| Summary: | Invasive species are a major cause of biodiversity loss in North American freshwaters. Once established, aquatic invasive species (AIS) can restructure food webs and change the abiotic or physical conditions of a water body. The ability of species distribution models (SDMs) to predict habitat suitability for AIS is of interest because they can be used to identify what species are likely to establish if introduced. Habitat suitability from SDMs could be used as a screening risk assessment to identify which species require more detailed risk assessments. However, SDM performance in areas that are spatially disjunct from their training has been understudied, specifically for aquatic species. I conducted two studies to assess the applicability of habitat suitability predictions created by SDMs to identify potentially invasive fish for further risk assessments. First, I compared the performance of three different SDM techniques (BIOCLIM, DOMAIN and, MaxEnt) for five different freshwater fishes (common carp Cyprinus carpio,burbot Lota lota, northern pike Esox lucius, lake trout Salvelinus namaycush, and brown trout Salmo trutta) when models were trained east of the Atlantic Ocean and evaluated in North America. DOMAIN models were best able to predict the potential distribution of fishes, but still broadly overpredicted the realized distribution of fishes. In areas with habitat conditions that were novel compared to training habitat conditions, predictions of habitat suitability were unreliable for all models. Based on the information from my first study, I applied the DOMAIN SDM technique to 44 fishes considered an invasion risk to the Laurentian Great Lakes to determine if habitat suitability could be used to identify fishes that are unlikely to invade. No fishes could be removed from consideration as potential invaders based solely on habitat suitability because of large areas of novel habitat conditions. Using habitat suitability as a screening measure to select fishes for further assessment is a questionable process due to uncertainties when predicting in novel habitat conditions and broad overpredictions of fishes’ realized distributions. Instead, trends in habitat suitability should be used to help identify where pathways intersect high habitat suitability for many species. M.Sc.
|
|---|