Prediction of outlet dissolved oxygen in micro-irrigation sand media filters using a Gaussian process regression

Abstract

Sand media filters are a key component of micro-irrigation systems since they help preventing emitter clogging, which greatly affects the system performance. Dissolved oxygen is an irrigation water quality parameter related to organic matter loading. Low values of dissolved oxygen can cause crop root hypoxia and, therefore, agronomic problems. Thus, an accurate prediction of dissolved oxygen values could be of great interest, especially if effluents are used in micro-irrigation systems. The aim of this study was to obtain a predictive model able to forecast the dissolved oxygen values at the outlets of sand media filters. In this study, a Gaussian process regression (GPR) model was used for predicting the output dissolved oxygen (DOo) from data corresponding to 547 filtration cycles of different sand filters using reclaimed effluent. This optimisation technique involves kernel parameter setting in the GPR training procedure, which significantly influences the regression accuracy. To this end, the height of the filter bed, filtration velocity and filter inlet values of the electrical conductivity, dissolved oxygen, pH, turbidity and water temperature were monitored and analysed. The significance of each variable on filtration performance is presented and a model for forecasting the outlet dissolved oxygen obtained. Regression with optimal hyperparameters was performed and a coefficient of determination of 0.90 for DOo was obtained when this new predictive GPR–based model was applied to the experimental dataset. Agreement between experimental data and the model confirmed the good performance of the latter.