Science on the edge of spatial scales: a reply to the comments of Williams (2001)

Guay et al. (2000) evaluated the ability of a numerical habitat model (NHM) to predict the spatial distribution of juveniles of Atlantic salmon (Salmo salar) in a river. The NHM that we used consisted of a hydrodynamic model predicting the physical characteristics (current velocity, water depth, substrate composition) of habitats under any given flow and a biological model assigning an ecological value ranging from 0 (poor habitat) to 1 (excellent habitat) to habitats using their expected physical attributes as independent variables. The main objective of Guay et al. (2000) was to compare the predictions of NHM based on two biological models to the spatial distribution of fish in a river. The new biological model that we developed, the habitat probabilistic index (HPI), predicted a significantly larger fraction of the local variations of fish density (r2 = 0.86) than the traditional habitat suitability index (HSI; r2 = 0.39). Williams (2001) questioned several methodological and fundamental aspects of the work of Guay et al. (2000). The points raised by Williams (2001) about Guay et al. (2000) can be grouped into problems of wording, problems of sampling sufficiency, and problems of spatial scales.