Report of the first working group on Predicting Extinction.

This working group met from 16-18 July at NCEAS. Participating
were:

Ray Hilborn, University of Washington working group coordinator

Loo Botsford, University of California, Davis

Mike Bradford, Canadian Department of Fisheries and Oceans

Dan Goodman, Montana State

Jeff Hutchings, Dalhousie University

Conrad Lamont III, University of Wisconsin
Martin Liermann, University of Washington
Steve Lindley, NMFS Tiburon Lab

Ram Myers, Dalhousie University

Paul Spencer, NMFS Tiburon Lab

Tom Wainwright, NMFS Newport Lab

This was the first of two meetings of this working group whose objective is to combine meta-analysis (the analysis of a large number of data sets to determine the probability of different population parameters) with Bayesian risk analysis to provide a methodology for making predictions about extinction and other risks of low density populations. The working group has three specific objectives; (1) to produce a paper demonstrating this approach for calculating extinction risk for coho salmon in California or Oregon, (2) to produce a paper summarizing the general approach and assessing its applicability to the general problem of predicting extinction, and (3) to review the availability of data sets for taxonomic groups other than fish that could be applied to the problems of predicting extinction.

At this first meeting we began with a review of our individual experience in the various proposed components of the approach which include meta-population modeling, Bayesian analysis, biology of coho salmon, and use of meta-analysis to understand low density processes. Presentations were made on each of these topics.

We then divided into several working groups, one charged with putting together a demonstration Bayesian risk analysis for coho without any meta population structure, a second charged with reviewing and assembling the available data on coho salmon population dynamics in fresh water, a third group exploring the appropriate spatial area to model in our full blow meta- population model, and a fourth group putting together a preliminary outline of a general "Science" paper on the use of the approach to predicting extinction.

The demonstration model from the first working group served to provide a common ground for everyone in the working group and a pilot example for those who have not done Bayesian analysis previously. Several members of the second group have been assembling data on coho from fresh water, primarily experimental watersheds where spawning numbers and subsequent smolts were counted. They found that they had a number of data sets where spawning numbers were very low (a few individual females) and the rate of spawner to smolt production seemed to be consistently about 50 female smolts per female spawner. There was no indication in these data of depensation or other low density effects.

The third group concluded that there would be no watershed in California that would likely have sufficient data for a good detailed worked example and that some southern Oregon watersheds were much more promising.

The fourth group constructed a preliminary outline of the paper.

The following work assignments were made for preparation for the next working group meeting.

- Implementation of the analysis for coho salmon: Lindley, Spencer, Wainwright, Botsford
- Further collection and analysis of coho data: Bradford, Myers
- Continued development of general paper: Botsford, Myers, Hilborn
- Exploration of applicability to other species/taxonomic
groups: Goodman, Hutchings, Liermann, Lamont