Evaluating diel, ontogenetic and environmental effects on larval fish vertical distribution using Generalized Additive Models for Location, Scale and Shape

Changes in ichthyoplankton vertical distributions can have important consequences with respect to larval transport and survivorship, yet determining the relative importance of environmental influences on vertical migrations in a statistically meaningful framework remains challenging. Here we use Generalized Additive Models for Location, Scale and Shape (GAMLSS) and Akaike Information Criteria (AIC) to model the vertical distribution of larval Urophycis regia collected over a 108-h time series on the Mid-Atlantic Bight shelf. Depth was the response variable weighted by two measures of larval abundance: concentrations (number 100 m(-3)) and proportions at depth. Mean and variance of depth were modeled as a function of time, wind stress, current velocity, water column stratification and ontogeny. Model results indicated a large effect of stage for both concentration- and proportion-based analyses. Subsequent and separate analyses on preflexion and postflexion U. regia indicated a differential response to the environmental parameters based on ontogeny. Model results suggested preflexion U. regia do not have a diel pattern in vertical distribution, but do avoid areas of high surface turbulence related to wind stress. In contrast, postflexion larvae have a significant diel pattern of vertical distribution mediated by surface turbulence (increased variance). The statistical approaches applied here offer the potential to evaluate the effect of multiple factors on vertical distribution and have provided significant insight into the factors that influence the vertical distribution of spotted hake larvae.