Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.
Predicting Waterborne Escherichia coli Particle Attachment Using Regression Tree Analysis
Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.orgCitation: 21st Century Watershed Technology: Improving Water Quality and the Environment Conference Proceedings, May 27-June 1, 2012, Bari, Italy 12-13704.(doi:10.13031/2013.41442)
Authors: Gregory S Piorkowski, Rob C Jamieson
Keywords: E coli, particle attachment, regression tree
Deterministic watershed models can adequately predict surface water E. coli concentration by incorporating settling and resuspension routines. The partitioning of E. coli between planktonic and particle-associated phases is a sensitive parameter in these models, but is not well understood. This study evaluates land use, hydrological, water quality and particle properties on percent E. coli particle attachment. Sixty water samples were retrieved from four monitoring stations between May and November 2011, capturing four major storm events, and analyzed for bacterial and particulate parameters. Multiple linear regression and regression tree models were built and evaluated for predicting E. coli particle attachment in surface water. Percent of E. coli particle attachment ranged between 48.2% and 94.4%, with an average of 67.3%. The regression models revealed that a combination of land use (%forested, %residential), water quality (EC, pH, DO, temperature), particulate (TSS, VSS, %organic fraction, turbidity) and particle size distribution (%sand, %silt, geometric mean diameter, ratio of interquartile particle diameters) properties were significant in predicting E. coli particle attachment. A parsimonious regression tree model that did not contain particle size distribution data was the strongest predictive model, exhibiting a low root mean squared error (13.2) and a high index of agreement (0.92). This study demonstrates the utility of regression tree models for estimating sensitive watershed model parameters.(Download PDF) (Export to EndNotes)