Predicting freshwater fish distributions using landscape-level variables

摘要

Management and conservation of aquatic systems requires the ability to identify species' historical, current, and potential distributions. We explored how a geographic information system can be used in conjunction with a few broad landscape variables to provide watershed-scale information useful for identifying diverse aquatic areas and predicting potential fish habitat. We developed species habitat profiles for all fish species that are known to occur in Pennsylvania. Five landscape variables were used to characterize a species' habitat profile to predict its statewide distribution: presence in a major drainage basin, presence in a physiographic region, median watershed slope, level of watershed disturbance, and watershed-stream size. Each of these variables was referenced to a small watershed boundary. Using these variables, we predicted a species potential habitat range. Distribution maps that we generated were then compared to known distributions with an average accuracy of 73%. While many collections have been made in Pennsylvania over the last 50 years, we determined that many areas still remain unexplored as potential sampling locations. Among those fishes whose predicted distribution was less than the actual sampled distribution, four receive special protection in Pennsylvania and one is federally endangered. Moreover, we determined that small watersheds (1:24,000 scale) in the Allegheny River drainage, in the Pittsburgh Low Plateau Section, of small size (3-4 order), with moderate slope (2-4%), and moderate watershed disturbance (25-75%) have the highest fish species richness. Our results should facilitate the conservation of fish species and our technique should be easily repeatable in other geographic areas.