Dynamics of surface runoff and its influence on the water quality using competitive algorithms in artificial neural networks

Abstract

This study was aimed at developing a modeling technique to accurately describe the hydrological interaction with non-point pollutants using Artificial Neural Networks (ANNs). Rainfall, surface discharge water, and nutrient concentrations (total nitrogen and total phosphorus) were monitored and used for ANN computation. A comparison study was conducted for two well-known algorithms in ANNs, Modular Neural Network (MNN) and Generalized Regression Neural Network (GRNN), to find a good modeling tool for the best management of the nutrients. The correlation coefficients (R) for the resulting predictions from the networks versus measured values were generally in the range of 0.70 to 0.75 in surface discharge forecasting, and 0.49 to 0.77 in nutrient predictions. Overall, MNN showed better simulation results to describe the water and nutrient mass dynamics. This study also discussed the issues of network optimization and computational efficiency. The practical implication in this study showed that the ANN technique performs well in predicting the rainfall-surface discharge process, and has relatively acceptable predictions in water quality forecasting.

Keywords:

• Modular neural network (MNN)
• generalized regression neural network (GRNN)
• rainfall
• surface discharge
• total nitrogen
• total phosphorus

Acknowledgement

The authors gratefully acknowledge the National Institute of Agricultural Science and Technology for providing the data to support this study.