Modeling Nitrate Loading Rate in Delaware Lakes Using Regression and Neural Networks

Abstract

The objective of this research was to predict the nitrogen-loading rate to Delaware lakes and streams using regression analysis and neural networks. Both models relate nitrogen-loading rate to cropland, soil type and presence of broiler production. Dummy variables were used to represent soil type and the presence of broiler production at a watershed. Data collected by Ritter & Harris (1984) was used in this research. To build the regression model Statistical Analysis System (SAS) was used. NeuroShell Easy Predictor, neural network software was used to develop the neural network model. Model adequacy was established by statistical techniques. A comparison of the regression and neural network models showed that both perform equally well. Cropland was the only significant variable that had any influence on the nitrogen-loading rate according to both the models.