Water Quality Analysis and Prediction Using Hybrid Time Series and Neural Network Models

Abstract

Chagan Lake serves as an important ecological barri er in western Jilin. Accurate water quality series predictions for Chagan Lake are esse ntial to the maintenance of water environment security. In the present study, a hybri d AutoRegressive Integrated Moving Average (ARIMA) and Radial Basis Function Neural Ne twork (RBFNN) model is used to predict and examine the water quality [Total Nitrog en (TN), and Total Phosphorus (TP)] of Chagan Lake. The results reveal the following: ( 1) TN concentrations in Chagan Lake increased slightly from 2006 to 2011, though yearly variations in TP were not significant. The TN and TP levels were mainly classified as Grad es IV and V, (2) The hybrid ARIMA and RBFNN model’s RMSE values for the observed and predicted data were 0. 139 and 0.036 mg L -1 for TN and TP, respectively, which indicated that the hybrid model describes TN and TP variations more comprehensively and accur ately than single ARIMA and RBFNN model. The results serve as a theoretical bas is for ecological and environmental monitoring of Chagan Lake and may help guide irriga tion district and water project construction planning for western Jilin Province. Keywords: ARIMA model, Chagan Lake, RBFNN model, Total N, Tot al P.