Predicting Dangerous E. Coli Levels at Erie, Pennsylvania, Beaches with Random Forests in R

• Activity Number: 619
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Michael Rutter*+
• Companies: Penn State
• Keywords: Water Quality ; Random Forests ; Statistical Learning ; R ; E. Coli ; Real time

Abstract

Presque Isle State Park in Erie, Pennsylvania is home to 11 public beaches on Lake Erie that attract more than 4 million visits annually. During the summer swim season, water quality is always a public health concern, specifically the presence of Escherichia coli (E. coli). While E. coli itself can be dangerous to humans, it is also an indicator of other, harmful bacteria that are much harder to detect. Given the risk to human heal, there is a need for a statistical model that attempts to predict E. coli levels before swimmers arrive at the park. A prediction model will be presented based on weather conditions and other data sources available on the internet using a random forest decision tree approach. The entire process of collecting data from the internet, generating predictions from the model, and presenting the results in a web page are done completely within R. The model is currently being used as part of the decision process for issuing beach warnings at Presque Isle State Park.