Expanded Beach "Nowcast" Modeling across WI


The proposed project will significantly expand operational nowcasting of beach water quality in Wisconsin by building multivariate models for high priority and impaired beaches using EPA’s Virtual Beach 2.0. In addition, we will test a complimentary system developed by USGS and a hybrid qPCR/multivariate model. We will provide hands-on training and technical assistance to beach managers to build long-term, local capacity to operate and refine the models, and will provide user-feedback to EPA and USGS to further enhance the tools and lower barriers to their use throughout the Great Lakes.


To reduce the number of Type I and Type II monitoring errors – and the overall number of beach closures in Wisconsin – the proposed project will establish a three-year project position within the WDNR Bureau of Science Services to facilitate a significant expansion of operational nowcast modeling across the state. The goal is to establish 20 operational nowcast models by the summer of 2013, including 10 or more high priority beaches. Major project tasks will include building initial nowcast models for candidate beaches using modeling tools and database systems developed by EPA and USGS; developing, testing, and refining step-by-step nowcast training modules; conducting 5 hands-on training workshops; providing technical assistance to beach managers and monitoring personnel engaged in operating, evaluating, or refining nowcast models; compiling and providing user-feedback and practical suggestions to EPA and USGS; and helping to coordinate complimentary tool development and database integration efforts led by EPA, USGS, and others. Version 2.0 of EPA’s nowcasting software Virtual Beach (scheduled for release in spring 2010) will be used to build multivariate predictive models for 27 high priority beaches across Wisconsin, plus 25 additional beaches that are included or proposed for inclusion on the state’s 303(d) Impaired Waters list (see Table 1 and attached Project Map). Of the 52 candidate beaches, 41 have adequate data for building nowcast models; i.e. 120+ contemporaneous observations of the response variable, E. coli, and multiple explanatory variables (e.g., wave height, turbidity, antecedent rainfall) collected during routine sanitary surveys or via USGS an NOAA hydro-meteorological stations (Francy and Darner 2006; Mednick 2009). The remaining 11 beaches will have adequate data by 2011 under separate GLRI proposals submitted by project collaborators at the University of Wisconsin (UW)-Oshkosh and Racine Health Department (HD), titled Comprehensive Sanitary Survey Project for High Risk Wisconsin Beaches – Northern Wisconsin and Comprehensive Sanitary Survey Project for High Risk Wisconsin Beaches – Southern Wisconsin. During the summer of 2010 WDNR will complete work underway to assemble a master database for the 52 candidate beaches comprised of historic (ca. 2003-2010) E. coli monitoring results and contemporaneous data on “beach conditions” collected via routine sanitary surveys, as well hydro-meteorological data from NOAA and USGS automated observing stations and NWS Cooperative Observer Program weather stations; i.e., antecedent rainfall, air temperature, wind speed and direction, cloud cover, stream flow, lake level. This work will be conducted in collaboration with the USGS Wisconsin Water Science (WWSC), which is assembling similar databases for selected beaches under the federal GLRI project titled: Beach Health – Nearshore Water Quality and Beach Closures, and will be guided by a comprehensive inventory of publically-accessible data for nowcasting conducted by WDNR (Mednick 2010). Continuous and categorical time variables (e.g., Julian date and “season”) will be included in order to account for potential inter- and intra-seasonal variation in FIB concentrations not explained by other variables (Mednick et al. 2009).


The proposed project addresses Nearshore Health and NPS Pollution Goal #4 (“High quality bathing beach opportunities…”) in the draft Great Lakes Restoration Initiative Action Plan for FY2010-2014. Specifically, the project will help to achieve the objective of “rapid testing or predictive modeling… at 33% of high priority beaches” by 2014, by instituting nowcast models at 10 (37%) of the 27 high priority beaches in Wisconsin, and a combined 20 (37%) of the 54 high priority and/or 303(d) impaired beaches in the state by 2013. The project directly responds to the Great Lakes Regional Collaboration Strategy’s call for state, federal, local, and tribal partners to create and improve predictive models in conjunction with sanitary surveys. Disseminating information and training tools on predictive modeling is part of both the Lake Michigan and Lake Superior LaMPs. Enhancing predictive models, model scope, and application is a goal of the interagency Ocean Research Priorities Plan - Great Lakes. Field-testing and the development of user guidelines for Virtual Beach is part of EPA’s Critical Path Science Plan. Lastly, the project will help to meet the goal of the Wisconsin Great Lakes Strategy goal of reducing closure dates at high priority beaches by 10% from base year 2006. The baseline accuracy assessment described under Project Background (Section 8) and numerous beach-specific studies (previously cited) strongly suggest that nowcast modeling will result in an overall reduction in beach closures.

Related Reports

Run Project Summary Report
View Umbrella-Projects
View Related-Projects

Great Lakes Restoration Initiative
Nearshore Health and NPS
Reports and Documents
Summary report of Beach Modeling Workshop in Columbus, OH 12/5-7/11
Grant report to EPA
GLRI grant progress
Appendix to final QAPP - Beach Monitoring QAPP
Signature page - WDNR PROJECT NAME: Expanded Beach "Nowcast" Modeling across WI
Final Approved QAPP
Appendix 2 to approved QAPP - Sanitary Survey
Activities & Recommendations
Information and Education
Expand operational nowcasting of beach water quality Lakes Michigan and Superior.