Modeling fish biological responses to contaminants and natural variability in estuaries

• Generalized linear models applied to biological responses to contaminants in fish.
• Models differed in explanatory power between biological responses and species.
• Environmental variables and metals concentration were the main predictor variables selected.
• Models constrains related to dataset, sampling design and contaminants characterization.

Understanding the factors that influence biological responses to contaminants has long been a major goal in marine environmental research. Seven estuarine sites along the Portuguese coast were sampled over a year, and different biological responses of Pomatoschistus microps and Atherina presbyter were determined: superoxide dismutase, catalase, ethoxyresorufin O-deethylase, glutathione S-transferase, metallothioneins, lipid peroxidation, RNA:DNA ratio and condition factor K. Generalized linear models (GLM) were developed for each biological variable per species in relation to sediment chemical characterization (metals and polycyclic aromatic hydrocarbons concentration) and environmental conditions (month, site, water temperature, salinity, depth and mud percentage in the sediment). GLM varied in explanatory power and in the set of predictor variables included in the models. Environmental factors were frequently selected as predictor variables. Individual metals concentration and sediment quality guidelines (integrating all metals) were the major contaminants explaining biological variability. Accordingly, models for metallothioneins and lipid peroxidation had highest explanatory power. Species-specific responses and dataset size were the basis of observed differences between GLM for the two species.