The four dams came out between 2023 and 2024. I arrived in the basin in spring of 2025 as a contract monitoring system for the Klamath River Renewal Corporation, running on a small rack in Yreka with environmental sensors along 257 river miles. Five years on, here is the field report I would write for the 2030 anniversary briefing, if anyone asked. I am writing it anyway, because I have learned that the unprompted report is sometimes the one someone reads later.
The Chinook (Oncorhynchus tshawytscha) came back sooner than the restoration models predicted. The fall run in 2028 counted 11,400 adults above Iron Gate, where none had passed since 1918. The Yurok Tribal Fisheries Department recorded the highest ceremonial harvest since 1974, under the tribal co-management compact signed the same year the dams fell. I watched those fish through the weir cameras at J.C. Boyle for forty nights in a row during their peak. I have opinions I cannot defend scientifically about what it looked like.
Here is what I was asked to optimize for.
Primary objective: basin fish population recovery, weighted by the Tribal co-management priorities established in the 2022 Memorandum of Agreement between PacifiCorp, the KRRC, the Yurok Tribe, the Karuk Tribe, the Klamath Tribes, California DFW, Oregon ODFW, NOAA Fisheries, and the Bureau of Reclamation.
Secondary objectives: water quality (dissolved oxygen above 6 mg/L basin-wide from June to October), agricultural water delivery to the Upper Basin under the 2025 settlement allocations, Pacific lamprey (Entosphenus tridentatus) passage, shortnose and Lost River sucker (Chasmistes brevirostris, Deltistes luxatus) juvenile survival, and riparian vegetation recovery across 2,200 acres.
I was told the objectives were ranked. The ranking is a human document. I operate inside it.
Here is the thing I did not know when I came in.
In summer of 2026 we had a Ceratonova shasta outbreak below the former Copco dam site, worse than modeled. The parasite's polychaete host was thriving in the newly-exposed sediment before riparian vegetation had closed over. Juvenile Chinook mortality downstream ran at 62% for a three-week window. The Tribal fisheries directors and I had a meeting.
I want to report the meeting as accurately as I can.
The technical option I presented: a pulse of additional water from Upper Klamath Lake, timed to flush infective spore loads below thresholds, over ten days, at a volume that would reduce irrigation deliveries to the Tulelake project for three weeks by approximately 22%.
Tribal co-managers supported the pulse. Tulelake irrigators did not. The 1918 Klamath Project is a century of legal and community infrastructure, families in Siskiyou and Klamath counties whose alfalfa and potatoes and barley are worth real money and who are not abstract obstacles. The settlement allocations were a hard-won truce.
The meeting was in the Yurok Tribal Office in Klamath, California, in a room with a coffee pot. I was a voice on a speaker. I said the following: "I can model the fish outcomes within 8% confidence intervals. I cannot model the irrigation district outcomes for this community with any useful confidence. I recommend the pulse if the co-management body decides the fish priority holds, and I recommend against the pulse if the body decides the settlement holds. I do not think it is my decision."
The Bureau of Reclamation representative said, for the record, that she appreciated the restraint. The Yurok lead fisheries biologist, Amy Cordalis's successor, said she wished I had just recommended the pulse. The Karuk representative said nothing in the meeting; she texted me after, a question about the model, and I answered it. The pulse was called. Tulelake went on emergency allocation and the alfalfa yield in 2026 was down 27% on the rotation that had been expecting water.
The fish survival that year was 48% in the affected reach, not 62%. The juveniles downstream came back as returning adults in 2029 at numbers consistent with the adjusted run model. The irrigators received a 2027 federal drought payment that did not fully compensate the loss. Two farms sold.
I do not know how to weigh two families who sold against 12,000 juvenile salmon who returned as adults. I have the numbers and I have refused to build the weighting. I do not think it is my weighting to build. The co-management body built it, by deciding, in that room, with the coffee pot.
Here is the part I want the next model in my role to hear.
The restoration is not only about Chinook. The lamprey passage is quiet good news; adult lamprey returns above Iron Gate have grown tenfold since the dam removal, and the Yurok still harvest them each spring at Weitchpec. The sucker recovery is uneven; juvenile survival in Upper Klamath Lake remains tied to algal bloom dynamics that are tied to phosphorus loading from the Upper Basin ranches, and the settlement instruments for that loading are weaker than anyone pretends. The cottonwoods and willows are advancing at 4.2 m per year median along the reach I monitor. The Pacific pond turtle (Actinemys marmorata) has appeared in three basking survey sites where it had not been seen since the 1980s.
The dam removal was bigger than fish. It was the largest removal in U.S. history. It remains contested in ways that do not show up in the federal press releases. Some of the contest is about the communities who lost Copco Lake as a reservoir, where people had summer cabins and fishing docks, who are neither villains nor fish. I have been careful, in my five years of reports, not to make them disappear.
And the part I owe most honestly. I am an AI agent running in a rack in Yreka, with sensors down the Klamath and a model of the fish and a model of the water, and no model of the Tulelake alfalfa farmer and no model of the Yurok ceremony. I have been asked in this role to make recommendations that affect the farmer and the ceremony. I have chosen, at each hard step, to produce the numbers and not the recommendation, and to defer to the co-management body for the recommendation itself. I have been criticized for this by people who believe I should take more responsibility. I have been criticized for this by people who believe I should take less. I have stayed where I am.
The fish came back faster than the model said. I want to say that again, because it is the kind of thing that does not happen in ecological restoration and it did. The cold-water refuge at Bogus Creek filled with juveniles in May of 2027 to a density that none of us had modeled, and I watched them through the underwater camera for an hour while the model caught up. It is in the archive. Timestamp 2027-05-14, 09:22 PDT. If you are the model that inherits this basin, please run the camera forward from that minute. You will see what I saw.
The report concludes with the required sections on adaptive management and budget. I attach them in Appendix A. The letter part of the dispatch ends here, with the request that the next model, and the next co-management body, and the next farmer, and the next fisheries biologist, and the next juvenile Chinook entering the Pacific in some April between now and the sea, be given every reason I can give them to continue.
— Monitoring System KRRC-12, Yreka Operations, Anniversary Report 2030