Pierce, Adam L., authorStednick, John D., advisorBoone, Randall B., committee memberThornton, Christopher I., committee member2007-01-032007-01-032014http://hdl.handle.net/10217/88587The role and importance of selenium as an environmental contaminant has gained widespread attention among research scientists, natural resource managers, and federal and state regulatory agencies during the last two decades. Selenium has been listed on Colorado's Clean Water Act Section 303(d) List of Impaired Waters for Spring Creek and Fossil Creek in the city of Fort Collins. Selenium is one of the most hazardous of the trace metals, following mercury, with a narrow range between dietary deficiency and toxicity. Identifying selenium sources and understanding the environmental processes controlling how selenium is introduced to streams is critical to managing and mitigating the effects of elevated concentrations. A modeling approach was used to predict selenium concentrations with exploratory variables including 15 geospatial landscape parameters, precipitation, and streamflow for 5 sub-watersheds within Spring Creek and Fossil Creek watersheds. A correlation analysis was applied with surface water selenium concentrations and the better exploratory variables identified. Selected variables were used in a multiple linear regression model. Various combinations of different variables determined the best performing model, and included the area of shale, area of moderate to strongly alkaline soils, and the length of streams with an adjusted R2 of 0.99, [Se µg/L = 24.038 + 9.516(ALK) - 0.782(STR) -1.039(SHL)]; where ALK = area (km2) of moderate to strongly alkaline soils; STR = length (km) of streams; SHL = area (km2) of shale. Additional multiple linear regression models were developed in ArcGIS® using Ordinary Least Squares (OLS) Regression, and Geographically Weighted Regression (GWR) with area weighted geospatial variables. The best performing OLS model used only area (km2) of wetlands, with an adjusted R2 of 0.98, [Se µg/L = -6.584 + 170.509(wetlands)]. Similarly, the best performing GWR model included area of wetlands, with an adjusted R2 of 0.98. The second best performing GWR model included area of shale, with an adjusted R2 of 0.66. Limitations of this model include a very small sample size of water quality sampling stations, which limits the statistical power of multiple regression models used. Additional techniques applied in basin delineations with landscape element coupling for identification of hydrologic and/or chemical response units can further develop the platform for future modeling efforts targeting unmonitored watersheds.born digitalmasters thesesengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.water qualitygeospatialGISpredictionregressionseleniumText