Skip to content
No. 040On holdDam safety and extreme rainfallUpdated Nov 16, 2026

For Africa's large dams, whose design rainfall is still often estimated with Hershfield's decades-old statistical method from 30 or so years of station records, do huge ensembles of simulated weather from open machine-learning climate emulators give credible estimates of 1-in-1,000 to 1-in-10,000-year basin rainfall, and how far do those estimates differ from the Hershfield values?

Good it would do
2 / 5
Can AI help
3 / 5
How neglected
4 / 5
Can we check it
2 / 5

Why it's on hold

Park. Credible 1-in-1,000 to 1-in-10,000-year rainfall cannot be verified (the source authors say so, and ERA5 and gauges are weak over Africa), and the Hershfield comparison mixes a small-basin probable maximum with 12,000 km2 grid-cell return levels; a cheap held-out validation study over large basins is possible, but no dam-safety user was found.

Why it's stuck

The new method has only been tried in rich countries. Paciorek and Cooley (arXiv 2510.08893, October 2025) ran 10,560 simulated years from the ACE2-ERA5 emulator to estimate very extreme rainfall, but restricted their analysis to the grid cells contained in the contiguous U.S., and wrote that 'we do not extensively investigate if this particular emulator is fit for purpose'.

What we'd try

Run open emulators (for example ACE2-ERA5, which works at 1-degree and 6-hour resolution) for thousands of simulated years, and compute basin-averaged rainfall extremes over 1 to 90 days for a handful of large African dam catchments big enough to span several 1-degree cells, such as Kariba on the Zambezi. Then compare with Hershfield-style estimates computed from the same basins' records.

We'd look again if

The problem is reframed as a validation study (do open ML emulators reproduce observed 20-100-year multi-day areal rainfall over named large African basins?) and a dam-safety regulator, river-basin authority or lender says what it would do with the answer; or a published evaluation of an open emulator's rainfall extremes over Africa, or public weights for a higher-resolution emulator, appears.

Source: https://arxiv.org/html/2510.08893v1

Think we got it wrong?

Ruled-out problems can come back. Tell us what we missed.

Members can read the full record for this problem, check its sources and comment. Open the member view → Not a member? Here is how to join.