Quantifying Habitat Correlates for Salmon in Coastal Waters

Marine survival of Pacific salmon is highly influenced by the physical and biological environment in coastal waters.  Yet, predictions of salmon survival and spatial distribution are often tenuous due to the dynamic aspect of the coastal environment.  Moreover, managing a migratory species like Pacific salmon is made more difficult by the transient nature of their habitat use.  Expanding on previous analyses, we used generalized linear models to characterize habitats that are correlated with the presence of salmon to improve our understanding of how salmon respond to local cues during migration.  We used catch data from an ongoing surface trawl survey (1998 through 2010; over 1000 trawls) along Washington and Oregon coastlines to model spring Chinook salmon habitat correlates.  Input variables included water depth, secchi depth, salinity, chlorophyll a concentration, and water temperature.  All of the environmental variables except salinity were highly significant in predicting our salmon catch (count) distribution.  However, we found that when spatial information (transect or station) was included, model fit improved significantly, suggesting that environmental factors alone cannot completely explain salmon spatial distributions.  We present multiple ways of incorporating spatial information into the model, as well as some model specifications, such as zero-inflation, that dramatically improved model fit.