In recent years, several contaminants have posed a threat to rivers, streams and lakes. The ability to analyse and anticipate water quality has emerged as an aid in the fight against the contamination of water. Various seasonal factors along with physicochemical properties influence water quality over time. Water quality data becomes time series data and the values of parameters change as meteorological conditions change over seasons at each location. Hence good time series analysis is required to forecast water quality. Considering the significance of Recurrent Neural Network (RNN) for time sequence data, this work is intended to build a water quality prediction model by learning seasonal patterns in the time series dataset. The dataset contains 10560 unique instances that describe both physicochemical and seasonal factors. Predictive models are developed using RNN and its variants GRU and LSTM and evaluated. Promising results are produced as a result of augmenting seasonal data with regular physicochemical properties while training the model.

River Water Quality, Prediction Model, Deep Learning Architectures, Physicochemical Parameters, Seasonal Parameters