Hail doesn’t need to touch every vine to wreck a harvest. A single storm, ten minutes long, can undo a year of work in a vineyard, which is why one premium wine-growing region in southwestern France spent nine years building a dataset precise enough to prove whether balloon-based hail mitigation actually works, rather than just assuming it did.
Severe Hail Frequency
-74.3%
Annual large hail events (≥1.9cm), down from 3.89/yr to 1.00/yr.
Mean Hailstone Size
-26%
Across all recorded events, from 1.66cm to 1.23cm.
Statistical Significance
p < 0.0000001
Ruling out a run of lucky years.
Operational Success
90–95%
2023–2024 launch records; 95% for the largest hailstorms.
Key Takeaways
- Annual frequency of severe hail events (≥1.9cm) dropped 74.3% in the treated zone, while a nearby untreated control zone rose 4.7% over the same period.
- Mean hailstone size across all recorded events fell 26% (1.66cm to 1.23cm), significant at p < 0.0000001.
- Nine years of data (2016–2025) and three independent analyses all had to agree before drawing a conclusion.
- Event-by-event radar analysis confirmed the same storms crossed the treated zone — the reduction came from mitigation, not storms missing the region.
- 2023–2024 launch records show a 90%+ operational success rate, rising to 95% for the largest hailstorms.
Setting Up a Study That Could Withstand Scrutiny
In 2021, the region’s wine council deployed Selerys’s LAICO SOBLI system: SKYDETECT radar coupled with smart balloons carrying hygroscopic salt. Rather than comparing a handful of storms before and after, the evaluation ran three separate analyses across 2016–2025 and required all three to agree before drawing a conclusion:
- A before/after comparison inside the treated zone
- A difference-in-differences comparison against a nearby untreated control zone within 15km
- An event-by-event review, cross-checking hail reports against radar storm tracks and actual balloon launch records
Hail was measured against two thresholds: 1.5cm and above, the point where agronomic damage starts, and 1.9cm and above, where crop losses become close to systematic.
The Results
Across all recorded hail events, mean hailstone size fell from 1.66cm before 2021 to 1.23cm after, a 26% decrease. The events also became more consistent: the standard deviation in size dropped from 0.67 to 0.29, meaning fewer of the outsized hailstones that do the most damage.
The result held up under statistical testing at a level rarely seen in field studies (p < 0.0000001), which rules out the possibility that the treated zone just had a run of lucky years.
For severe hail specifically, at or above 1.9cm, the treated zone saw an 11% reduction in mean size while the untreated control zone saw a 16% increase over the same period, a 27 percentage point gap between the two. Annual frequency of large hail events in the treated zone dropped 74.3%, from an average of 3.89 events a year down to 1.00. The control zone, over the same stretch, rose 4.7%, putting the gap between the two zones at roughly 79 percentage points.
| Severe hail (≥1.9cm) | Treated zone | Untreated control zone |
|---|---|---|
| Mean hailstone size, change | -11% | +16% |
| Annual frequency of events, change | -74.3% (3.89 → 1.00/yr) | +4.7% |
In real operational conditions, 2023–2024 launch records showed better than a 90% success rate overall, rising to 95% for the largest hailstorms when balloons launched ahead of the storm’s arrival rather than reacting to it.
Ruling Out the Obvious Objection
The first question anyone asks about a result like this: did the storms just start missing the vineyards? The event-by-event analysis answers that directly. Radar tracks confirmed the same storms that produced the reduced hail did in fact cross the treated zone. They weren’t deflected or diverted around it. The reduction came from the balloons intercepting the storm’s hail-forming process, not from the region getting lucky with storm paths.
That distinction matters, because national hail data for the same period shows the opposite trend almost everywhere else. Hail days at the 1.5cm threshold rose nationally across France over 2006–2024, while the treated zone held a limited, largely flat exposure through the same window, a pattern consistent with active mitigation working against a worsening regional baseline rather than a naturally quiet decade.
How the Balloon Network Gets There in Time
Hail forms fast. A storm cell’s working life is roughly 20 minutes, and the real window to intervene before hail actually forms is closer to 10. SKYDETECT radar, monitoring a 30km radius, flags a developing storm cell early enough for the SOBLI network to respond within that window: Selerys’s balloon systems reach their target in under five minutes from the decision to launch, without a technician needed on-site and without the flight authorization delays that ground or slow other seeding methods.
A single launch station stays fixed at the same location season after season, which fits a vineyard’s needs better than a mobile or aircraft-based system: the same protection returns every year without redeploying infrastructure.
What It Adds Up To
Nine years of data, three independent analytical approaches, and one consistent conclusion: 74% fewer severe hail events, smaller and more uniform hailstones when hail did occur, and a better than 90% operational success rate under real storm conditions. For an industry where a single storm can determine a vintage, that’s the kind of evidence that moves a technology from promising to standard practice.
Frequently Asked Questions
How much did severe hail frequency actually drop?
Annual frequency of severe hail events (≥1.9cm) in the treated zone fell 74.3%, from an average of 3.89 events a year down to 1.00. A nearby untreated control zone rose 4.7% over the same period, widening the gap between the two zones to roughly 79 percentage points.
How was this result validated?
Three independent analyses had to agree before drawing a conclusion: a before/after comparison inside the treated zone, a difference-in-differences comparison against an untreated control zone within 15km, and an event-by-event review cross-checking hail reports against radar storm tracks and balloon launch records, all across nine years of data (2016–2025).
Could the storms have just missed the vineyards instead?
No. Event-by-event radar analysis confirmed the same storms that produced the reduced hail did cross the treated zone rather than being deflected around it. The reduction traces to the balloons intercepting the storm’s hail-forming process, not to the region getting lucky with storm paths.
Is the result statistically significant?
Yes. Mean hailstone size across all recorded events fell from 1.66cm to 1.23cm, a 26% decrease, significant at p < 0.0000001 — a level that rules out a run of lucky years.
How reliable is the system in real operating conditions?
2023–2024 launch records show a better than 90% operational success rate overall, rising to 95% for the largest hailstorms when balloons launched ahead of the storm’s arrival rather than reacting to it.
