spread inland over a high-resolution elevation model. Counting, for each 10 m grid cell, the share of ensemble members that flood it gives a street-level flood probability map. The map is recomputed every forecast cycle (every 6 hours) on an ordinary server and published as map tiles, so local authorities can see which streets are likely to flood rather than one coastline-wide warning.
- Parametric storm-surge models (general technique)
- "Bathtub" inundation mapping on elevation models (common practice in flood studies)
- Ensemble tropical cyclone track forecasts (issued by major forecast centres)
Problem
Official storm-surge warnings are often issued for long stretches of coast and as a single "most likely" scenario. Local authorities in small towns need to know which streets are likely to flood, and how sure the forecast is, to decide evacuations. Full hydrodynamic surge models per ensemble member are too slow and expensive for small towns to run every forecast cycle.
Detailed description
The method takes the ensemble of forecast cyclone tracks (30), each with central pressure and maximum wind radius along the track (32), a coastline split into segments about 500 m long (34), and a digital elevation model of the town at 10 m resolution or finer (36).
For each ensemble member and each coastline segment, a peak surge height is estimated from the cyclone's pressure drop, wind speed and distance to the segment, using a simple parametric surge formula calibrated once against historical storms for that coast, plus the predicted astronomical tide at the time of closest approach.
The surge height at each segment is spread inland: a grid cell floods if its elevation is below the surge height of the nearest coastline segment and it is connected to the sea through cells that also flood (a "bathtub" fill with connectivity). For each cell, the flood probability is the fraction of members that flood it. The result is stored as map tiles in three classes (over 60 %, 20–60 %, under 20 %) plus the full percentage.
The whole calculation for 50 members and a town of 20 km² runs in a few minutes on a single server, so it can be repeated for every new forecast cycle.
Example
A cyclone ensemble of 50 tracks is issued at 00 UTC. For segment 12 of the coastline, 31 members give a surge above 1.8 m at high tide. The harbour street (38), at 1.4 m elevation and connected to the sea, floods in those 31 members: probability 62 %, shown in the highest class. A school two streets inland at 2.6 m floods in 6 members (12 %).
Variants
Instead of the parametric formula, a fast surrogate model (for example a neural network trained on full hydrodynamic runs) can estimate the surge per segment.
Wave run-up and river discharge can be added to the surge height where data is available.
Instead of probability classes, the map can show the expected flood depth, or the earliest time each street is expected to flood.
The same tiles can feed automatic SMS alerts to registered residents of grid cells above a chosen probability.
Drawings
Figure 1 shows a plan view. The solid curve is the main forecast track and the dashed curves are other ensemble members (30). Circles mark track points where pressure and wind radius are given (32). The coastline (34) runs along the bottom. Blocks (36) are buildings on the elevation model; the dashed outline marks cells flooded in most members, including the harbour street (38).