Storm-Level Independent Verification of Operational Cloud Seeding: Seven Seasons of Utah's Program, 2016-2025

Bryce Judy 

Operational winter cloud seeding programs claim precipitation increases of 3-15%, but the evaluations supporting these claims are produced by program operators using target/control methods that cannot separate seeding effects from operator storm selection: storms are seeded precisely when atmospheric conditions favor the target areas. This study performs, to the author's knowledge, the first independent storm-level evaluation of an operational program built entirely from public records. We assemble a machine-readable record of Utah's operations across seven winters (2016-17 and 2019-20 through 2024-25): 181 generator sites and 525 seeded storm periods totaling 37,773 generator-hours, transcribed from the operator's seasonal reports (the 2024-25 set published online, six earlier seasons provided by the Division of Water Resources), together with per-storm 700 hPa winds from the HRRR archive and daily precipitation from 546 SNOTEL stations in seven states. Three research designs of increasing selection-robustness are applied. A naive seeded-versus-unseeded storm comparison yields an implausible +335% for the largest program, quantifying the selection artifact. Same-sector near-miss controls reduce the pooled estimate to +24% [+4, +50]. The most robust design, a wind-rotation test in which plume exposure is recomputed from each storm's own wind field per generator, yields, pooled over seven seasons (937 program-events, 143 seeded), +5.8% [-0.0, +12.1] in-plume during seeded storms against a placebo of +3.0% [+0.1, +6.2] in-plume during unseeded storms. The placebo is now resolvably positive, revealing a wind-aligned orographic gradient that the smooth control does not fully absorb; the placebo-corrected seeding effect is +2.7% [-2.1, +8.8] per seeded storm. Seven seasons therefore remain consistent with a small positive operational effect and with zero, and exclude per-storm effects above ~9%, ruling out the upper half of the 3-15% claims. Carried through to water at the State's own baseline, the estimate implies an expected ~95,000 acre-feet per year (38% of the 249,600 claimed) at ~$52 per acre-foot, with a 17% probability that the program removes water from its target areas. All data and code are released.