A panel of experts assembled by NMFS has come out with a scathing criticism of one of the region's main tools for assessing salmon habitat and potential productivity, known as EDT [Ecosystem Diagnosis and Treatment]. The complicated model, developed by Mobrand Biometrics, is being embraced by salmon recovery players throughout the region, from Washington's Methow watershed to Oregon's Sandy River Basin, where the Portland Water Bureau is using it in planning efforts. The Power Planning Council has spent thousands of dollars developing EDT in the regional Framework Process for fish and wildlife, to help inventory and recover fish stocks in 52 Columbia River sub-basins.
But the six professors who make up the Salmon Recovery Science Review Panel say the model is too complicated. The panel, all well-regarded experts in fields like evolutionary ecology, conservation biology and theoretical and mathematical ecology, met in Seattle in early December to hear from regional modelers. Their findings were recently posted on the NMFS Recovery planning web site. "The more complex models become, the more easily one can twist them to do almost anything, and the less reliable they become," the panel said in its report. They singled out EDT with its 45 habitat variables as a case in point. "...the incorporation of so many variables into a formal model renders the predictions of such a model virtually useless."
The panel said EDT "exemplifies how modeling should not be done. It is over-parameterized, includes key functional relationships that cannot be known and cannot be tested, creates a false sense of accuracy, yet introduces error and uncertainty. Its very complexity makes it difficult to determine the effect of various assumptions and parameter values on the model's behavior and relation to data. The attempt at quantification through subjective 'expert opinion' compounds these fatal weaknesses, especially the model's inability to confront and improve with confrontation of data."
The panel said NMFS' CRI [Cumulative Risk Initiative] model, on the other hand, "offers the possibility of direct confrontation with data."
The panel recommended development of various modeling approaches along with experiments--"particularly experiments involving barging, to resolve a key issue--the size of indirect or deferred hydroelectric-induced mortality smolts suffer as they move downstream."
A NMFS review of EDT said that the model was "filling a void" because it's the only tool that is ready to use for summarizing any habitat in terms of "predicted fish numbers." But the reviewers said that important limitations need to be recognized when it is used for recovery planning. "In most regions, the majority of the habitat descriptions that form the foundation of EDT will not be based on actual measured data," the January 29 memo said.
"Instead, EDT will have to be built up from derived data (conclusions derived from measured data) or the best guesses of experts." They said EDT provides a "false sense of precision by using non-linear equations with coefficients involving several significant digits, when in fact there is no way 'expert opinion' could be so precise."
But the NMFS reviewers--Peter Kareiva, who helped developed the CRI model, and John Stein, who heads the Northwest Science Center's Environmental Conservation division--said the limitations of EDT do not imply that it has no value, but "suggest that EDT should not be oversold as a recovery planning tool and that guidelines for its use are needed. In fact, this is true for any model."
http://www.nwframework.org/ecol_work.html
http://research.nwfsc.noaa.gov/cbd/trt/rsrpdoc2.pdf
This page created and maintained by Chehalis River Council
Send comments or questions to the: Chehalis River Council