Case study · national coastal flood screening
The largest areas of connected coastal flooding are around Ada, the Volta estuary and Keta lagoon. This screening shows where the sea could reach today and with 2050 sea level, so the detailed 2D stage goes where it would change decisions for the people who live there.
Independent consultancy, separate from my river-flood dissertation. A static screening of where the sea could reach, not a 2D flow model. Values are estimates.
1-in-100-year total water level: 1.06–1.20 m above model mean sea level
Window: Volta estuary, Ada and Keta lagoon, connected flooding at 1 in 100 today
57 sections, 566 km of coast: 1-in-100-year total water level
1.06 m1.20 mStatic connected-inundation screening with provisional forcing; terrain FABDEM 30 m (CC BY-NC-SA 4.0), shown partially.
Before the flood: understanding risk
For each section I built a 32-year hourly record of tide, storm surge and wave setup, then fitted an extreme-value model to it. The map shows the resulting extreme water level; the profile beneath reads it along the coast from west to east.
Hover, tab to or choose a section. Levels are relative to the tide-and-surge model’s mean sea level, with no local vertical-datum correction. Section forcing is provisional: all 57 assignments are pending review, and 43 are flagged for node distance.
Water levels: GTSM v3 hourly reanalysis; waves: ERA5 (Copernicus Climate Change Service). Extreme-value analysis, sections and mapping: my own work (2026). Boundaries: OCHA/HDX COD-AB (GSS).
Along most of the coast the 1-in-100-year level varies by little more than a tenth of a metre. What varies far more is the land behind it: low-lying lagoons and deltas flood at levels that steeper coasts shrug off. That is why the screen goes on to map connected flooding against terrain, and why the priority sites for detailed modelling sit at the Volta delta, the Songor lagoon and the Densu delta.
How the screening works, in nine short steps: what each step does in plain terms, how I did it, and what it showed.
Walkthrough
Gauge → model check → waves → extremes → whole coast → 30 m maps
Scroll through the nine steps; each chart draws as you reach it.Choose a step to show its chart. Forcing provisional
Loading the chart for this step. Each step’s result is also given in the text.
Step 1 of 9 · Measured sea level
A tide gauge records the actual height of the sea every hour. It is the yardstick for everything that follows.
My approach
Result
30,254 valid hours. Highest level 2.14 m above gauge zero, on 27 October 2011.
Step 2 of 9 · Model check
A global model (GTSM) estimates tide and storm surge for every stretch of coast. Before relying on it, I compared it with the gauge, hour by hour.
My approach
Result
A close match: r = 0.94, RMSE 0.135 m. In the biggest storms the model runs about 0.17 m low.
Step 3 of 9 · Waves
Breaking waves pile water up against the beach (wave setup). The tide-and-surge model leaves this out, so I added it from ERA5 wave data.
My approach
Result
In the biggest wave event (4 September 2008), setup added 0.11–0.54 m, depending on the slope.
Step 4 of 9 · Extreme values
Six years of records hold thousands of high tides. To estimate rare extremes, I kept only the separate storm peaks above a high threshold.
My approach
Result
166 separate storms in six years at Accra–Tema, above the 98th-percentile level, with a 24-hour window.
Step 5 of 9 · Extreme values
A statistical curve fitted to those storms gives the level reached, on average, once in 1, 10 or 100 years, with a band for the uncertainty.
My approach
Result
1-in-100-year level 1.13 m at all three pilot sites (about 1.07–1.22 m); waves add 0.11–0.12 m.
Step 6 of 9 · Whole coast
I repeated the analysis for 57 sections of coast, each about 10 km, using 32 years of hourly data (1993–2024).
My approach
Result
1-in-100-year level 1.06–1.20 m along the coast; waves add 0.09–0.16 m.
Step 7 of 9 · Whole coast
Those levels are laid over the land to find the low ground connected to the sea: today, with 2050 sea level, and in a stress test.
My approach
Result
132.3 km² today, 182.4 km² in 2050, 963.8 km² in the stress test. Today most is at the Songor lagoon (Ada) and the Densu delta.
Step 8 of 9 · 30 m extents
Coarse national maps blur where water actually goes. Zooming into the Volta delta, the same flooding is shown at 1 km, 90 m and 30 m.
My approach
Result
87 km² at 30 m. The same flooding covers 97 km² of 90 m cells and 239 km² of 1 km cells: about 2.7 times too much.
Step 9 of 9 · 30 m extents
A simple “bathtub” floods every low spot, even lagoon basins cut off from the sea. My method keeps only land the sea can reach.
My approach
Result
Bathtub: 646 km² in this window; connected: 87 km². Nationally 709.8 against 132.3 km². In the stress test the lagoon connects: 864.9 km².
Data: UHSLC Takoradi tide gauge (h231c); GTSM v3 hourly water levels and ERA5 hourly waves (Copernicus Climate Change Service); FABDEM v1-2 30 m terrain (Fathom and University of Bristol, CC BY-NC-SA 4.0), shown partially and for illustration only. My analysis (2026). National forcing assignments are provisional, pending review. Static connected-inundation screening, not a 2D hydrodynamic model.
Ten decisions behind the numbers. Choose one to see what I did, why, and how it moved the result.
Learn and adapt: the future will differ from the past
The same extreme levels, raised by 0.25 m for 2050 sea level and then by a further 0.5 m as a stress test, show where connected low ground would flood. It is a screen of exposure, not a flood forecast.
Static connected inundation at 30 m. Area is on a logarithmic scale so that all three scenarios can be read; hover for area, mean depth and maximum depth.
Terrain: FABDEM v1-2 (Fathom and University of Bristol, CC BY-NC-SA 4.0). Forcing provisional.
Static connected inundation; illustrative windows only, not the full national map. Choose a place and a scenario. Water depth is the extreme level minus ground level, connected to the sea; no flow, timing, defences, overtopping or drainage are modelled.
Terrain: FABDEM v1-2 (Fathom and University of Bristol), CC BY-NC-SA 4.0, shown partially and with the owner’s agreement; commercial use requires a FABDEM licence. Boundaries: OCHA/HDX COD-AB (GSS).
A national screen says where flooding could reach. Where that would change a decision, the next tier is a full 2D hydrodynamic model that represents how water actually moves. For towns and cities, that detailed tier follows the World Bank’s Urban Flood Risk Handbook (Ferguson et al., 2023), “a roadmap for conducting an urban flood risk assessment in any city in the world”.
Only the first stage has been built. No 2D coastal model has been run for Ghana; the later stages are the proposed next steps.
Priority sites from the 1-in-100-year flooded area by section (sections 46–48 and the Densu delta today; the Keta lagoon under the stress test).
I build full 2D hydrodynamic models at 10–30 m resolution. I used TUFLOW and Flood Modeller 2D at Hamstall; HEC-RAS 2D and SFINCS are other options for river and coastal work.
These Ghana coastal results are static connected-inundation screening with provisional national forcing. They show where low ground is connected to the sea at a given level; they do not show how fast water arrives, how long it stays, or what defences and drainage do. A 2D model at the priority sites would.
See the 1D–2D model I built jointly with an Environment Agency colleague at Hamstall Ridware
What these results can and cannot support.
Provisional forcingAll 57 section assignments are provisional and await review; 43 are flagged because the nearest model node is more than 15 km away.
No vertical-datum reconciliationModel mean sea level, the tide-gauge zero and the terrain datum (EGM2008) have not been reconciled, so absolute levels carry an offset uncertainty.
Static, not hydrodynamicNo overtopping, wave run-up, breaching, flow timing or drainage. Connectivity along the Volta channel under the stress test is partly an artefact of the static method.
TerrainFABDEM 30 m (CC BY-NC-SA 4.0), shown partially with the owner’s agreement. Commercial use requires a FABDEM licence.
Hazard onlyThe screen maps hazard and flooded area. Coastal exposure and damage have not been finalised and are not shown.
Data: GTSM v3 hourly water levels and ERA5 hourly waves (Copernicus Climate Change Service); UHSLC Takoradi tide gauge; FABDEM v1-2 (Fathom and University of Bristol, CC BY-NC-SA 4.0; commercial use requires a FABDEM licence); boundaries OCHA/HDX COD-AB (Ghana Statistical Service).
Independent consultancy (2026). Results are modelled screening estimates with provisional forcing; they are not official figures of the Government of Ghana or the World Bank.