Pragmatic approaches to modeling recruitment in fisheries stock assessment: A perspective

Our tasks in performing stock assessment are to estimate what is in the population now, what is likely to be there in the near-future, and how we should best extract yield without imperiling future reproduction or yield. There are many considerations for modeling recruitment in a stock assessment, and I walk through the typical sequence of decisions an analyst must make when setting up their stock assessment. I begin with the deceptively simple decision of the first modeled age, progress through the treatment of fecundity, and into the important decision of parameterizing the stock recruit relationship (SRR). Subsequent assessment steps such as projections and estimation of reference points build on these earlier decisions. The consequences of many of these decisions are amplified or obviated, depending on whether or not a null SRR model (mean or median with deviations) is implemented. Examples of expanding or attenuating consequences are provided within each decision. The potential for misspecification related to these decisions is noted, and advice to reduce that misspecification as well as suggested diagnostics are offered. As a high level summary, I recommend fitting the null SRR, using weight as a surrogate for fecundity and conducting sensitivity analysis if there are data suggesting significant departure from isometry. If estimating a non-null SRR, move away from the steepness parameterization (return to the original alpha,beta parameterization), and do not begin at age 0. Lastly, exercise caution when incorporating covariates -they may be hitchhikers posing as drivers.