Hybrid Algorithm for early Detection of Water Pollution Impact on Environmental Indicators using Wavelet Techniques and RBF Neural Network Learning

The present study examines the impact of water pollution on the environment with the aim of detecting early abnormalities or significant changes in the water pollution indicators. A hybrid algorithm based on wavelet techniques and radial basis function neural network learning using high-frequency surrogate relation is introduced. Important qualitative indicators such as phosphate, nitrate, and chemical oxygen demand (COD) in the water bodies have uncertainties with variations such as dependence, and effectiveness of physical and chemical factors. In the first step, the high-frequency time series of the main TP index is obtained through the surrogate model and compared with GARCH techniques. By using the wavelet transform, the noise components of the time series are removed and pre-processed. In the next step, it is created by using the neural network to identify the main characteristics of water quality. In the last step, the contamination threshold is calculated based on the estimated base pattern for analyzing statistical patterns. The results show that the proposed algorithm has high stability and accuracy because using the surrogate technique has extracted a more accurate model of the behavior of the required water variables. It can be used to manage surface runoff in watersheds to preserve the environment and improve water quality.