Fine-scale modeling of riverine Odonata distributions in the northeastern United States

The distributions of riverine dragonflies and damselflies (Insecta: Odonata) were modeled at the scale of individual river segments across the northeastern United States, a 784,982 km2 region spanning from Ohio, West Virginia, and Virginia northeast to Maine. The species distribution modeling approach was reviewed with respect to Odonata, and several modeling techniques were compared. Species locality data assigned only to U.S. county is prevalent in invertebrate databases, and it was found that using these data for modeling tends to overpredict the geographic distributions of species. Several techniques to compensate for geographic sampling bias, which is also a characteristic of these largely opportunisticly collected databases, were compared, though the optimum method (thinning data, biased background sampling, or no treatment) depended on the dataset. Watersheds and riparian zones are ecologically relevant for riverine organisms, and models with catchment- and local-scale predictors outperformed models based only on climate. This fine-scale modeling approach is appropriate for the conservation of freshwater diversity, because individual river segments containing suitable species habitat can be identified and prioritized. Climate change is expected to reduce the available habitat for riverine Odonata within the northeastern U.S., though some rivers may serve as climatic refugia, and conservation of these rivers and their watersheds is essential.