Rigorous meta-analysis of life history correlations by simultaneously analyzing multiple population dynamics models

Correlations among life history parameters have been discussed in the ecological literature for over 50 years, but are often estimated while treating model estimates of demographic rates such as natural mortality ( M ) or individual growth ( k ) as "data." This approach fails to propagate uncertainty appropriately because it ignores correlations in estimation errors between parameters within a species and differences in estimation error among species. An improved alternative is multi-species mixed-effects modeling, which we approximate using multivariate likelihood profiles in an approach that synthesizes information from several population dynamics models. Simulation modeling demonstrates that this approach has minimal bias, and that precision improves with increased number of species. As a case study, we demonstrate this approach by estimating M / k for 11 groundfish species off the U.S. West Coast using the data and functional forms on which pre-existing, peer-reviewed, population dynamics models are based. M / k is estimated to be 1.26 for Pacific rockfishes ( Sebastes spp.), with a coefficient of variation of 76% for M given k . This represents the first-ever estimate of correlations among life history parameters for marine fishes using several age-structured population dynamics models, and it serves as a standard for future life history correlation studies. This approach can be modified to provide robust estimates of other life history parameters and correlations, and requires few changes to existing population dynamics models and software input files for both marine and terrestrial species. Specific results for Pacific rockfishes can be used as a Bayesian prior for estimating natural mortality in future fisheries management efforts. We therefore recommend that fish population dynamics models be compiled in a global database that can be used to simultaneously analyze observation-level data for many species in life history meta-analyses.