Invertebrate species are increasingly being considered for listing under various conservation acts, yet basic information needed to guide conservation efforts is commonly lacking. This is especially the case for rare taxa, considering the difficulties in determining their geographic distributions. Detecting rare taxa can be improved by using species distribution models to predict their presence at new locations. We created distribution models for two rare, endemic dragonflies in Ozark/Ouachita region, Gomphurus ozarkensis and Somatochlora ozarkensis. Presence data was gathered from OdonataCentral, an online opportunistic database of odonate occurences as well as from field surveys, museum and photographic records. Environmental predictor variables were downloaded from StreamCat, an online dataset of landscape metrics for streams in the conterminous United States. Presence data were analyzed with Maxent software and predictions were made onto background stream segments within the study region. Streams with the highest predicted probability of occurrence for both taxa were mapped, and a subset were sampled.
Results/Conclusions
Distribution models were built with 56 and 71 presences for G. ozarkensis and S. ozarkensis, respectively, and 12 environmental predictors. Area under the Curve (AUC) values, which compare the false positive rate of prediction from a model with the true positive rate, indicated useful models were created for both G. ozarkensis (AUC=0.852) and S. ozarkensis (AUC=0.852). Percent coniferous forest was found to be important for both taxa, while stream base flow was important for S. ozarkensis and human population density was most important for G. ozarkensis. One new locality was found from predicted locations. Currently, we are ranking streams based on a combined score from RandomForest and Maxent models, and streams with the highest scores will be sampled. We will assess whether this combined model approach will better guide surveys for locating these rare species. Conservation efforts for rare taxa can be misguided due to a lack of basic information regarding these species, however, using a combined model approach has been suggested to maximize available data and help guide efforts to allocate limited resources in a useful manner.