This study examines potential interactions among the environmental variables likely to affect larval walleye pollock Theragra chalcogramma feeding in the sea. Walleye pollock larvae were sampled from Shelikof Strait, Gulf of Alaska, and from the eastern Bering Sea, with corresponding environmental data. Variables used in our study were time spent feeding, seawater temperature, light, prey density, wind speed and standard length of the larvae. We applied an additivity test to detect the presence of potential interactions among these variables and adopted an expansion of generalized additive models (GAMs) that allowed the inclusion of interaction terms in a non-parametric regression analysis. After testing all possible 2-way interactions among these variables, we found 4 significant terms: (1) time feeding-standard length, (2) temperature-light, (3) light -wind speed and (4) prey density-standard length. The most influential interaction term was the light -wind speed interaction, which caused a decrease of the model-generalized cross-validation (GCV, whereby lower values indicate more parsimonious models) from 0.0402 (for a model with all variables but no interaction term) to 0.0290 (for a model with all variables and this interaction), and an increase of the model explained variance by 7% (R 2 = 0.89 versus 0.82). This result indicates that the effect of wind speed (turbulence) on larval walleye pollock feeding is dependent upon the amount of light available. This may be due to vertical movements in the water column by both walleye pollock larvae and their prey in response to turbulence. KEY WORDS: Larval fish feeding · Walleye pollock larvae · Generalized additive models
Resale or republication not permitted without written consent of the publisher