Yellow Lampmussel stacked species distribution model results across the Atlantic Slope, U.S.

Species distribution models are useful for wildlife management by identifying suitable habitats and communities for focal species. Identifying these key relationships may allow for predictions of a species in undersurveyed or unmanaged areas, providing a way to fill data gaps and save resources. However, when species or community distribution models are limited in spatial extent (i.e., to a state or other locality), results may not be applicable across an entire range. Flora and fauna experience different environmental and biotic relationships across biogeographic spatial gradients, making range-wide models important for regional management and status determination. In the study that accompanies this data release, we explore how inferences differ between range-wide versus spatially-restricted distribution models of the Yellow Lampmussel (Lampsilis cariosa), a U.S. federal At-Risk freshwater mussel species, and the community of fishes and other mussels it occurs with using stacked species distribution models (SSDMs). We collated co-occurring freshwater mussel data and host fish data from across L. cariosa’s range and made predictions at the range-wide (n = 16 states; ~750,000 km^2) and, when possible (n ≥ 50 observations), state (n = 5 states; ~100,000 +/- 30,000 SD km2/state) spatial extents. Our state-based SSDMs had similar spatial predictions to our range-wide model, but different relationships between abiotic and biotic variables and L. cariosa. These different relationships are likely due to the “niche truncation” effect of SSDMs across smaller spatial extents, or local responses to the environment that are nonrepresentative across space, and the uneven availability of host fish and co-occurring mussel distribution data. Our range-wide model predicted L. cariosa presence in Vermont, New Hampshire, and Delaware, three states where L. cariosa is presumed to not occur (Vermont) or is presumed extirpated (New Hampshire, Delaware). By avoiding effects of niche truncation, allowing predictions into undersurveyed areas, and providing a range-wide model to integrate with future modeling or survey efforts, our study supports the value of range-wide modeling efforts and applications for management. We grouped our model predictions from the range-wide and state-based models to three different hydrologic unit code (HUC) watershed levels: HUC8, HUC10, and HUC12. Therefore, there are 6 shapefiles associated with this data set (range-wide [rw] across 3 scales, state-based [sb] across 3 scales). The .csv file titled "data_sources.csv" indicates where survey and museum records came from, including institutions and points of contact, as well as the date each database was queried.