UW Aquatic & Fishery Sciences Quantitative Seminar

James Thorson


Recent and future developments in geostatistical index standardization: the case for a global bottom trawl database


Governments spend millions of dollars annually conducting bottom trawl surveys of bottom-associated fishes. In many cases, bottom trawl data span many decades, and represent a valuable source of information regarding distribution shifts caused by climate or abundance trends caused by fishing. However, trawl data are often analyzed using design-based estimators that ignore any potential benefit of spatial information. Furthermore, bottom trawl data are often analyzed separately for different surveys, which prevents interpreting data from nearby regions (e.g., fish populations spanning Washington and British Columbia). In this presentation, I outline ongoing efforts to develop a spatio-temporal model and global database for bottom trawl data. I first review recent research regarding the benefits of geostatistical models for estimating an index of abundance relative to conventional design-based estimators. Next, I outline statistical developments and tools to account for “re-transformation bias”, which arises when estimating a nonlinear function of random effects using maximum likelihood methods (e.g., population abundance given spatial estimates of log-density). I show how to deal generically with retransformation bias using computational methods involving automatic differentiation, which is now implemented in Template Model Builder. Finally, I present potential estimators for shifts in the spatial distribution for marine fishes (using West Coast groundfishes as a case study). This application illustrates important differences between model-based and conventional estimators for range shifts, where the latter are likely biased due to shifts in the spatial distribution of sampling. I conclude by discussing ongoing and future research using a growing database of bottom trawl data that currently spans three continents.

Quantitative Seminar Home