Graduation Year


Document Type




Degree Name

Master of Science (M.S.)

Degree Granting Department

Marine Science

Major Professor

Joan B. Rose, Ph.D.

Committee Member

Valerie J. Harwood, Ph.D.

Committee Member

John T. Lisle, Ph.D


An Escherichia coli ribotype profile library was created for the Tampa Bay watershed for application to bacterial source tracking of human or non-human fecal contamination to the bay and surrounding rivers. Bacterial source tracking uses a variety of methods to identify the source of fecal contamination, not just the quantity of the indicator bacteria present, allowing implementation of effective management practices. The 700 confirmed E. coli isolate library was equally distributed among four sources: human, bird, dog and cow. The library was tested for applicability for source tracking within the Tampa Bay watershed. The DuPont Qualicon RiboPrinter® Microbial Characterization System, a fully automated ribotyping instrument, was used to generate this library. The ribotype patterns were analyzed by software packages from the RiboPrinter® and BioNumerics®. Proficiency isolates were used to test the accuracy of the library. The RiboPrinter® defined 166 ribogroups, whereas, BioNumerics® defined 234 ribogroups for the same isolates. While the numbers of ribogroups differed, percent classifications in the four source categories were similar when normalized data were compared between software packages. The RiboPrinter® found dog, bird, and cow isolates represented in 34, 38, 29 and 31% of the ribogroups, respectively. BioNumerics® found dog isolates, bird, human and cow isolates in 32, 35, 31 and 30% of the ribogroups, respectively. The RiboPrinter® and BioNumerics® software respectively assigned 58% and 32%, of the proficiency isolates to their correct source categories. The average rate of correct classification, as determined using BioNumerics® and Jackknife analysis, ranged from 72.57% for human isolates to 77.14% for cow isolates. Conversely, the individual holdout analyses showed low rates of correct classification, with most individuals less than 50% correctly classified, of individuals (a single sample containing multiple isolates). The individual holdout analyses and the proficiency isolate data indicate low correct classification rates(less than 60%), suggesting, the results of the Jackknife analysis over-estimates the rates of correct classification. Currently, this library does not offer the discrimination needed for most bacterial source tracking applications and a large library containing the diversity of isolates will be needed. The 700 isolates and the ribogroups that they form provide foundation for future work.