Modelling Site Occupany By Spotted Gar in Lake Erie Wetlands While Accounting for Imperfect Detection

Determining which sites are occupied by a species and the characteristics of a site that influence its occupancy are important steps in correctly identifying and protecting the critical habitat of a species at risk.   By definition, species at risk are rare and, thus, often may be difficult to detect.  Failure to account for detection probability when detection is less than perfect (detection probability < 1) will lead to underestimation of site occupancy and may bias estimates of the effect of site covariates.  In this study we have estimated detection probability of Spotted Gar, a Threatened species in Canada, and modelled site occupancy based on detection probability and sample site covariates to determine critical habitat and to predict distribution and potential colonization of currently unoccupied Great Lakes wetlands in Canada.  Repeated sampling of multiple sites in three coastal wetlands of Lake Erie was conducted using fine-mesh fyke nets.  Detection probability and site occupancy were estimated using the software PRESENCE v 6.4 and Akaike Information Criteria scores were used to rank candidate models to determine which covariates influenced site occupancy.   The highest weighted models were used to determine the potential suitability of sites in Cootes Paradise Marsh for establishment of Spotted Gar.