We evaluated the ability of numerical habitat models (NHM) to predict the distribution of juveniles of Atlantic salmon (Salmo salar) in a river. NHMs comprise a hydrodynamic model (to predict water depth and current speed for any given flow) and a biological model (to predict habitat quality for fish using water depth, current speed, and substrate composition). We implemented NHMs with a biological model based on (i) preference curves defined by the ratio of the use to the availability of physical conditions and (ii) a multivariate logistic regression that distinguished between the physical conditions used and avoided by fish. Preference curves provided a habitat suitability index (HSI) ranging from 0 to 1, and the logistic regression produced a habitat probabilistic index (HPI) representing the probability of observing a parr under given physical conditions. Pearson's correlation coefficients between HSI and local densities of parr ranged from 0.39 to 0.63 depending on flow. Corresponding values for HPI ranged from 0.81 to 0.98. We concluded that HPI may be a more powerful biological model than HSI for predicting local variations in fish density, forecasting fish distribution patterns, and performing summer habitat modelling for Atlantic salmon juveniles.

Historical changes in the distribution of herbaceous wetland plant associations were inferred from the hydrological regime of Lake Saint-Pierre, a 312 km 2 broadening of the St. Lawrence River (Quebec, Canada), to assess the cumulative effects of human interventions and climatic variability. Relative abundance index (height · percent cover) of wetland plants in 630 field quadrats sampled at 13 sites (1999)(2000)(2001)(2002) were used to derive a model predicting the occurrence of nine herbaceous plant classes with a 71% (24-84%) accuracy. Wetland types included seasonally dry (meadows), mudflats and wet (low marshes and submerged) assemblages. Over the 1961-2002 period, the total surface area of Lake Saint-Pierre herbaceous wetlands ranged between 11 (in 1972) and 128 (in 2001) km 2 and was negatively correlated (Spearman r ¼ )0.86, p < 0.0001) to average water level during the current growing season. Within-season variability and level conditions over the previous season defined 5 marsh assemblages characterized by different species composition, relative abundance and diversity. Significant hydrological variables included quadrat elevation, water depth, number of days flooded and depth variability experienced over the current and/or previous growth seasons. The hydrological model suggests that for a given level, wetland plant assemblages differed markedly whether the multi-year sequence of water levels was rising or falling. Lake Saint-Pierre alternated between three broad-scale wetland configurations, dominated by meadows and open marsh with floating-leaved vegetation (in the 1960s), scattered tall Scirpus marshes (in the 1970s and early 1980s) and closed marsh with aggressive emergents (since 1996). The strong response of Lake Saint-Pierre wetlands to hydrological conditions in the current and previous growth seasons underlines their vulnerability to future water level variations resulting from regulation and climate variability.