Assessment of the transferability of biological habitat models for Atlantic salmon parr (Salmo salar)

We assessed the transferability of the habitat suitability index (HSI) and the habitat probabilistic index (HPI) between two rivers. Transferability was measured by the ability of HSI and HPI models developed in the Sainte- Marguerite River to predict the distribution of Atlantic salmon parr (Salmo salar) in the Escoumins River. HSI and HPI were based on the pattern of utilization by fish of water depth, current velocity, and substrate size. HSI was developed using the preference curve approach, and HPI was developed using a multiple logistic regression. Predicted values of HSI and HPI in Escoumins River ranged from 0 (poor habitat) to 1 (excellent habitat). Fish density in habitat patches assigned different HSI or HPI values ranged from 0 to 1 fish·100 m -2 . Only HPI adequately predicted local variations in parr density (r 2 = 0.84) in habitat patches of Escoumins River. Our results suggest that HSI is less transferable between rivers than HPI. Differences in substrate size between the two rivers is suspected to impede the transferability of the HSI model. We also argue that the mathematical structure of HPI provides a larger degree of flexibility that