Assessment of Atlantic salmon (Salmo salar) habitat quality and its uncertainty using a multiple-expert fuzzy model applied to the Romaine River (Canada)

Many tools have been developed to evaluate environmental flows, including physical microhabitat models like PHASBIM and HABSCORE, which require habitat suitability curves. Unfortunately, the models and curves are often used in stream-specific applications and are rarely easily exportable. With the aim to address this shortcoming, we developed several habitat suitability indices sets for three Atlantic salmon (Salmo salar) life stages (young-of-the-year (YOY), parr, spawning adults) with the help of fuzzy logic modeling. Using the knowledge of twenty-seven experts, from both sides of the Atlantic Ocean, we defined fuzzy sets of four variables (depth, substrate size, velocity and Habitat Suitability Index, or HSI) and associated fuzzy rules. When applied to the Romaine River (Canada), median curves of standardized Weighted Usable Area (WUA) were calculated and a confidence interval was obtained by bootstrap resampling. Despite the large range of WUA covered by the expert WUA curves, confidence intervals were relatively narrow: an average width of 0.095 (on a scale of 0 to 1) for spawning habitat, 0.155 for parr rearing habitat and 0.160 for YOY rearing habitat. In addition, Student's t-test showed significant differences in predicted HSI between presence and absence, for parr and YOY, and RM_ANOVA showed significant differences for parr only. When considering an environmental flow value corresponding to 90% of the maximum reached by WUA curve, results seem acceptable for the Romaine River. Generally, this proposed fuzzy logic method seems suitable to model habitat availability for the three life stages, while also providing an estimate of uncertainty in salmon preferences.