UW Aquatic & Fishery Sciences Quantitative Seminar
Approaching a Unified Theory for Fish Recruitment Variability, Assessment Information, and Unbiased Estimation.
Richard Methot and Ian Taylor
Integrated analysis (IA) assessment models commonly estimate annual recruitment as a lognormal deviation from a central tendency calculated from the mean spawner-recruitment relationship (SRR) and the standard deviation of the lognormal distribution (sigmaR). This approach allows the IA models to span long time series that begin in data-poor, lightly fished eras and extend into data-rich, fully exploited eras. The assertion that the parameters of the SRR are stationary (or can be modeled through climate/ecosystem links) provides a robust estimation system. Two simplifications in this approach are that sigmaR is constant and that sigmaR is the correct quantity to use in calculating the offset between the SRR and the expected value for log(recruitment). First, ecological considerations for marine fish suggest that much of the variability in recruitment occurs before the life stages with density-dependent survival. Simulation studies presented here confirm the conclusions of Minto, Myers and Blanchard (Nature, 2008) that this phenomenon causes variation in recruitment to be a function of spawning stock abundance. Second, the current IA practice of using sigmaR as the sole basis of the bias correction produces an over-correction because the assessment data are never perfectly informative about the degree of recruitment variability. A modified procedure that also takes into account the measurement error is evaluated.