Hierarchical analysis of multiple noisy abundance indices

Fisheries analysts often rely on relative abundance indices for assessing stock status. However, trends in abundance can be difficult to discern when there are multiple indices or when correlation among indices is weak or negative. In this paper, I present a hierarchical framework for analyzing multiple, noisy indices with the goal of estimating a single time series of relative abundance. An implicit assumption is that each index is measuring the same quantity (relative abundance) but that each is subject to process error (attributable to variation in catchability, spatial distribution, etc.) in addition to an estimable level of within-survey variance (i.e., sampling error). I use simulation to explore estimator performance under a number of scenarios, including several that violate underlying assumptions. In general, the hierarchical approach produced estimators with reasonable properties. I illustrate the method with an analysis of seven fishery-dependent catch-per-unit-effort indices of Spanish mackerel (Scomberomorus maculatus) off the US Atlantic coast and provide several suggestions for how this approach can be used in practice.