There are myriad barriers to aquatic connectivity beyond dams, with culverts at road crossings primary among them. UGA will lead the effort to develop a database of these non-dam blockages and model the likelihood that each is a barrier to fish movement, including mussel hosts. This process, described in more detail below, will result in a GIS point layer with numeric attributes that describe the likelihood that any given crossing is a blockage to fish passage. This data will be incorporated into the dam database to produce a database of all known and potential barriers in the region. This unified database will form the unit of analysis for the subsequent connectivity assessment in which each of the barriers will be assessed for its potential to benefit migratory & resident fish if removed or bypassed.
The UGA-USFWS team will use the National Inventory of Dams (NID) and other state data as available (e.g. DOT bridge and culvert data) to locate known structures. We will use Google Earth, NAIP photos, and field surveys to enhance those data to include high-resolution information on obstructions for a set of 12-digit HUCs, stratified across the SALCC region and capturing variation in land use, elevation, and position in the larger river basin. For road-stream crossings and water body outlets, classification models potentially including MaxEnt and Random Forests will be used to estimate likelihood of an obstruction and obstruction type based on covariates such as slope, watershed position/stream order, land use, the SARP riparian condition index, and other characteristics. Estimates of relative permeability for potential barriers will be generated from the literature combined with the barrier type and information from the field surveys (see below).
For barriers not associated with roads and water bodies, such as low-water crossings on private roads not captured in electronic databases, we will use multiple linear regression (model selection via an information-theoretic approach) to estimate the density of small obstructions at the 10 or 12-digit HUC scale. These data will be used in conjunction with land-use, topographic, and population density data to estimate the frequency of barriers which are not captured in the point layers developed above. While we acknowledge that these areal calculations are not directly useful in the connectivity assessment, they are necessary to estimate the accuracy of the other methods.
We will investigate the use of additional datasets, such as the Southeast Aquatic Resources Partnership (SARP) riparian condition index, land use derived from SALCC urban growth models, and stream slope (associated with NHD Plus datasets) as an indicator of naturally occurring dispersal barriers. This uniform region-wide barrier estimation project would aid efforts to address data gaps by identifying areas that are both barrier-dense and currently unsurveyed or where the confidence estimates around model predictions are wide.
We will perform field surveys of a subset of representative watersheds to augment the surveys already collected in the SSBGIS, targeting areas that are underrepresented in the dataset. Within the SALCC footprint, there are three datasets of stream barriers in the SSBGIS that will be most useful for developing our models. However, these cover US Forest Service lands (road crossings with 3-filter passability scores, gathered by the Center for Aquatic Technology Transfer in Blacksburg, Virginia), coastal plain areas (dams, culverts, water control structures in the North Carolina coastal plain, developed by the NC Department of Environment and Natural Resources), and an Aquatic Species Obstruction Inventory for North Carolina, collected by the North Carolina (Division of Land and Water Resources) and the US Army Corps of Engineers to identify small dams (less than 15 feet high), and dams that were previously undiscovered. We will collect field measurements of obstructions at approximately 450 points to verify the data within these datasets and sample areas where gaps exist in the current data coverage, such as private lands the Georgia and South Carolina piedmont. We plan to employ a smartphone-based rapid-survey tool developed by the USGS and USFWS Fish Passage Program for use in culvert assessments for a subset of these surveys to assess the utility of this inexpensive protocol relative to the more detailed and time-intensive traditional methodologies. This field data would be used both for model training and validation.
