The Ohio Nowcast has been providing real-time beach advisories to the public on the basis of predictive models since 2006. In support of the nowcast, data were collected during the recreational season of 2008 to validate and refine predictive models at two Lake Erie beaches. Predictive models yield data on the probability that the single-sample bathing-water standard for E. coli will be exceeded. Field personnel collected or compiled data on Escherichia coli (E. coli) concentrations as well as variables expected to affect these concentrations, including manual and automated measurements of turbidity, wave height, and water temperature; lake level; and radar and airport rainfall amounts. Two new variables were measured during 2008 - photosynthetically-active radiation at Huntington (Bay Village) and foreshore head at Edgewater (Cleveland). (The foreshore is a strip of land along a body of water between low and high water marks.) The performance of the nowcast was monitored during 2008. The Huntington nowcast yielded a greater percentage of correct responses (84.9 percent) than did the previous day's E. coli concentration (75.2 percent). In contrast, at Edgewater, the nowcast yielded a slightly higher percentage of correct responses (61.0 percent) as compared to the previous day's E. coli concentration (56.5 percent), but both percentages were relatively low. Lake levels in 2008 were significantly higher than levels in the data used to develop the Edgewater models (2004-7), confounding their abilities to provide correct responses. At Edgewater during 2008, the strongest relation (as measured by Pearson's correlation) was between E. coli concentrations and the difference in foreshore head over the past 24 hours (r=0.48), a variable not included in the models. At Huntington, photosynthetically-active radiation on the previous day showed a significant negative relation to E. coli concentrations (r=-0.33) during 2008. Refined models were developed for Huntington and Edgewater using data collected from 2005-8. The refined models included the variables wave height, log turbidity, radar or airport rainfall, and day of the year in various combinations for different dated segments of the recreational season. Water-resource managers will determine which models to apply to the Ohio Nowcast for issuing water-quality advisories in 2009.