Nowcast Modeling of Escherichia coli Concentrations at Multiple Urban Beaches of Southern Lake Michigan

US Geological Survey, Great Lakes Science Center, 1100 N. Mineral Springs Road, Porter, IN 46304, USA.
Water Research (Impact Factor: 5.53). 01/2006; 39(20):5250-60. DOI: 10.1016/j.watres.2005.10.012
Source: PubMed


Predictive modeling for Escherichia coli concentrations at effluent-dominated beaches may be a favorable alternative to current, routinely criticized monitoring standards. The ability to model numerous beaches simultaneously and provide real-time data decreases cost and effort associated with beach monitoring. In 2004, five Lake Michigan beaches and the nearby Little Calumet River outfall were monitored for E. coli 7 days a week; on nine occasions, samples were analyzed for coliphage to indicate a sewage source. Ambient lake, river, and weather conditions were measured or obtained from independent monitoring sources. Positive tests for coliphage analysis indicated sewage was present in the river and on bathing beaches following heavy rainfall. Models were developed separately for days with prevailing onshore and offshore winds due to the strong influence of wind direction in determining the river's impact on the beaches. Using regression modeling, it was determined that during onshore winds, E. coli could be adequately predicted using wave height, lake chlorophyll and turbidity, and river turbidity (R2 = 0.635, N = 94); model performance decreased for offshore winds using wave height, wave period, and precipitation (R2 = 0.320, N = 124). Variation was better explained at individual beaches. Overall, the models only failed to predict E. coli levels above the EPA closure limit (235 CFU/100 ml) on five of eleven occasions, indicating that the model is a more reliable alternative to the monitoring approach employed at most recreational beaches.

Download full-text


Available from: Meredith Becker Nevers
  • Source
    • "turbidity or 24 h rainfall total (Brandt et al., 2006;Olyphant and Whitman, 2004). Multiple linear regression (MLR), with the model estimated by ordinary least squares (OLS) is the most popular regression method for FIB nowcast models (Nevers and Whitman, 2005;Francy and Darner, 2007;de Brauwere et al., 2014). However, OLS is well-known for drawbacks like overfitting, difficulty of variable selection, and the inflexibility of its linear modeling structure (Ge and Frick, 2007). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Epidemiological studies indicate that fecal indicator bacteria (FIB) in beach water are associated with illnesses among people having contact with the water. In order to mitigate public health impacts, many beaches are posted with an advisory when the concentration of FIB exceeds a beach action value. The most commonly used method of measuring FIB concentration takes 18-24 h before returning a result. In order to avoid the 24h lag, it has become common to "nowcast" the FIB concentration using statistical regressions on environmental surrogate variables. Most commonly, nowcast models are estimated using ordinary least squares regression, but other regression methods from the statistical and machine learning literature are sometimes used. This study compares 14 regression methods across 7 Wisconsin beaches to identify which consistently produces the most accurate predictions. A random forest model is identified as the most accurate, followed by multiple regression fit using the adaptive LASSO.
    Full-text · Article · Nov 2015 · Environmental Modelling and Software
  • Source
    • "Lake Michigan, U.S.A. (Nevers and Whitman, 2005) and oceans, e.g. in England (Kashefipour et al., 2002). Nevers and Whitman (2005) applied a statistical model based on correlations between bacteria and other water quality parameters to predict bathing water quality. A similar approach was applied by Frick et al. (2005) for Lake Erie, U.S.A.. Viegas et al. (2012) developed a hygienie early warning system for a beach at the Portuguese coast based on a deterministic modeling approach, taking into account tides, currents, and wind. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The lower Ruhr River is located in a densely populated and industrialized area in Northrhine-Westphalia (NRW) in western Germany. Due to upgrades of sanitary infrastructure, such as wastewater treatment plants (WWTPs) and combined sewer overflows (CSOs), and a decline of industrial production, water quality of Ruhr River has been constantly increasing over the past decades. One effect is a growing attractiveness of the Ruhr for bathing and water sports. In order to enable future bathing in the lower Ruhr, this study investigates methods for predicting the permissibility of bathing, according to the microbial water quality regulations of the Bathing Water Ordinance of Northrhine-Westphalia (NRW-BWO). On basis of the European Commission Bathing Water Directive, the NRW-BWO defines methods for the assessment of bathing water quality on basis of bacterial threshold concentrations of Escherichia coli (E. coli) and intestinal enterococci (Int. Ent.). Furthermore, if the bathing water is subject to short-term pollution, the NRW-BWO requires the installation of an early warning system to prevent bathers' exposure. Laboratory detections of both bacteria species from water samples are not suitable to be used in an early warning system. Online measurement devices for bacteria showed to be not sensitive and accurate enough to reliably indicate an exceedance of the threshold values. Thus, the application of a prediction model is appropriate. In total, four different modeling approaches were developed and compared to provide short-term predictions of bacterial concentrations: (i) statistical modeling based on linear correlations between hydro-chemical parameters, such as ammonia and turbidity, and bacteria, (ii) modeling based on artificial neural networks (ANNs), which consider non-linear correlations between hydro-chemical and climate parameters and bacteria concentrations, (iii) a balance model, which considers all in- and outflows, both in terms of water quality and quantity, along a stretch of the lower Ruhr River, and (iv) binary modeling based on precipitation rates, as rainfall is assumed to trigger high bacteria loads in the river. It could be shown that ANNs allow the most accurate prediction of bacterial concentrations in the lower Ruhr River. However, the model performance varies among different stretches along the Ruhr River. This indicates that local conditions, e.g. distance to next upstream WWTP or CSO, are essential and need to be further investigated. The binary model which considered rainfall effects also provided acceptable short-term predictions. For example, at all potential bathing spots, after two days following substantial precipitation amounts, bathing would have been allowed. The balance model showed the weakest results, which is mainly due to data gaps, as time series of bacterial loads from tributaries, WWTPs and CSOs had to be estimated. As a next step, high resolution bacterial measurements following CSO discharge events are planned in order to develop a concise picture of processes determining bacterial concentrations at the Ruhr River. Copyright © 2015 Elsevier GmbH. All rights reserved.
    Full-text · Article · Jun 2015 · International journal of hygiene and environmental health
  • Source
    • "Statistical FIB models rely heavily on the selection of a best subset of predictor variables. This selection is often completed in a stepwise fashion (Nevers and Whitman, 2005; Gonzalez et al., 2012; Gonzalez and Noble, 2013), where predictors are added or dropped according on their coefficient significance. Stepwise subset selection methods can occasionally fail to find the true best subset of predictors (Berk, 1978) especially when selection is influenced by multicollinearities (which can alter the calculated statistical significance of variables), or the curse of dimensionality (i.e. a large number of candidate models) makes a robust model selection difficult. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Although the relationships between meteorological conditions and waterway bacterial contamination are being better understood, statistical models capable of fully leveraging these links have not been developed for highly urbanized settings. We present a hierarchical Bayesian regression model for predicting transient fecal indicator bacteria contamination episodes in urban waterways. Canals, creeks, and rivers of the New York City harbor system are used to examine the model. The model configuration facilitates the hierarchical structure of the underlying system with weekly observations nested within sampling sites, which in turn were nested inside of the harbor network. Models are compared using cross-validation and a variety of Bayesian and classical model fit statistics. The uncertainty of predicted enterococci concentration values is reflected by sampling from the posterior predictive distribution. Issuing predictions with the uncertainty reasonably reflected allows a water manager or a monitoring agency to issue warnings that better reflect the underlying risk of exposure. A model using only antecedent meteorological conditions is shown to correctly classify safe and unsafe levels of enterococci with good accuracy. The hierarchical Bayesian regression approach is most valuable where transient fecal indicator bacteria contamination is problematic and drainage network data are scarce. Copyright © 2015 Elsevier Ltd. All rights reserved.
    Full-text · Article · Jun 2015 · Water Research
Show more