Inter-annual stability from ensemble modelling

Ensemble modelling for fisheries analyses is increasing, and may improve on single-model approaches through better representation of uncertainty, reduced potential for bias and greater stability in results. Stability, defined here as deviations from model estimates as each year of data is added, may be due to the use of multiple models (rather than periodic changes to a single-base case model), and from the buffering effect of characterizing the central tendency with a set of models. However, stability against the addition of new data, although logically appealing, has not been explored for fisheries stock assessment. We use the Pacific halibut ensemble as an example, and provide a simple simulation to explore the general behavior of results from an ensemble of models. Counter-intuitively, we found the models in the halibut example showed high temporal correlations among deviations. However, we found that a small number of models with low among-model deviation correlations could show a stability benefit. Among-model deviation correlations may be a useful diagnostic for analysts developing ensembles, or those performing sensitivity testing of single-model assessments.The accepted manuscript in pdf format is listed with the files at the bottom of this page. The presentation of the authors' names and (or) special characters in the title of the manuscript may differ slightly between what is listed on this page and what is listed in the pdf file of the accepted manuscript; that in the pdf file of the accepted manuscript is what was submitted by the author