Modeling of Dissolved Oxygen Concentration Using Different Neural Network Techniques in Foundation Creek, El Paso County, Colorado

Abstract

The aim of this study is to examine the accuracy of two different artificial neural network (ANN) techniques, the multilayer perceptron (MLP) and radial basis neural network (RBNN), to estimate dissolved oxygen (DO) concentration. The ANN results are compared with multilinear regression (MLR) model. The neural network model is developed using experimental data collected from the upstream (USGS Station No: 07105530) and downstream (USGS Station No: 07106000) stations on Foundation Creek, CO. The input variables used for the ANN models are water pH, temperature, electrical conductivity, and discharge. The determination coefficient (R2), mean absolute error (MAE), and root mean square error (RMSE) statistics are used for the evaluation of the applied models. The MLP and RBNN models are also compared with MLR model in estimating the DO of the downstream station by using the input parameters of the upstream station. Comparison results indicate that the RBNN model performs better than the MLP and MLR models.