Groundwater quality prediction based on LSTM RNN: An Iranian experience

Groundwater quality prediction has practical significance for the prevention of water pollution. Based on the exogenous variables which are effective on water quality indicators, this paper proposes a new method with new effective parameters based on LSTM RNN for groundwater quality index prediction. The effective parameters on the groundwater quality index include rainfall rate, temperature, and humidity, and groundwater abstraction was collected. Monthly time series data selection was done from five different locations in the Damavand region in Iran from 2009 to 2021. Neural network architecture is tested by “f-score” tested to obtain the best neural network performance. A comparison of the real value and the result of the prediction show that the water quality index prediction has been done sensibly and quite properly in most cases.