Water Quality Prediction Based on an Improved ARIMA- RBF Model Facilitated by Remote Sensing Applications

摘要

Remote sensing technique are great used to assess and monitor water quality. An efficient and comprehensive method in monitoring water quality is of great demand to prevent water pollution and to mitigate the adverse impact on the livestock and crops caused by polluted water. This study focused on a typical water area, where eutrophication is the main problem, and thus, the total nitrogen was chosen as an important parameter for this study. The research contains two parts. The first part is the methodology development, an algorithms was proposed to inverse the total nitrogen (TN) concentrations from the field imagery acquisition. The squared correlation coefficient between the inversion values and measured values was 0.815. The second part is the deduction of water quality parameter (TN) from upstream to downstream. An improved hybrid model of Autoregressive Integrated Moving Average (ARIMA) model and Radial basis function neural network (RBF-NN) was developed to simulate and forecast variation trend of the water quality parameter. We evaluated our method using data sets from satellite. Our method achieved the competing predicting performance in comparison with the state-of-the-art method on missing data completion and data predicting. Generally, the evaluation results indicated that the developed methods were successfully applied in forecasting the water quality parameters and filling in missing data which cannot be inversed in space by satellite images due to the cloud and mist interference, and were of promising accuracy.