Title: Predicting and forecasting water quality using deep learning

Authors: Ahmad Debow; Samaah Shweikani; Kadan Aljoumaa

Addresses: Higher Institute for Applied Sciences and Technology (HIAST), Damascus, Syria ' Higher Institute for Applied Sciences and Technology (HIAST), Damascus, Syria ' Higher Institute for Applied Sciences and Technology (HIAST), Damascus, Syria

Abstract: During the last years, obtaining water with acceptable quality for human consumption or even for agricultural applications is a big challenge in many places around the world. Water quality (WQ) can be defined by various factors like pH, turbidity, dissolved oxygen (DO), nitrate, temperature, total and faecal coliform. Therefore, prediction and forecast of WQ have become vital in order to monitor and control pollution. In this paper, 4-stacked LSTM models are developed to predict and forecast water quality index (WQI). Many algorithms are applied in this context to prepare the data like K-NN and annual mean, also for data analysis and features selection. The best prediction model is to predict without total coliform and RMSE value is 0.027, and the best forecasting method is filtering data with RMSE = 0.013. Models in this research can contribute in water management to avoid pollution as possible as we can.

Keywords: chi-squared; correlation matrix; deep LSTM; dissolved oxygen; forecasting; faecal coliform; K-NN; prediction; total coliform; water quality index; WQI; water quality class.

DOI: 10.1504/IJSAMI.2023.129858

International Journal of Sustainable Agricultural Management and Informatics, 2023 Vol.9 No.2, pp.114 - 135

Received: 30 Jun 2022
Accepted: 30 Aug 2022
Published online: 31 Mar 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article