Predictive Models for Determination of E. coli Concentrations at Inland Recreational Beaches Ayokunle Christopher Dada & David P. Hamilton Received: 1 March 2016 /Accepted: 3 August 2016 /Published online: 29 August 2016 # Springer International Publishing Switzerland 2016 Abstract Given the 24-h turn-around time before swimming advisories are released, advisories issued to protect public health really only indicates ‘it may be unsafe to swim yesterday’. Predictive modelling for Escherichia coli concentrations at inflow-impacted beaches may be a favourable alternative to the current, routinely criticised monitoring approach. Using a total of 482 sets of meteorological and bacteriological data covering 14 swimming seasons, as well as environmen- tal data of 10 inflow streams, this study developed models that could be used for predicting E. coli concen- trations at five Lake Rotorua beaches. The models in- clude predictor variables such as wind speed, antecedent rainfall, suspended solids at Puarenga, Utuhina and Ngongotaha stream inflows and particulate inorganic phosphorus concentration at Puarenga stream inflow. The combined 2011–2012 models had an average- adjusted R 2 of 0.73, root mean square error (RMSE) of 0.33 logCFU/100 mL and captured 38 % of the variance in the validation data when used to predict E. coli con- centrations for an additional 2 years (2013–2014). Among the individual beach models, predictive accura- cy ranged from 88.89 to 92.31 % for the three beaches considered in the study. The developed models can provide a faster estimation of E. coli condition, poten- tially assisting local beach managers in the decision process related to swimming advisories issuance. Keywords Beach advisory . Water quality . Indicator bacteria . Water quality model . Inland beaches . Lake inflows 1 Introduction For many decades, there has been a lot of published information on studies conducted in different settings of the world, all with a view to assessing and minimising risks associated with recreational water quality (Zmirou et al. 2003; Soller et al. 2010, 2014; Dada et al. 2012; Ahmad et al. 2013; Devane et al. 2015). Yet, today’ s water quality monitoring is faced with a number of challenges. First and perhaps most critical is that the conventional tests used as a protective measure to safe- guard swimmers from exposure to pathogens are usually associated with a delivery time of at least 24 h before the results can inform swimming advisory decisions (Wade et al. 2003; Gonzalez et al. 2012). This is added to the infrequent nature of the tests, testing frequencies often restricted to three times a month and isolated to summer seasons due to monitoring costs. The absence of a real- time water quality prediction system could be hamper- ing efforts aimed at reduced pathogen exposure and public health protection generally. Additionally, since bacterial concentrations often change overnight and Water Air Soil Pollut (2016) 227: 347 DOI 10.1007/s11270-016-3033-6 Electronic supplementary material The online version of this article (doi:10.1007/s11270-016-3033-6) contains supplementary material, which is available to authorized users. A. C. Dada (*) : D. P. Hamilton Environmental Research Institute, Faculty of Science and Engineering, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand e-mail: cdada@waikato.ac.nz