Swim advisories are issued by beach managers on the basis of standards for concentrations of bacterial indicators—Escherichia coli
) or enterococci for freshwaters and enterococci for marine waters. The analytical methods for these organisms, however, take at least 18–24 hours to complete. Recreational water-quality conditions may change during this time, leading to erroneous assessments of public-health risk. As a result, some agencies have turned to modeling to obtain near-real-time estimates of recreational water quality.
Techniques such as multiple linear regression (MLR) are used to develop
multivariable statistical models on the basis of relations between
fecal-indicator bacteria concentrations and variables known or suspected to
affect their concentrations in a particular water body. The sources of fecal
contamination do not need to be identified in order to develop and use
statistical models. Multivariable statistical models (hereinafter
“predictive models”) are being developed and tested in many areas of the
USA. They are used for beach closure or advisory decisions at a few areas in
the Great Lakes including the Ohio Nowcast.
The Ohio Nowcast is the result of multi-year partnerships on several projects between the U.S. Geological Survey (USGS) Ohio Water Science Center (OWSC), and other federal, state, and local agencies
and universities. Current
and past partners include Cleveland Metroparks, Cuyahoga County Board of Health (CCBH),
Erie County Health Department, Lake County General Health District,
Muskingum Watershed Conservancy District, Northeast Ohio Regional Sewer District (NEORSD), Ohio Department of Health (ODH), Ohio Department of Natural Resources (ODNR), Ohio Lake Erie Office, Ohio Water Development Authority,
University of Toledo, U.S. Environmental Protection Agency, and the U.S. National Park Service.
History of statistical modeling in Ohio
In Ohio, the use of predictive models was first explored with one year data at three Lake Erie beaches (Francy
and Darner, 1998
). This effort was expanded to include three additional Lake Erie beaches, one inland lake, and two seasons of data collection (Francy and others, 2002
). Predictive models were subsequently developed for five Ohio Lake Erie beaches with 2-4 years of data, depending on the beach (Francy and others, 2006
). The best model for each beach was based on a unique combination of variables that explained changes in
concentrations. The “variables” included turbidity (water clarity), rainfall, wave height, water temperature, day of the year, and lake level. The model from Huntington (Bay Village) was validated using data collected during a independent year, leading to implementation of the Ohio Nowcast for Huntington in 2006.
At the same time, a rapid analytical method was tested at river sites within
the Cuyahoga Valley National Park (Brecksville, Oh)
models with turbidity as an explanatory variable were developed
(Brady and others
. Continuing the work in Lake Erie, data were collected during the recreational season of 2007 to test and refine predictive models at Huntington, Edgewater (Cleveland), and Villa Angela (Cleveland) (Francy and Darner, 2007
). The Huntington and Edgewater models performed well, and Edgewater was added to the Ohio Nowcast
in 2008. Data were collected during 2008 to operate the nowcast at Huntington and
Edgewater and monitor the performance of the models
(Francy and others, 2009)
In 2011, Maumee Bay State Park was added and in 2012, Villa Angela was added to the Ohio Nowcast.
The USGS has been working with local agencies to test and develop predictive models at other Lake Erie beaches and at some Ohio inland lake beaches. During 2014, Nickel Plate (Huron, Oh), Vermillion
Main Street Beach (Vermillion, Oh), Mentor Headlands State Park (Mentor, Oh), Fairport Harbor Lakefront Park Beach (Fairport Harbor, Oh) will be added to the Ohio Nowcast. We will continue to test models at Lakeview (Lorain, Oh), Lakefront Park (Huron, Oh), Tappan Lake (Deersville), and Seneca Lake (Senecaville).
The steps to develop statistical models are data collection; exploratory data analysis; model development, selection, and diagnosis; determination of model output values; and model validation and refinement. These steps are described in detail with examples in
Francy and Darner (2006)
Predictive modeling is a dynamic process; that is, models should be
continuously validated and refined to improve predictions and better protect
Tools to compile data and develop predictive models are available on-line through a
USGS report. These tools include software for creating predictive models (Virtual Beach) developed by U.S. Environmental Protection Agency; software to process weather data from the nearest National Weather Service airport site (PROCESSNOAA); and a spreadsheet to process lake-level data from the National Oceanic and Atmospheric Administration (NOAA). Also in the report are results from a study where the USGS worked with 23 local and state agencies to improve existing operational nowcast systems at 4 beaches and expand the use of predictive models in nowcasts at an additional 45 beaches throughout the Great Lakes. The predictive models were specific to each beach, and the best model for each beach was based on a unique combination of environmental and water-quality variables. During validation of 42 beach models in 2012, the models overall performed better than the current method to assess recreational water quality (previous day’s
E. coli concentration).
and Plona, M.G., 2009, Relations Between Environmental and Water-Quality
Variables and Escherichia coli in the Cuyahoga River With Emphasis on
Turbidity as a Predictor of Recreational Water Quality, Cuyahoga Valley
National Park, Ohio, 2008: U.S. Geological Survey Open-File Report 2009–1192,
Brady, A.M.G., 2007, Rapid method for
Escherichia coli in the Cuyahoga River: U.S. Geological Survey Open-File Report 2007–1210, 5 p.
Bushon, R.N., and Plona, M.G., 2009, Predicting recreational water quality
using turbidity in the Cuyahoga River, Cuyahoga Valley National Park, Ohio,
2004—7: U.S. Geological Survey Scientific Investigations Report 2009–4192,
D.S., Darner, R.A., 1998, Factors affecting Escherichia coli concentrations at
Lake Erie public bathing beaches: U.S. Geological Survey Water-Resources
Investigations Report 98-4241, 41 p.
D.S., Gifford, A.M., Darner, R.A., 2002, Escherichia coli at Ohio bathing
beaches— distribution, sources, wastewater indicators, and predictive modeling:
U.S. Geological Survey Water-Resources Investigations Report 02-4285. 120 p.
Francy, D.S., Brady, A.M.G.,
Carvin, R.B., Corsi, S.R., Fuller, L.M., Harrison, J.H., Hayhurst, B.A., Lant,
J., Nevers, M.B., Terrio, P.J., and Zimmerman, T.M., 2013, Developing and
implementing predictive models for estimating recreational water quality at
Great Lakes beaches: U.S. Geological Survey Scientific Investigations Report
2013–5166, 68 p.
Francy, D.S. and Darner, R.A., 2006, Procedures for developing models to predict exceedance
of recreational water-quality standards at coastal beaches: U.S. Geological
Survey Techniques and Methods 6-B5, 34 p.
Francy, D.S., Darner, R.A., and Bertke, E.E.,
2006, Models for predicting recreational water quality at Lake Erie beaches:
U.S. Geological Survey Scientific Investigations Report 2006-5192, 13 p.
Francy, D.S., and Darner, R.A., 2007, Nowcasting
beach advisories at Ohio Lake Erie beaches: U.S. Geological Survey Open-File
Report 2007–1427, 13 p.
Francy, D.S., and Darner, R.A.,
2009, Testing and refining the Ohio Nowcast at two Lake Erie beaches in 2008:
U.S. Geological Survey Open-File Report 2009–1066, 20 p.