The salmon dilemma: How good is the model?


By Joel Gallob Of the News-Times

March 15, 2006

The possibility of a zero-catch salmon season this year arises not just from what has been happening in the Klamath River in recent years, but also from this year's prediction by the computer model federal scientists and regulators use to predict how many fish may come back up that river (and the other salmon streams, as well).

But how good is that model?

One fisherman at the Ocean Salmon Industry Group meeting in Newport recently noted the model had predicted 144,000 Klamath River four-year-old fall Chinook coming back to that river to spawn in 2002, yet the 2002 post-season estimate came to only 66,000 fish.
Craig Foster, project biologist with the Oregon Department of Fish and Wildlife, admitted he model does not always work that well. "We over-predicted last year," he admitted.

But an under-prediction is as likely, historically, as an over-prediction, he said.

A chart comparing preseason estimates and postseason numbers for the California Central Valley Chinook, given as part of a slide presentation by ODFW Program Manager Curt Melcher, showed that sometimes the preseason estimates were too high, and sometimes too low, for the fish produced by the Sacramento River and other rivers of the California Central Valley. Sometimes the model was quite close to accurate, as in 2004 when the prediction of 870,000 Chinook was only slightly above the actual figure. Sometimes, as in 2001, the prediction (of just over 600,000) was way short of the post-season figure (in that year, 881,000). Other times, as in 2005, the model was way off in the other direction. In that year, it called for about 1,700,000 Chinook to return, and only about 843,000 did.

Since 2000, according to another chart, the Oregon Coast Natural (OCN) coho returning spawner abundance has been consistently underestimated by ODFW. The estimate for the wild coastal coho for 2001 was almost two thirds short of the 163,000 the postseason estimate said returned. In 2002, at about 60,000, it was way short of the 305,000 OCN spawners the postseason estimate found. In 2003, the roughly 115,000 figure was again short of the 279,000 the postseason estimate reported. In 2004, the roughly 150,000 expected was close to the 197,000 the postseason estimate found. And in 2005, the preseason estimate of 150,000 was right on target.

The Klamath model, specifically, said Melcher, had done well in previous years, "but now is doing poorly. The Klamath fish seem to be easier to catch now than they were in previous years."

But the variability in the accuracy of the pre-season forecasts, argued ODFW fish biologist Craig Foster, does not mean that predictive errors by the model - and so by ODFW or the Pacific Fisheries Management Council - necessarily hurt the fishermen.

"It goes both ways," he said. An under-estimate may unnecessarily constrain a season's catch, but an over-estimate may set a quota higher than, in retrospect, would have been biologically warranted.
 
 


+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NOTE: In accordance with Title 17 U.S.C. section 107, any copyrighted
material  herein is distributed without profit or payment to those who have
expressed  a  prior interest in receiving this information for non-profit
research and  educational purposes only. For more information go to:
 http://www.law.cornell.edu/uscode/17/107.shtml



Source:  http://www.newportnewstimes.com/articles/2006/03/15/news/news24.txt