May, Liliane, Clean Water Institute, Lycoming College, One College Pl, Williamsport, PA 17701, maylili@lycoming.edu; Rieck, Leslie O., Department of Biology, Lycoming College, 1 College Place, Williamsport, PA 17701, rieck@lycoming.edu; Kaunert, Matt, Clean Water Institute Lycoming College 1 College Place, Williamsport PA 17701, kaunert@lycoming.edu; Bohlin, Emily, Department of Biology, Lycoming College, 1 College Place, Williamsport, PA 17701, bohlin@lycoming.edu.
Pennsylvania has approximately 80,000 miles of streams, many of which are coldwater streams with native brook trout (Salvelinus fontinalis) populations that are important to recreation and serve as an excellent bioindicator. Unfortunately, many of these streams are threatened by resource extraction, declining water quality, and changes in land use. The Pennsylvania Fish and Boat Commission is tasked with surveying state streams and protecting streams containing naturally reproducing brook trout populations. The Unassessed Waters Initiative (UWI) began in 2010 as a joint effort between PFBC and partners (e.g., institutions of higher education, non-profit organizations) to identify high priority watersheds through standardized electroshocking and water quality surveys. Lycoming College’s Clean Water Institute (CWI) joined the program in 2010 and has participated every year since. Between 2012 and 2024, CWI surveyed 638 streams, documenting wild brook trout populations in Lycoming, Tioga, Clinton, Clearfield, and Centre counties. Here, we centralized and organized CWI’s UWI dataset and used it to investigate physical and geological indicators (predictors) of wild trout presence. All data were entered into Microsoft Excel then edited to remove errors and standardize the units and formatting of all variables. We then used the data on the 365 streams from that dataset for which all data was complete. Of those 365 streams, 139 (40.8%) contained no brook trout. Using the predictors of stream wetted width (m), dissolved oxygen (mg L-1), temperature (°C), pH, and conductivity (μS cm-1), we completed a principal components analysis in JMP 18 on the log-transformed predictors, retaining principal components (PCs) with eigenvalues > 1. Using these two retained PCs as predictors, we used AICc-based model selection in R 4.4.1 to investigate 12 possible models (all possible combinations of PCs 1 and 2), all using a zero-inflated negative binomial general linear model. We considered models well supported when ΔAICc < 2. PC1 explained 37.4% of the variation in predictor variables and was negatively associated with dissolved oxygen and positively associated with temperature, alkalinity, and conductivity. PC2 explained 18% of the variation in predictor variables and was positively associated with stream wetted width and negatively associated with pH. Four models had ΔAICc < 2, none including a significant predictor PC for the count portion (i.e., explaining trends in streams in which brook trout were present) of the zero-inflated negative binomial model. PC1 (related negatively to DO and positively to temperature, alkalinity, and conductivity) was significant, with a positive coefficient, in all four best-supported models, while PC2 was not significant in any best-supported model. These results suggest that dissolved oxygen and temperature may be the primary determinants of trout abundance when trout are already present, but that other factors may influence their absence from an area (e.g., barriers, low flow). Future work should expand the dataset to include more of the statewide UWI dataset and add a wider array of variables, including landscape factors (e.g., barriers, hydrology).
Brook Trout, Unassessed Waters, Modeling