COS 61-5
How invaded are we, really?: Aquatic invasive species occurrence datasets under-represent species distributions and misinform predictive models

Wednesday, August 7, 2013: 9:20 AM
L100I, Minneapolis Convention Center
Alexander W. Latzka, Center for Limnology, University of Wisconsin - Madison, Madison, WI
Scott Van Egeren, Wisconsin Department of Natural Resources, Madison, WI
M. Jake Vander Zanden, Center for Limnology, University of Wisconsin-Madison, Madison, WI
Background/Question/Methods

Species distribution models are becoming increasingly common in ecology, enabled by the growth of digital, public species occurrence records. However, these records are sometimes biased in their spatial and taxonomic coverage. With regard to invasive species, these records may be used to estimate range size and frequency of occurrence within a range, both of which are important in determining landscape-level impacts and designing optimal landscape-scale management. These records are also used to construct predictive models and prioritize management efforts. Here, we compare overall rates of invasion and model outcomes based on these records to those based on field surveys. In a field survey of 458 lakes in Wisconsin, we documented the presence or absence of 6 common aquatic invasive species (AIS). For the same set of lakes, we compared our results to invasion records compiled by state management agencies. From both datasets, we (1) estimated the total proportion of lakes invaded statewide, and (2) constructed predictive species distribution models using logistic regression. 

Results/Conclusions

A crucial finding of this research is that biological invasions may be far more prevalent than indicated by species occurrence records. Of the 458 lakes surveyed, 338 were invaded by at least one AIS. However, only 227 of these lakes were invaded in occurrence records. This under-representation was more substantial for small, inaccessible lakes and less well-known species compared to large accessible lakes and AIS “poster-child” species, respectively. By scaling up our survey results, we estimate that approximately 39% of lakes in the state are invaded, a nearly fourfold increase from known invasions in occurrence records (11.9%). Additionally, logistic regressions based on the records were significantly worse at predicting known invasions and predicted significantly fewer lakes to be suitable than those based on surveys for 2 species. When the models were used to make predictions of lake suitability across the landscape, substantial disagreement occurred between records-based models and survey-based models for nearly all species. Therefore, use of existing species occurrence records for assessing and predicting invasions, while informative, needs to be done with caution and in conjunction with more representative datasets.