by Art Rangno
How we fooled ourselves about the effectiveness of cloud seeding.
The Colorado River Basin Pilot Project (CRBPP, 1970-75) is still the most ambitious, expensive randomized orographic cloud seeding project in US history. It featured an unprecedented amount of instrumentation, including up to 87 recording gauges, telemetered high-altitude weather stations, some above the tree line, telemetered seeding generators, radars, and, later, aircraft measurements. The target area in the San Juan Mountains of SW Colorado was over 1200 square miles (~3100 km2). Avalanche hazards and snowpack were closely monitored, with additional support from the Soil Conservation Service and the US Geological Survey.
An independent contractor who did not know whether days were seeded or not, measured precipitation. Simultaneously during the CRBPP, research was underway due to a separate million dollar grant to explore possible ecological impacts of the extra snow that was expected to be produced and the possible effects of silver iodide being used to seed.
The CRBPP aimed to demonstrate at a high level of competence that cloud seeding could increase snowfall on a determinant basis. It was going to be the centerpiece of all the Bureau of Reclamation’s cloud seeding efforts, one even having its own documentary film, “Mountain Skywater,” a testimony to the CRBPP’s perceived importance. Obtaining a statistically significant result that seeding had increased snowfall would then support widespread operational seeding after the CRBPP to address looming water shortages in the West.
The CRBPP’s foundation mainly rested on a single, spectacular report of increases (50-100%) in snow produced by cloud seeding obtained in a four and a half season randomized experiment at Climax, CO, (“Climax I”) conducted by Colorado State University (CSU) scientists, hereafter, “the experimenters”). By 1969, preliminary reports indicated that equally as spectacular large percentage increases in snowfall as had occurred in Climax I were accruing in a second, “confirmatory” randomized experiment at Climax (“Climax II”), and from a separate, seasonally randomized one targeting Wolf Creek Pass in southwest Colorado. In all these experiments including the ongoing ones, statistically significant snow increases were linked to same meteorological conditions; high temperatures in the mid-troposphere (>-20°C or –>-23°C at 500 hPa). With the Climax I results in hand, and with the promising preliminary results from the two other experiments, the Bureau of Reclamation’s cloud seeding research division funded the design of the CRBPP to CSU scientists and their consultants in June 1968.
The two on-going experiments finished in 1970, the same year that the first random decisions were drawn in the CRBPP. The statistical results of the newly completed experiments were nearly identical with those in the Climax I experiment and were compatible with the preliminary results.
In the meantime, a key report on ice crystal concentrations and ice nuclei at Climax appeared in 1968 indicating there were too few ice nuclei to produce natural snow in clouds with top temperatures >-20°C. It was asserted by the experimenters that high 500 hPa temperatures had partitioned increases in snowfall because wintertime storm cloud tops hovered around 500 hPa. The 1968 report seemed to explain why seeding had worked with high 500 hPa temperatures. It was when those temperatures were >-20°C that the large percentage increases in snowfall occurred on seeded days. A leap had been made.
Further cementing the view about where cloud tops resided was that in all three experiments the water equivalent precipitation per experimental day dropped precipitously as 500 hPa temperatures rose above -20°C on the control days, but not on the seeded days where the daily amount continued to rise, presumably due to seeding. Confidence was thus further instilled that the temperature at hPa 500 really had captured those at cloud top; the declining amounts of snowfall on control days represented clouds that were ineffective precipitators due to a lack of ice crystals and needed to be seeded.
The final, convincing proof, again, in all three experiments was that cloud seeding had increased the length of time that snow fell and not its intensity. Snow fell for longer periods at the same intensity as natural snowfall on seeded days, a finding compatible with seeding from ground generators using small amounts of silver iodide as was the case in all these experiments. The inference was made that a thick, non-precipitating cloud layer must exist before and after natural snowfall events and that those had been made to snow.
At this point, the reader can perhaps understand why, during the several major planning conferences held by the CRBPP sponsor, the Bureau of Reclamation’s Division of Atmospheric Water Resources Management, that there was no questioning of the underlying foundations of the Colorado experiments; they all fit together so well.
Thus, a great edifice of statistical results with a supporting physical foundation that explained why increases in snowfall had occurred gave confidence that cloud seeding had been proved by the CSU experiments. A scientific consensus formed, validated by the National Academy of Science-National Research Council’s Panel on Weather and Climate Modification who wrote in 1973, ironically, as the CRBPP was plummeting into statistical failure:
“Hence, in the longest randomized cloud-seeding project in the United States (at Climax, CO), involving cold orographic winter clouds, it has been demonstrated that precipitation can be substantially increased and on a determinate basis.”
A newspaper article starts a downhill slide
After a November 1975 newspaper article that began on the front page of the Durango Herald in which a CRBPP meteorologist announced that he was going to reanalyze all the CSU experiments, the “wheels of success” began to fall off the CSU experiments immediately. The experimenters themselves decided to review their experiments (again) after the article came out. It’s quite bad if an outsider finds your errors before you do.
By 1977 the experimenters were acknowledging that, remarkably, both Climax experiments had enjoyed, “lucky random draws” in the category where the huge (50-200%) percentage increases in snowfall had been reported, the very reports that spurred the funding of the CRBPP. The huge percentage increases in snowfall were due to natural storms that were heavier on the seeded days than the control days in the high temperature category. The results of both Climax experiments were verbally retracted at the 7th Conference on Planned and Inadvertent Weather Modification at Banff in 1979. At that same conference this writer showed that the seasonally randomized Wolf Creek Pass experiment (WCPE), had also enjoyed a “lucky random draw” that created the misperception of large percentage increases in snowfall and statistically significant increases in runoffs. The CRBPP was situated in San Juan Mountains of southwest Colorado due to the apparent success of the WCPE and its heavier warm storms compared to those at Climax.
On a remarkable evening at Banff in October 1979, ALL the successes reported by the experimenters on which the CRBPP had been based were swept away.
Airborne measurements by the University of Washington and the University of Wyoming late in the CRBPP found several critical problems, the most critical one being that ice crystal concentrations were far higher in clouds with high cloud top temperatures than had been reported in 1968 by the experimenters, making the clouds largely unsuitable for cloud seeding. The primary pillar supporting the seeding successes had been removed.
Another cycle of reanalyses was to occur in the following years by the experimenters looking to resuscitate their lost cloud seeding successes, and by external skeptics who countermanded the that work. Ultimately, the view that there had been no cloud seeding induced increases in snowfall in any of the CSU experiments prevailed (with rare exception) when the dust settled by 1995.
How could our best scientists, and not just the NAS-NRC Panel on Weather and Climate Modification, be so fooled by ersatz cloud seeding successes backed by equally ersatz experimenter “ripe for seeding” cloud reports?
Answer: The ersatz reports made it into peer-reviewed literature.
Not one person who reads Climate etc. will find these many factors concerning the peer-reviewed literature surprising:
- The inherent, presumed high credibility of peer-reviewed journal literature led to a false consensus on cloud seeding by our best scientists. The gullibility factor.
- The failure of the “firewall” of peer-review that allowed an amazing amount of faulty literature into the journals (dozens of pages): reviewers of the experimenters’ manuscripts did not ask enough of the authors.
- The polarization in the cloud seeding domain and in any other such domain, leads to poor reviews due to partisans who author papers with questionable results and then recommend to a journal editor those reviewers who will support their viewpoint.
- Authors can, and do, recommend to a journal editor those reviewers they don’t want to have review their manuscript. This writer has been de facto blacklisted in the domain of his expertise. (W. R. Cotton, 30 October 2024, personal communication).
- Journal readers failed to comment on suspect results, including those who knew there were problems.
- Insiders that knew there were problems also remained on the sidelines (as did this writer; at least for most of the CRBPP).
- The CSU experimenters themselves did not question their initial success, Climax I, one they had created by choosing control stations mid-way through it instead of beforehand. Once hard-wired, there were no further increases in snowfall during Climax I and through all of Climax II (after errors were corrected) suggesting “cherry-picking” to bring out a seeding effect the experimenters sincerely believed had occurred.
- Randomization, thought to remove storm and experimenter bias, failed on both counts. The former due to the very small sample sizes in which the CSU successes in the warm category were hinged, a generally unrecognized fact.
- The sponsor of the CRBPP did not recognize all the problems that would bedevil them once the CRBPP started. Rather, the agency focused on planning the CRBPP and brought in many outside experts who also did not recognize the problems ahead.
- All the holes in the experimenters’ hypotheses could have been, and most would say, should have been, discovered before the CRBPP began. Reports from an earlier orographic experiment at Steamboat Springs, Colorado, pretty much laid out the problems with the experimenters’ hypotheses before the CRBPP began.
- In June 1968 $468,000 grant for the design of the CRBPP went to CSU and their consultants, an act that made sense. But that grant put enormous pressure on CSU to have the two ongoing experiments, Climax II and the WCPE, be successes. Nineteen years later numerous errors in the Climax II “confirmatory” experiment data were discovered that mainly benefitted seeded days in the category where seeding increases in snowfall were expected. There was no Climax II success following the corrections, nor had there been a “lucky draw.” The latter was created by the errors in storm day assignments.
“Skepticism is the essence of science,” as Physicist Richard Feynman may have said. But it was MIA concerning the seeding successes on which the CRBPP was based. That it was missing was somewhat understandable given the mountain of circumstantial evidence in support of those successes.
What was missing in all the Bureau of Reclamation planning conferences was input from everyday weather forecasters in the CRBPP domain who would have easily pointed out the holes in the experimenters hypotheses, such as that cloud tops did not hover at 500 hPa, or that the water equivalent precipitation per day did not decrease at hPa 500 temperatures above -20°C to -23°C, as examples. Practical weather experience was missing.
Despite its outcome, the CRBPP has an enduring legacy that resides in the comprehensive data collected across diverse synoptic environments in a complex mountainous region, incorporating a dense array of gauges, telemetered weather station data, frequent rawinsonde profiles during storms, detailed airborne and ground-based studies of ice crystal concentrations and characteristics, and thorough post hoc analyses conducted by official evaluators to identify possible seeding effects. Perhaps the most enduring aspect of the CRBPP will be the studies of the San Juan biota 50 years ago in the context of “climate change.”
In summary, the CRBPP had no prospects of confirming that cloud seeding would increase snowfall due to the illusory results it tried to replicate. We were too ready to believe.
I was the only meteorologist/forecaster with the CRBPP for all five of its seasons. Thus, this is an insider’s viewpoint, and due to my experiences during the CRBPP, became skeptical of all cloud seeding literature. It was a fruitful perception. I went on, with Prof. Peter V. Hobbs in tow, over the next 25 years to reanalyze less than six peer-reviewed experiments from the Skagit in Washington State to those in Israel. Not one of those “successes” was what it claimed to be; even the “ripe-for-seeding” cloud descriptions were wrong where they could be checked.
Addendum
So why is so much commercial cloud seeding going in the West during the wintertime right now in view of null results in precipitation on seeded days in two recent major randomized wintertime experiments in Wyoming and Israel[1]? The bottom line is that it’s a win-win situation for both funding agencies and seeding purveyors who get paid. The funders appear to be doing something about a drought to their constituents, and don’t even have to care about what the evidence is concerning their projects. They don’t have to make the purveyors of seeding adhere to rigorous standards of conduct and evaluations of their efforts. Purveyors of cloud seeding and even their academic allies are pushing cloud seeding despite this recent evidence from Wyoming and Israel. Too, most articles in the media are biased toward positive views because the reporters don’t know any better and don’t interview skeptics.
So, when cloud seeding purveyors come “a knockin’” to your county, state, water district, etc., claiming that they will increase your water supply, mitigate a drought:
Caveat emptor!
If one does come “a knockin,” demand these elements, which in effect, forces them to do “the science.” The confident purveyors of cloud seeding services will hardly mind; the charlatans will refuse to implement these, and will give various specious reasons. Namely, turn the commercial project into a viable scientific experiment:
- Select controls or co-variates before the first day of seeding.
- Monitor and save model forecasts of precipitation for the target. Some purveyors may only operate when substantial precipitation is forecast for the target while none is predicted for upwind (usually lowland stations) , namely, when a natural storm looks like a seeding result.
- Randomize the project and make it blind; those who measure the snowfall do not know if a day is seeded. The cloud seeding company decides if a day should be one where a random decision is called based on its expertise.
- Mandate that your project/experiment is evaluated by independent experts, such as those in university statistical departments.
- Provide signs in the target area that alert motorists and skiers that cloud seeding operations are taking place when they do. If a car slides off a road due to snow falling and snow on the road and someone is killed or injured, there may be a claim to be made. The seeding company will be glad to tell you that they made things more treacherous by increasing the snowfall, or starting a snowfall when it otherwise would not have occurred! (Not!!!!)
[1] Seeding increases in snowfall have been reported on several occasions from project sponsored by a hydroelectric company in Idaho. The details, IMO, are not sufficient yet for an external evaluation of the reliability of those results. Morevoer, the clouds in your area may be different than those targeted in Idaho.
Bio notes. Art Rangno was a Research Scientist for 30 years at the University of Washington Cloud and Aerosol Research Group (CARG). He held the position of flight meteorologist or flight scientist on more than 700 research flights. He is one of the world’s most experienced scientists in airborne studies of clouds. He has co-authored many publications with the late Professor Peter V. Hobbs concerning cloud microstructure, and in particular, on the origins and concentra- tions of ice particles in clouds. Art has looked closely at a series of cloud seeding experiments published in the peer-reviewed literature and found that they were not the successes they were reported to be by the experimenters who conducted them. In 2005 Art and Peter Hobbs received a prize from the World Meteorological Organization and the United Arab Emirates for their work in weather modification and their critiques of cloud seeding experiments.
The post Cloud Seeding History: Looking Back at the Colorado River Basin Pilot Project appeared first on Climate Etc..
via Climate Etc.
August 8, 2025 at 05:38PM
Iowa Climate Science Education 