A Village Lost 300 Homes. Forty-One Generators Were Burning Silver Iodide Upwind.

VILLENEUVE-D'ALARIC, France — Between 15:52 and 21:48 on the evening of 23 August, forty-one ground generators burned silver iodide into the air over the Alaric valley.

Thirty of those burns appear in the operating association's published log.

Eleven do not.

At 19:52 a tornado came off the ridge, stayed on the ground for eleven minutes, and took the roofs off roughly three hundred homes in a village of eleven hundred people. The prefecture records nineteen injuries and no deaths.

What is not in dispute

France has modified its weather in public since 1961.

Ground-based generators, arranged in arcs across the hail corridors of the south, burn a silver iodide solution when a convective forecast crosses a threshold. The network is real, it is subsidised in part, its positions are mapped, and a version of its burn log is published every quarter. None of this is hidden and none of it is new.

What is new is the contract underneath it, which this newspaper obtained on 12 August and publishes today in full.

Clause 9

The operating company works for a consortium of eleven mutual insurers under a nine-year service agreement.

Most of it is what you would expect: coverage arcs, response times, silver iodide consumption, an availability guarantee, a penalty schedule.

Clause 9 is not. It provides a bonus, calculated per hectare of insured area per event, for redirection convective attestée — attested convective redirection — payable on certification by a loss adjuster that a cell tracked away from a covered area.

The clause is nine years old. So far as this newspaper can establish, nobody has ever claimed under it.

The eleven

At 18:31, eleven generators on the southern arc were lit and burned for three hours and fifty-six minutes.

They are in the telemetry. They are not in the published log. The operator explains why below and the explanation is a real one: the published log covers the subsidised association programme, and these were private-contract burns, which have never been logged anywhere a member of the public could see them.

That is a convention, not a concealment, and this newspaper says so. It is also a convention under which the only complete record of what was burning over a valley on the night it lost three hundred roofs sits on one company's server, subject to a ninety-day retention policy, and reached us because a director chose to send it.

The 9.4 kilometres

The operator's own post-event note, written the following morning and given to us unredacted, records that the cell crossed 9.4 kilometres south of the track its 16:00 model run had predicted.

The note offers no explanation for this and does not appear to think one is required. Forecast models are wrong about tracks constantly, by more than this, for reasons that have nothing to do with anybody's generator.

We print it because it is the only sentence in the company's own paper that puts the burn and the track in the same frame, and because a newspaper that had left it out would have been choosing.

What the science says, and it is not what the contract says

The thirty-one randomised seeding trials this newspaper holds at PX-1986 find a median effect on hail energy of about four per cent. In nineteen of the thirty-one the confidence interval crosses zero.

Prof. Aurélie Steenkamp-Loubet, who read the telemetry for us, is blunt: a surface generator releases a nucleant at the bottom of the atmosphere and hopes the updraught takes it somewhere useful, and nothing in that description permits steering a supercell.

The directrice générale of the operating company agrees with her, in writing, in the reply printed below.

So the position is this. There is a contract that pays for moving storms. There is an industry that says it cannot move storms. Both of those are true at once, and the reason they can both be true is that no one has to reconcile them, because the burns that would settle it are not published.

What this desk takes from it

Not that a village was steered into. We cannot show that and we are not going to arrange the page so a reader concludes it for us.

What we take is smaller and worse. A programme that has run for sixty-five years, that is paid by insurers, that holds a clause pricing the redirection of severe convection, publishes two thirds of what it burns and keeps the rest for ninety days on a private server.

The operator has now proposed a combined public log for the third time in five years. The association has refused it twice. If it refuses again, we will print the refusal, and we have been promised it by the person who will receive it.