April 23, 1999
Re: PATH Process Workshop
Dear Dr. Foster and Commissioners:
Simple underlying truths can be obscured through immensely complicated processes, particularly when the participants in those processes are carefully shielded from inquiry by those who know the defects in the process. Thus the Commissioners could sit through four hours of presentations and never be told that both the CRiSP and FLUSH models predict higher salmon survival through and around eight dams than from dam removal. This result is based on actual measurements of fish survival in the river, combined with the actual measurements of the survival of fish in barges.
It is only by tacking a series of assumptions about what happens after juvenile salmon leave the Columbia River that the PATH result emerges: dam removal is better for fish than a status quo with transportation. In other words, the PATH results are exactly opposite of the CRiSP and FLUSH modeling results: even though both those models predict higher survival in a status quo with transportation, the additional modeling steps (about which little was disclosed yesterday) produce a higher likelihood of recovery with dam removal.
Obviously, something important is going on in these extra modeling steps, and NMFS has pointed out that one of the driving forces is assumptions about "delayed mortality" from transportation. Researchers have carefully measured the survival of fish released from barges as compared to fish migrating in the river for as long as the radio transmitters can be detected (below Bonneville to the estuary), and find that barged salmon survive at the same or slightly higher rates.
So the PATH results depend significantly on the assumption that transportation kills fish somehow and somewhere we have not yet measured and understood. Those of us who have watched the fishery agencies claim gradually morph their theories from upriver "lethal slackwater pools", to downriver "lethal slackwater pools", into mysterious mechanisms by which fish die months or years after contact with barges or dams are highly skeptical, to say the least. By dividing measured transportation-benefit ratios by roughly a factor of three based on assumptions about "delayed mortality", PATH declares—contrary to thirty years of data measuring the effectiveness of transportation—that transporting more fish reduces the probability of recovery.
The PATH proponents emphasized two points over and over again: (1) there was extensive peer review and participation in the PATH process; and (2) disputes about modeling, and the CRiSP and FLUSH models in particular, do not matter because dam removal is better no matter what models you use.
As to the adequacy of peer review, no one ever saw critical details of the FLUSH model; instead, we are told that "diagnostics" were enough—as if one could tell whether a car had a rotary or piston engine worked by pressing on the accelerator pedal and seeing the car move forward. Mr. Marmorek admitted that the Expert Panel providing peer review was not even given the model estimates of passage survival; much other information was omitted as well. How can peer review mean anything when the peer reviewers fly over at 60,000 feet?
Earlier this week, one member of the "Scientific Review Panel" warned PATH:
". . . after seeing your comparisons of passage model predictions of reach survival to observed reach survivals . . . I no longer trust your assessments about the range of uncertainty in recovery predictions under alternative policies, in particular I do not trust your finding that there is a very high long term recovery probability under the dam removal options (where the passage models become the dominant factor causing differences among policies in predicted performance.)"
No one saw fit to disclose these comments to the Commission.
As to the question of modeling, CRiSP and FLUSH predict only juvenile salmon survival, not what happens to the juveniles in the estuary and the ocean, not the survival of adults ascending the river, and not the production of juveniles from spawning adults. Whether the PATH results are the same with CRiSP and FLUSH misses the point: the real question is whether the PATH results are the same under different assumptions throughout the lifecycle.
Unfortunately, PATH did not consider a reasonable range of assumptions throughout the lifecycle. I have submitted a lengthy series of questions to the PATH managers aimed at uncovering just how contra-factual the assumptions are (attached). The questions were never answered.
For brevity, I will give but one example. The PATH model runs assume that there would be a very substantial increase in adult survival by removing the four Lower Snake Dams: 50% in the case of fall chinook, and a lesser number for spring chinook that is not clearly disclosed in any documentation we can find. No alternative hypotheses were examined by PATH, even though radio-tag data suggests adults move faster upriver through the Lower Snake River than they would through a "natural river", so that adult survival gains may well be nonexistent. If the models were changed to show no benefits to adults from dam removal, the benefits of dam removal would obviously drop significantly—enough to allow the status quo to beat dam removal if "delayed mortality" assumptions are removed as well.
By hard-wiring enough of these kinds of assumptions into the modeling, PATH produces the result that dam removal is better. That result is the product of the assumptions. As Howard Schaller put it: "you can make models do whatever you want them to do". State and tribal fishery agency staff want the dams out, and the PATH process is best understood as a process to make the models support that goal.
Sincerely,
James L. Buchal
Return to Other Salmon Materials
Return to www.buchal.com