Solveig Marchetti-Rand Randomised Her Own Rain Programme. The Result Closed It.

GREAT FALLS, Mont. — Four per cent, with a confidence interval running from minus five to plus thirteen.

That is what six randomised winters of cloud seeding over the Rocky Mountain Front returned, and Solveig Marchetti-Rand printed it on the second page of her own programme's final report, over her own signature, in January 1989.

She had directed the programme for twenty-one years. The legislature struck its appropriation five months later. Thirty-one people lost their jobs on 30 June, and she countersigned every separation notice, including the last one, which was hers.

She died at home in Great Falls on 2 August, aged 91.

The envelopes

In the autumn of 1982 she told her advisory board that the operational programme would from then on run as a randomised crossover. Every flight would seed or not seed according to a sealed envelope, and the envelopes were cut in the state statistician's office from a published table of random digits, in a room she was not allowed to enter.

The board voted five to two against. Her enabling statute gave the director operational discretion, and she used it.

The design fit on one sheet. The pilot carried the envelope on his knee-board and opened it after take-off, once the aircraft was in cloud and the day had already been judged seedable on the meteorology alone. Ridge generators ran to the same envelope. Forty-two gauges in the target basin and twenty-one in the control were read by staff never told which flights had been which.

She also signed an undertaking, still bound into the back of the report, not to look at any result until all six winters were flown. There were no interim looks.

The number

One thousand and sixty-one flights reached the decision point between the winters of 1982–83 and 1987–88. Seventy-one were later excluded by her, for reasons below. That left 990 — 494 seeded, 496 not.

The seeded basins caught four per cent more precipitation. The interval ran from minus five to plus thirteen: the trial could not rule out that seeding had done nothing, and could not rule out that it had done rather a lot.

Set against the thirty-one randomised trials in our vault, PX-1986, it is unremarkable — wide, centred a little above zero, indistinguishable from most of the field. That is the ordinary condition of the science. It was not the condition of a programme that had been telling farmers it worked since 1961.

The seventy-one

In two winters the envelopes were got at. Delmar Ochoa-Tiedemann, who flew 388 of the sorties, says ground crew opened some at the morning briefing in 1984–85 and again in 1986–87 — "somebody wanted to know whether it was worth driving up to the generators in the dark."

She excluded those flights and listed them in Appendix C by date and tail number. Sixty-one were seeded flights, and on the raw gauge figures they were wetter than the ones she kept. Excluding them made her own result smaller, and nobody outside the programme would have known if she had not.

The board chair's objection

Harlan Vosberg-Pace, who chaired the board and voted against, is 88 and has not changed his mind. His argument is not a stupid one and this newspaper will not pretend it is. Farmers in four counties paid a levy for a service. Randomising meant withholding that service on roughly half the good days, for six years, without telling them. "You can call it a trial," he said. "They bought rain."

She answered him in writing, in the two sentences that are the closest thing to a manifesto she left: that the levy bought a programme, not a promise, and that a service nobody has measured is a service nobody can price.

Two pages about grass

From 1989 until 2011 she inspected rain-gauge exposure for a river-basin authority in the Missouri headwaters. Four hundred and ten gauges. Grass height, fence lines, the distance to the nearest cottonwood.

In her first autumn out of work she wrote a two-page siting standard for field staff, amended once, in 1996. We have found it copied into fourteen state manuals.

The eighty-eight words

She discussed the trial on the record exactly once. In 2004 a graduate student posted a questionnaire to nine hundred former weather-modification staff; two hundred and fourteen came back, and nobody has ever cited it.

Under "Anything else you would like us to know?" she wrote eighty-eight words. The last of them: "I would like it recorded that I have been inspecting gauges since 1989 and that the gauges are worse than the clouds."

What this desk takes from it

The programme came back in 1993 without randomisation and has seeded every winter since — thirty-three seasons, currently $2.1m a year. No controlled trial has been attempted in the state in the thirty-eight years since her last randomised flight.

Her daughter's letter is below and is harder on us than on the legislature. We asked eleven questions about the winters and one about the gauges, which is the ratio her mother predicted.

The obituary she would have wanted is two pages long and concerns the siting of rain gauges. This is the other one.