Predictive Modeling of Bacteria Concentrations at Presque Isle State Park, Erie, Pennsylvania Active

Background

The Lake Erie beaches in Pennsylvania are a valuable recreational resource for Erie County. Approximately 7 miles of Lake Erie shoreline lies within Presque Isle State Park in Erie, Pa. Monitored public bathing beaches at Presque Isle make up less than 1 mile. Concentrations of Escherichia coli (E. coli) at monitored beaches in Presque Isle State Park occasionally exceed the single-sample bathing-water standard of 235 colonies per 100 milliliters resulting in potentially unsafe swimming conditions and prompting beach managers to post public advisories or to close beaches to recreation.

The use of E. coli as an indicator of recreational water quality has been largely effective in determining when fecal contamination is present; however, there are drawbacks with using it as the only indicator. Concentrations of E. coli may change significantly between the time of sample collection and the reporting of results (18-24 hours). A more rapid method that some managers of recreational waters have adopted is the use of water-quality and environmental variables as surrogates for fecal-indicator bacteria that include, for example, precipitation, wind speed and direction, streamflow, wave height and turbidity to predict when concentrations of fecal-indicator bacteria will exceed recreational standards. These emerging techniques may supplement the use of E. coli as an indicator of fecal contamination.

Predictive Bacteria Modeling, 2004-2006

To supplement the current method for assessing recreational water quality (E. coli concentrations from the previous day), the USGS Pennsylvania Water Science Center, in collaboration with the Erie County Department of Health, developed predictive regression models for E. coli at Presque Isle Beach 2 during the 2004-06 recreational seasons (May - September each year).

Results

• The strongest correlations were observed between E. coli and turbidity. Weaker correlations were observed between E. coli and streamflow, precipitation, and number of birds.
• Statistical relations also were observed between E. coli and wind direction, water-current direction, and wave height. Higher median concentrations of E. coli were observed when winds were blowing from the north/northeast; the recreational water-quality standard for E. coli was exceeded when water-current directions were northeast/east or southwest; and wave heights 3 feet and greater had higher E. coliconcentrations.
• All water-quality and environmental variables that were related to E. coli were considered in model development; models were developed at Presque Isle Beach 2 using combinations of data collected from 2004 to 2006.
• The 2005-06 dataset yielded the best predictive model for E. coli with the explanatory variables turbidity, wave height, and number of birds. It is not clear why excluding the 2004 data yielded the best model.
• The 2005-06 model was better at predicting when the recreational water-quality standard for E. coli would be exceeded than the current method of assessing recreational water quality (E. coli concentrations from the previous day).

Ongoing Studies, 2010-2015

The USGS Pennsylvania Water Science Center is working with Erie County Department of Health, Pennsylvania Department of Conservation and Natural Resources at Presque Isle State Park, and Regional Science Consortium at the Tom Ridge Environmental Center to expand the use of operational nowcast systems (based on predictive models) to Presque Isle State Park beaches. Funding from the Great Lakes Restoration Initiative is helping to make this work possible. During 2010, the USGS is collecting data for predictive bacteria model development for the rapid assessment of the recreational water-quality of Presque Isle Beaches using factors such as wave height, turbidity, number of birds on the beach, lake-current direction, rainfall, and wind direction.

The ultimate goal is the development of a web-based nowcast system that estimates current bacteria conditions much like a weather forecast estimates future weather conditions. Current bacteria conditions are estimated using a predictive regression model that utilizes weather and environmental data to predict whether or not E. coli levels will exceed an established threshold and a beach closing or water-quality advisory needs to be issued. Three or more years of data collection and model development typically are needed before a nowcast system can be established.