Characterizing brook trout (Salvelinus fontinalis) occurrence using landscape-level terrestrial characteristics of watersheds in the Arrowhead Region of Minnesota

Link To Share This Poster: https://cdmcd.co/LdDzG7
Live Discussion Link: https://cdmcd.co/3wqGX7

Lisa Elliott and Patrick Landisch, Department of Forest Resources, University of Minnesota, St. Paul, MN, William Severud, University of Minnesota, St. Paul, MN, Mark Nelson, Forest Inventory and Analysis, USDA Forest Service, St. Paul, MN, Jody Vogeler, Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, CO, Joseph Knight, Forest Resources, University of Minnesota, Saint Paul, MN

Background/Question/Methods
Land use, land cover, and disturbance within riparian areas and across entire watersheds may have both direct and indirect effects (e.g., through impacts on water quantity, quality, and temperature) on the ecology and distributions of aquatic species. Remote sensing technologies allows us to quantify such landsape-scale watershed components across large geographic extents. We used Boosted Regression Trees to model the influence of landscape characteristics on the distribution of brook trout (Salvelinus fontinalis) within part of its native range in Minnesota’s Lake Superior basin. We obtained watershed-level attributes from the 2016 National Land Cover Database, Landsat time series-based forest canopy disturbance data (1974-2018), the Forest Inventory and Analysis database, delineations of land ownership and protection status, and combined these with electrofishing data and historical stocking records. We created variable-width riparian buffers for watersheds within Minnesota’s Lake Superior basin using the Riparian Buffer Delineation Model, assessed landscape characteristics within and outside riparian areas at multiple hydrologic unit code (HUC) scales, and applied these data to our machine learning algorithm (Boosted Regression Tree) analysis of brook trout occurrence.
Results/Conclusions
Boosted regression tree analysis shows that forest land use in the watershed as a whole and tree canopy cover within riparian areas are particularly important for distinguishing occupied watersheds from unoccupied watersheds. Riparian area variables including developed land cover, multiple use lands, percentages of most recent fast disturbance, percentage of flooding, and percentage of unprotected use lands also contributed explanatory power to the model, as did watershed variables associated with the total watershed including percentage of most recent fast disturbance and impervious surface cover. We also identified interactions between tree canopy cover and developed land cover, tree canopy cover and percentage of most recent fast disturbance, and percentage of most recent fast disturbance and forest land use. These results emphasize the critical role of terrestrial management in aquatic systems. This application of landscape ecology offers a strategic approach for incorporating terrestrial characteristics into watershed management and conservation planning.