The field of bacterial source tracking (BST) has been rapidly evolving to meet the demands of water pollution analysis, specifically the contamination of waterways and drinking water reservoirs by point source and nonpoint source pollution. The goal of the current study was to create a BST library based on carbon-utilization patterns (CUP) for predicting sources of E. coli in a watershed, to compare this library to an antibiotic-resistance analysis (ARA) library previously published for the same isolates, and to determine the efficacy of using a composite dataset which combines data from both datasets into a single library for predicting the source of unknown isolates. This was accomplished by generating a CUP dataset and a composite ARA-CUP dataset for the E. coli isolates from known fecal sources within a watershed. These libraries were then used to predict the sources of E. coli isolates collected from 13 water sites in the same watershed and compared in regard to predictive accuracy. The dominant sources of E. coli in the South Bosque watershed were cattle as identified by all three methods. The 6-source composite library had higher average rates of correct classification (96.7%), specificity (99.2%), positive-predictive value (99.1%), and negative-predictive value (96.8%) than either the ARA or CUP 6 source libraries (ARCC 80.1% and 86.7% respectively). The current study is the first field study to compare two phenotypic methods, Antibiotic Resistance Analysis (ARA) and Carbon Utilization Profiling (CUP). This study is also the first to combine both of these methods to create a composite “toolbox” type approach.