August 14, 2007
Contact:  Christina S. Johnson, California Sea Grant, 858-822-5334, csjohnson@ucsd.edu

When Are Stock Assessments Worse Because They Attempt to Model Too Much?

fish

Pacific sardines schooling in Kelp Forest Exhibit at the Monterey Bay Aquarium. (c) Monterey Bay Aquarium/Rick Browne

Commercial fishers sometimes disparage stock assessments as being unreliable – bad measures of the status of a fishery, and therefore not a good starting point for setting rules on fishing.

Yasmin Lucero, a NOAA Fisheries/Sea Grant graduate population dynamics fellow, who just earned her doctorate in ocean science at UC Santa Cruz and is now off to Seattle for a postdoctoral fellowship at NOAA’s Northwest Fisheries Science Center, is sympathetic. She has been grinding through the math that supports industry’s (and biologists’) concerns.

In particular, she has been looking at estimates of adult biomass as a function of time from two very simple stock-assessment-like models. One model used two parameters to predict fish biomass; the other used three parameters.

The more complicated model, in theory, should have produced better results, since, in theory, it incorporated more biologic detail than the two-parameter model. Not so. Sometimes it did, but not always.

The take-home message is that “complicated is not necessarily better,” Lucero said. “A very simple model that accurately reflects what you know and don’t know is preferred to a complicated model that adds biological complexity for which you have no data.”

Her project was a digital thought experiment; she can’t make any statements on which fish stock assessments might be off because no specific species was analyzed. The goal, however, is to go in this direction. “We’d like to develop guidelines that would help managers select the right model for each species in question, given the data and biology available,” she said.