A Regression Model for the Prediction of Chlorophyllain Lake Okeechobee, Florida

Lake Okeechobee water quality data from 1980 to 1993 were used to develop a model for predicting chlorophyll a concentrations. Data collected by the South Florida Water Management District (SFWMD) and the University of Florida included stage, temperature, total nitrogen, total phosphorous, wind velocity and chlorophyll a. The lake was divided into five zones based on previous studies. Exploratory data analyses were performed to determine spatial and temporal scales of variation in chlorophyll a levels. The result of a backward testing strategy based on Type III sum of squares was a linear model in which the principle predictor of chlorophyll a concentration was log10 total phosphorus, followed by log10 total nitrogen and then by log10 temperature, die latter of which can be considered a seasonality term. The model suggests that location in the lake is important in predicting chlorophyll a since statistically different slopes describe the relationship between the predictors (temperature, total phosphorus and total nitrogen) and the response (chlorophyll a) for each zone. Stage is a highly significant term (p < 0.0001), and wind is absent as a predictor in the final model. The high importance given to the interaction terms involving zone suggests that location is a prime determinant of chlorophyll a concentration, perhaps as a surrogate of (unmeasured) individual station (or ZONE) wind effects or fetch length.