Temporally varying natural mortality: sensitivity of a simple virtual population analysis and an exploration of alternatives

Cohort reconstructions currently applied in Pacific salmon management estimate temporally variant exploitation, maturation, and juvenile natural mortality rates but require assumed (typically invariant) values of adult natural mortality, resulting in unknown biases in the remaining vital rates and precluding studies of factors influencing adult natural mortality rates. We explored the sensitivity of cohort reconstruction results to misspecification of mean mortality rate and to the amount of variability around the mean. Under the common assumption of adult natural mortality of 20%, error (RMSE) in estimated vital rates is generally small (<= 0.05) when adult natural mortality rates are low to moderate (<= 40%). The greatest error is in maturation rates, as well as relative error in juvenile survival/age 2 abundance. The ability of cohort reconstruction methods to estimate temporal patterns in juvenile natural mortality is adequate (rho > 0.75) except in the case of high adult natural mortality (>= 60%) coupled with high variability (CV >= 0.40). Simulations suggest alternative methods that allow estimation of adult natural mortality rates perform well only when assumptions of shared year and age effects are met and rates are time variant.

Draft manuscript intended for publication in Natural Resource Modeling.