328 



Fishery Bulletin 101(2) 



v/inter-run chinook salmon level within 50 years. The low 

 probability of recovery predicted by the density-dependent 

 model is due in part to reductions in productivity at moder- 

 ate population sizes. The median equilibrium winter-run 

 chinook salmon population size, given by (/J-i-zi- C6S)//3 with 

 S = 512,000, is 18,100, which is below the recovery target 

 of 20000. 



Discussion 



Predation by striped bass and effect of stocking 



The results presented here indicate that striped bass pre- 

 dation may be a nontrivial source of mortality for winter- 

 run chinook salmon. According to our analysis, the current 

 striped bass population of roughly 1x10^ adults consumes 

 about 9% of winter-run chinook salmon outmigrants. By 

 comparison, 85,000 northern squawfish consume about 

 11% of juvenile salnionids passing through the John Day 

 Reservoir on the Columbia River (Rieman etal., 1991), 

 based on prey consumption rates and predator and prey 

 abundances. Jager et al. (1997), using a spatially explicit 

 individual based model, estimated that between 13% and 

 57% of fall-run chinook fry were consumed by piscivorous 

 fish in the Tuolumne River, California. The predation rate 

 by striped bass on winter-run chinook salmon juveniles 

 inferred from the time series of their abundances appears 

 plausible in light of these comparisons. If striped bass 

 predation is truly in this range, a significant increase in 

 striped bass abundance could substantially increase the 

 risk of winter-run chinook salmon extinction and reduce 



the likelihood of recovery. A limited program aimed at 

 stabilizing the striped bass population at its recent size 

 might pose an acceptably small risk; the model indicates 

 with 95% certainty that the stabilization program would 

 add less than 3.1% to the baseline extinction risk of 28%. 

 Although this analysis suggests that striped bass preda- 

 tion may be a significant risk factor for winter-run chinook 

 salmon, striped bass eradication would not be enough to 

 ensure recovery of winter-run chinook salmon. In the fol- 

 lowing two subsections, we discuss how data limitations 

 and model uncertainty influence the results and our inter- 

 pretation of them. 



Model uncertainty 



Model uncertainty arises from our ignorance of the exact 

 processes driving population dynamics. Although there is 

 a well-developed statistical basis for model identification 

 and selection (Burnham and Anderson, 1998), different 

 models may fit the data equally well yet make quite dif- 

 ferent predictions (Pascual et al., 1996). In such cases, one 

 should consider a variety of models and ensure that impor- 

 tant conclusions are upheld by all of them (Beissinger and 

 Westphal, 1998). 



Population dynamics and PVA models can be very sensi- 

 tive to the presence and form of density dependence in the 

 model. Because the work presented here was concerned pri- 

 marily with the change in extinction risk posed by a change 

 in striped bass abundance, it is encouraging that the prob- 

 ability of quasi-extinction was not sensitive to assumptions 

 about density dependence. We presented results of both 

 a Ricker-type density dependent model and a density- 



