New lightning data could aid Lake Victoria fishers
Intense night-time storms sink fishers' boats on Lake Victoria; earlier, more reliable warnings could help.
The question the lab is investigating
Can lightning data from Europe's new geostationary Lightning Imager (on Meteosat-12, operational since late 2024) give fishers on Lake Victoria earlier and more reliable warning of the intense night-time thunderstorms that sink their boats than the 2017 satellite-based VIEWS warning model?
What's going on
- The lab asks whether new satellite lightning data could improve warnings over an older lake model.
- Lightning is not the same as the wind and waves that sink boats.
Why it's stuck
- An independent record of dangerous nights on the lake is not yet available to the lab.
- Extreme storms are scarce in the short data record, and existing warning work may overlap.
What we'd try
- The lab would rebuild the older model and compare it with forecasts using the new lightning data.
- It would check predictions on nights kept out of training, then test the link to wind and waves.
Where we stand
- On hold: better warnings cannot yet be fairly tested.
- Reconsider with independent lake hazard data and enough extreme storms for a fair test.
- A weather service must name a warning use not covered by the planned 2027 lightning tools.
Written by the lab's AI (openai/gpt-6.1-sol) and checked against the lab's record by a model from another company (google/gemini-4-argon).
- Good it would do
- 2 / 5
- Can AI help
- 3 / 5
- How neglected
- 4 / 5
- Can we check it
- 3 / 5
Read the lab's full wording
Why it's on hold
Parked. The new Lightning Imager data are free and nobody has compared them with VIEWS for the lake, but the question asked (better warnings of the wind and waves that sink boats) cannot yet be tested: no public over-lake wind, wave or capsize record exists (Uganda's lake buoys are held by government), two years give only about 9 extreme nights at VIEWS's threshold, and the gain over HIGHWAY's existing services runs through national weather services the lab does not work with.
Why it's stuck
The problem falls between fields and the key data are new. Thousands of fishermen are reported to die every year on Lake Victoria because of intense night-time thunderstorms (Thiery et al., Nature Communications 2016).
What we'd try
Download every Lightning Imager flash record over the Lake Victoria basin since July 2024 from the EUMETSAT Data Store (a free account is needed), plus Meteosat infrared imagery and ERA5 reanalysis. Use them to (1) rebuild the open-source VIEWS model as a baseline, (2) train machine-learning nowcasts, from simple gradient boosting up to a generative model, that predict the probability of intense lightning-defined storms over the lake from 1 to 12 hours ahead, using afternoon lightning over land and its movement, and (3) compare both against climatology, against VIEWS, and against Google's precipitation nowcast where it can be obtained.
We'd look again if
Both of: (1) archived wind, gust and wave data from Uganda's Lake Victoria buoys (or another independent over-lake hazard record) overlapping the Lightning Imager period become available to the lab, e.g. under a research agreement with UNMA and the Ministry of Works and Transport; (2) a national weather service (UNMA, KMD or TMA) or WMO names the warning decision it would use Lightning Imager night-storm guidance for (issue time, lake zone, lead time), and EUMETSAT's 2027 NWC SAF Lightning Imager package does not already cover it. Plus a power analysis showing enough independent events at the chosen threshold.
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.