After the July 4, 2025 Guadalupe River flood killed about 135 people, millions were told that cloud seeding did it. The factual record — assembled from a state permit file, an operator's flight log, and a regulator's review — was that the last seeding mission in the region had released about 70 grams of silver iodide into two small clouds near Runge, Karnes County, two days earlier and roughly 150 miles from Kerr County, and that the clouds dissipated within hours RECORD. The flood came from the remnant moisture of Tropical Storm Barry feeding a mesoscale convective complex that dropped 10 to 20 inches of rain MEASURED.

That dispute was settled by exactly the kind of analysis this page describes: where and when did seeding actually occur, and what could that much material physically do? Seeding leaves fingerprints in five independent layers of public evidence. Here is each layer, what it can prove, and — just as important — what it cannot.

1 · The registry — was anyone even operating?

Before a gram of silver iodide flies in the United States, there is paper: a NOAA Form 17-4 filed at least ten days ahead, a state permit with a defined target area, and in most western programs a published map of ground generator sites RECORD. The first forensic question is always answered here: was any permitted operation within plume range, in season, at all? For most of the country, most of the time, the answer is no — and the analysis is over before it starts. See The Records.

2 · The flights — where was the aircraft?

Seeding aircraft carry transponders like everyone else. Their registrations are in the FAA's public registry (Weather Modification Inc.'s Piper Senecas fly out of Fargo; the UAE's King Airs out of Al Ain), and community ADS-B networks — adsb.lol, ADS-B Exchange, OpenSky — record their tracks, with full daily history dumps published openly MEASURED. A seeding sortie has a distinctive fingerprint: repeated straight parallel legs flown crosswind at the altitude where the cloud is −8 to −15 °C, by an aircraft registered to a weather modification operator. Some Texas districts publish their flight tracks outright. Mountain terrain can shadow receivers, so a missing track is weaker evidence than a present one — a rule that recurs in every layer.

Live on this site: the fleet watchlist — /data/registry/fleet.json, 24 aircraft with Mode-S hex codes, the join key to the public ADS-B record. It took three passes of forensics to build: the operators that register under their own names (WMI's Fargo Senecas, the Texas association fleets, SOAR, a DRI drone); the King Airs WMI flies through holding companies at its own street address (grep the registry by address, not name); and a contract tail confirmed from a photograph in Utah's own annual report — NAWC owns no aircraft and leases SOAR's Cessna 414 for the Northern Utah aerial program. The remaining gaps are themselves findings: Idaho's reports never print a registration, and Rainmaker's seeding drones fly under Part 107 with no Mode-S transponders at all — the newest aircraft in the industry are the least trackable.

3 · The environment — could seeding even have worked?

Glaciogenic seeding requires supercooled liquid water in a cloud whose top sits roughly in the −8 to −25 °C window, with flow to carry the plume over the target. Those conditions are checkable after the fact from radiosondes, icing pilot reports, and model fields (HRRR cloud liquid water at sub-freezing levels) MODELED. A claim that seeding produced weather in an atmosphere that had nothing to seed refutes itself — this is the cheapest and most decisive forensic test, and it is the one the Seedability section will run continuously.

4 · The radar — is the track actually visible?

The SNOWIE experiment settled that seeding can be seen: in January 2017 over Idaho's Payette basin, aircraft AgI legs produced sets of parallel reflectivity lines — the famous zig-zag signatures — trackable for hours as they drifted downwind, and yielding roughly 100 to 275 acre-feet of snow per case when integrated (French et al. 2018; Friedrich et al. 2020, PNAS) MEASURED.

But the detectability theory that followed (Zaremba, Rauber et al. 2023, J. Appl. Meteor. Climatol.) is the honest half of the story: a seeding signal needs roughly a 5 dB reflectivity enhancement over the natural background to stand out, and in SNOWIE the lines appeared only in the cases with weak natural precipitation. Seeding into an already-snowing storm — the normal operational mode, because that is where the supercooled water is — is generally radar-invisible. Dual-polarization moments (a pulse of KDP and ZDR from the burst of small pristine crystals AgI creates) can strengthen an attribution, and satellite cloud-top phase can catch glaciation tracks in thin supercooled decks. No automated seeding-track detector for NEXRAD exists in the published literature; building one — as a confirmation tool primed by flight tracks and generator plumes, never as a free-running oracle — is on this site's roadmap.

Geometry honesty: KCBX (Boise) sees Idaho's aircraft seeding levels well; KMTX on Promontory Point overshoots the Wasatch valleys by 1–4 km, so Utah's ground-generator plumes are mostly below the beam; the San Juan programs are effectively unradared. Any radar claim on this site will carry the MRMS Radar Quality Index as a mask for where the mosaic is actually blind.

5 · The ground — did anything land?

The last layer is what reached the surface. Gauge-corrected QPE (MRMS, Stage IV), SNOTEL stations sitting exactly in the mountain target areas radar cannot see, and CoCoRaHS observers give the accumulation record MEASURED. And there is a gold standard: silver in the snowpack. Trace-chemistry studies (Boise State / Idaho Power, thousands of ICP-MS samples) detected silver enrichment in 88% of ground-generator cases within 70 km of the generators — at parts-per-trillion levels, far below any toxicity threshold but analytically unambiguous — and, strikingly, in zero of the aircraft-seeded cases MEASURED. Chemistry closes the chain, and it also bounds it: past about 80 km, nothing is detectable, which is itself forensic information when someone blames a program 150 miles away.

Registry first — the NOAA filings and state reports normalized into a geocoded database of programs, target areas, generator sites and operator aircraft. Then flights — a watchlist of the seeding fleet against open ADS-B history, with a classifier for the seeding-leg pattern. Then environment — the seedability record, so every claimed event can be checked against whether the atmosphere was seedable at all; first run at season scale in Case 003. Then radar — line detection on the archived Level II data downwind of tracked flights, dual-pol overlays, MRMS accumulation and quality masks. Then verification — SNOTEL, CoCoRaHS and gridded QPE deltas, with the silver-in-snow literature as the chemical anchor. Worked public case studies at each step; no free-running "detector" ever published without its error bars.

Prior art, credited: cloudseedwatch.com correlates known seeding aircraft with weather heuristically; state pages publish reports; academic pipelines exist for single campaigns; and Bryce Judy's seeding-verification built the reproducible statistical rung — public-data causal designs with built-in placebos, adopted here as Case 002. Nobody yet combines the registry, the flights, the radar, and the verification under one standard of evidence. That is the lane.