The use of generalized additive models to examine relationships between environmental variables and commercial catch rates

No abstracts are to be cited without prior reference to the author.

Generalized additive models (GAMs) have become an accepted method for assessing non-linear effects of factors on catch rates of commercial species. Catch data reported by the University of Massachusetts School for Marine Science and Technology study fleet from 2006 to 2009 were standardized to catch per unit of effort for Atlantic cod (Gadus morhua), haddock (Melanogrammus aeglefinus), and windowpane flounder (Scophthalmus aquosus). GAMs were employed to determine how fluctuations in catch were influenced by environmental conditions. Explanatory variables included year, season, time of day, latitude, longitude, depth, and bottom temperature. Models were built with stepwise forward selection based on the Akaike information criterion value and deviance explained. Successive models were tested for significant differences with a Chi-square test. Results indicated that spatial variables described the majority of the explained deviance. Other variables were significant, but the contribution to the explained deviance was small. A disadvantage of having a large number of explanatory variables is that an optimal model may be selected in which many variables have significant effects. Another possible artefact of including many explanatory variables is that correlation may occur, which affects model fit. While GAMs can characterize spatial and temporal patterns in catch rates, selection of explanatory variables has implications for the results. If identifying trends in catch rates or shifts in species abundance are an intended outcome, then certain variables such as depth and water temperature may need to be included. Despite these technical challenges, GAMs appear to be a robust method for identifying factors associated with catch rates.