Improving Distribution Model Precision and Accuracy By Introducing Species Absences From Historical Fish Collections

Most current fish distribution models are based on presence-only records whereas presence/absence data can greatly improve model precision and accuracy. The presence-only records are collected from fisheries field surveys which are conducted routinely at related locations. If we can identify survey locations where fish presence records occur for multiple species, we can derive  absence for other species. Massive presence datasets are required to obtain a realistic absence dataset for a species. By using Geographic Information Systems and fisheries atlas for individual states and continental scale records, we were able to create a comprehensive, historical presence/absence dataset, including all  freshwater fish species within the extent of the contiguous United States. We will outline the automated procedures used to digitize the atlases and create historical species distributions.  When these data are overlaid on NHDplusV2 streamlines, we estimate that presence/absence records exist for over 200,000 inter-confluence stream segments, representing a very high sampling intensity, and an unprecedented resource for improving the precision and accuracy of fish distribution models.