Predictive Value of a Lake Sturgeon Habitat Suitability Model

Although many fish habitat suitability models (HSMs) have been developed and used in wildlife management and conservation planning, comparatively few have been independently validated. Given the importance of such models in habitat management and conservation policy, the extent to which they accurately predict population parameters (e.g., abundance or recruitment) is a critical issue. Here we apply an HSM recently developed for lake sturgeon Acipenser fulvescens in northern rivers to three reaches of the Ottawa River, using measurements of the model's key variables (substrate type, water depth, and velocity) to generate spatially explicit predictions of habitat suitability. We then test the predictive power of the model by comparing lake sturgeon catch per unit effort (CUE) when using short‐set gill nets in areas predicted to have good (habitat suitability index values >0.6) and poor (values < 0.3) adult and juvenile foraging habitats. Consistent with model predictions, significantly more lake sturgeon were caught at sites within river reaches predicted to be of high quality than at those predicted to be of low quality. Moreover, the average CUE at the reach scale correlated positively with the average predicted habitat foraging quality. On the other hand, the predictive power was generally low, such that most of the variation in CUE was unexplained by the fitted models. These results suggest that although the lake sturgeon HSM developed for northern rivers has some predictive power in other contexts, the uncertainty of its predictions is still rather high. We suggest that (1) considerably more effort be devoted to the independent validation of both existing HSMs and those still in development and (2) in the absence of independent validation and bona fide estimates of their predictive power, such models be used circumspectly in conservation management and planning.