The mathematical models used to make predictions on the nowcast are based on measurements of environmental and water-quality variables known to influence E. coli levels. These are called "explanatory variables" and include measurements of rainfall, wave height, and water clarity.
During 2010, two models will be used at Huntington and three models will be used at Edgewater. Which model is used is based on the date; that is, a different model is used at the beginning of the season than at the end of the season. Using different mathematical formulas and different variables in the models for different times of the year will hopefully minimize the percentage of incorrect predictions provided by the models in 2010. All of the models are developed with radar rainfall, obtained from the National Weather Service through a series of data transfers. In the event the radar data are not transmitted successfully, a model for each beach was developed as a backup using rainfall data obtained from the National Weather Service site at Hopkins International Airport.
At Huntington, the variables used in the models for the 2010 nowcast are as follows:
Subseason 1 (May 18−July 23): (1) log turbidity (the logarithm of the measured water clarity), plus the wave height, (2) radar rainfall from six 4-km grids surrounding the beach during the previous 24 hours, and (3) day of the year. The threshold probability for issuing an advisory using this model is 23 percent.
Subseason 2 (July 24−Sept 7): (1) log turbidity plus the wave height and (2) radar rainfall from six 4-km grids surrounding the beach during the previous 48 hours with the most recent rainfall receiving the most weight. The threshold probability for issuing an advisory using this model is 32 percent.
Backup model with airport rainfall (entire season): (1) log turbidity plus the wave height, (2) airport rainfall during the previous 48 hours with the most recent rainfall receiving the most weight, and (3) day of the year. The threshold probability for issuing an advisory using the model is 34 percent.
Subseason 1 (May 1−June 13): (1) log turbidity (the logarithm of the measured water clarity) plus the wave height and (2) radar rainfall from two 4-km grids surrounding the beach during the previous 48 hours with the most recent rainfall receiving the most weight. The threshold probability for issuing an advisory using this model is 27 percent.
Subseason 2 (June 16−Aug 10): (1) log turbidity plus the wave height, (2) radar rainfall from two 4-km grids surrounding the beach during the previous 48 hours with the most recent rainfall receiving the most weight, and (3) day of the year. The threshold probability for issuing an advisory using this model is 28 percent.
Subseason 3 (Aug 11−Sept 15): (1) log turbidity plus the wave height, (2) radar rainfall from two 4-km grids surrounding the beach during the previous 48 hours with the most recent rainfall receiving the most weight, and (3) day of the year. The threshold probability for issuing an advisory using this model is 31 percent.
Backup model with airport rainfall (entire season): (1) log turbidity plus the wave height and (2) airport rainfall during the previous 48 hours with the most recent rainfall receiving the most weight. The threshold probability for issuing an advisory using this model is 30 percent.
