Case study
The Tana kept rising. The forecast saw it days out.
Published 2024
In late November 2023, El Niño short rains sent the Tana River over its banks at Garissa, displacing tens of thousands across north-eastern Kenya. The build-up was slow and foreseeable: ensemble forecasts carried the signal for days. We reconstructed what a probabilistic, catchment-scale warning would have shown, run by run, as the flood wave routed down the basin.
1 · What happened
El Niño short rains, and a slow-rising river
The 2023 El Niño super-charged Kenya's October–December short rains. Over the Tana basin (34,404 km² draining Mt Kenya and the Aberdares eastwards to Garissa), weeks of heavy rain saturated the ground and filled the Seven Forks reservoirs. The Kenya Red Cross warned of critical Tana levels on 8 November; the river burst its banks around 25 November near Madogo, and Garissa's peak flood extent was reached on 26 November. Across north-eastern Kenya the floods killed scores of people and, by mid-December, displaced more than half a million.
2 · The forecast, run by run
Watch the warning build as the flood wave routes to Garissa
Each ICON ensemble run (40 members, every 12 h) is fed through the Tana catchment model. Drag the slider, or press play, to step from five days out to the eve of the Garissa peak and watch the flood probability, the ensemble hydrograph against flood thresholds, and the warning status change. Select a reach down the main stem, from the Mt Kenya headwaters to Garissa, to read the forecast, thresholds and exceedance at that point and watch the wave grow downstream.
Catchment & reaches
Garissa (basin outlet) · 34404 km² drained · flood levels illustrative, area-scaled per reach.
Forecast issued · Garissa
Chance of exceeding each flood level
≈ 1 of 40 forecast scenarios cross it
Forecast flood hydrograph · 40-member ensemble · Garissa
| Predicted crest onset (Garissa) | Thu 23 Nov, 11 UTC |
| Observed Tana peak · Garissa | Sun 26 Nov 2023 |
| Forecast catchment rain (ensemble median, 24 h) | 6.5 mm · observed 60.1 mm |
Mon 20 Nov, 00 UTC · 144 h out · Garissa. Modelled chance of a ≥2-yr flood: 2.5%. Predicted onset Thu 23 Nov, 11 UTC, vs the observed Garissa peak on 26 Nov.
3 · Forecast vs reality
The probability climbed days before the peak
Probability of a ≥ 2-yr flood vs lead time
How to read the ensemble: probability, not the median
The chart on the left shows one number per forecast run: the chance that flow tops the 2-year flood level at Garissa. An ensemble is 40 parallel forecasts of the same weather. Rather than ask whether the single middle member crosses a threshold, you count how many members do; that share is the exceedance probability. Days ahead the members disagree and the probability is modest, so a single deterministic line can look like a miss. As the event nears, the members converge and the probability climbs. A flood service acts on that rising probability, which is why operational systems such as GloFAS and EFAS verify on exceedance probability rather than a median hit or miss.
4 · What this means
The signal was there, for days. The warning is the product.
Run after run, the ensemble placed heavy rain across the Tana headwaters and the catchment model routed it into an escalating, probabilistic flood warning at Garissa, with days of lead as the wave built and moved downstream. The warning was near-certain days before the flood, and the modelled crest fell within about two days of the observed peak. On a large, slow-responding basin the science offers a long warning horizon; what was missing was a system to turn it into a local, catchment-specific, probabilistic warning people could act on. That is RA-FEWS.
Event Reconstruction Challenges
These challenges are specific to reconstructing a 2023 event from the only public historical archive. None of them apply to live RA-FEWS operation: in production the system ingests native-resolution operational forecasts, uses thresholds calibrated per catchment, and is not constrained by a coarse research archive.
Forecast. DWD ICON-EPS (40 members) from the ECMWF/TIGGE archive, every 12 h across the 120 h horizon; observed rainfall from NASA GPM-IMERG; both run through the operational RA-FEWS catchment model (SCS-CN → unit hydrograph → variable-parameter Muskingum–Cunge routing with channel transmission loss) over the 31 Tana sub-basins to Garissa. The basin spans many grid cells, so the raw forecast is used directly, with no neighbourhood upscaling.
Reservoir regulation is out of scope. The Seven Forks dams on the upper Tana store and release flow; the model treats the basin as natural runoff. At the flood peak the reservoirs were full and passing their inflow, so the timing is representative, but the absolute discharge is illustrative.
Flood thresholds are provisional. The HQ2–HQ50 levels (600–2,300 m³/s at Garissa) are placeholders for readability, not a gauged flood-frequency analysis; per-reach values are area-scaled from the outlet. Read the probabilities as rising flood risk, not calibrated design floods.
Data: DWD ICON-EPS via TIGGE (CC BY 4.0); GPM IMERG, NASA. Event figures from OCHA, IFRC/Kenya Red Cross and contemporaneous reporting (2023–24). Cycles issued after the ~25 Nov onset are omitted from the plot.
References
- OCHA, Kenya: Heavy Rains and Floods Impact and Response (as of 20 December 2023)
- IFRC / Kenya Red Cross Society, Kenya: El Niño Floods 2023 Emergency Appeal (MDRKE058)
- ReliefWeb, Kenya: El Niño Flash Floods Rapid Needs Assessment (November 2023)
- Characteristics, drivers, and predictability of flood events in the Tana River Basin, Kenya (2024)