PRIORITIZATION OF SITES FOR CONSERVATION OF NATIVE FRESHWATER MUSSELS (UNIONIDAE) IN INDIANA USING SPECIES DISTRIBUTION MODELING

Freshwater mussels in the family Unionidae are an ecologically important and imperiled taxonomic group, suffering declines in abundance and diversity across much of their historical range. Similar to other geographical regions, unionids have declined precipitously in the state of Indiana, prompting a statewide inventory of unionid species occurrence. Some global drivers of unionid declines such as human exploitation, water pollution, and habitat fragmentation, have been identified. However, in the state of Indiana, USA, specific factors contributing to local extirpations have generally not been identified, and underlying environmental relationships are under-described for many unionid species. To improve understanding of environmental drivers of unionid decline and relationships between unionid species distributions and environmental conditions, we used data from detections of live unionids and unionid shell material, and classified observations of eight unionid species (Amblema plicata, Actinonaias ligamentina, Lampsilis fasciola, Ptychobranchus fasciolaris, Villosa iris, Eurynia dilatata, Alasmidonta marginata, Alasmidonta viridis) into total (present = detections of either live individuals or shell material only vs. absent = neither live individuals nor shells detected) and constrained (present = live detections vs. absent = shell detections only) sample distributions. We used these presence/absence data and various environmental datasets to develop predictive ensemble models (based on combined logistic regression and boosted regression tree models) of both total and constrained sample distributions for these species. We aimed to elucidate environmental factors related to species occurrence and ranked residuals from our model predictions to identify sites of high conservation priority for each species. For total sample datasets, watershed size and soil depth to bedrock were important predictors, while percent of impervious surfaces and percent of sandy soils in the watershed were important predictors of constrained sample datasets. While the directions of some effects of environmental variables on unionid presence were inconsistent among species, our models based on total samples displayed consistently negative effects of number of dams and percent of natural land use in the watershed, and consistently positive effects of soil depth to bedrock. Among constrained sample models, we observed a consistently negative effect of percent of impervious surfaces on species presence. We identified priority conservation sites for total sample distributions of all eight study species, and for constrained distributions of seven study species. Our results provide important insight into environmental effects on native unionid distributions in the state of Indiana, and our study demonstrates a unique application of species distribution modeling by incorporating unionid shell observations into presence/absence data and focusing on model residuals for conservation prioritization.