Development of the support vector regression-particle swarm optimization simulation-optimization model for the assessment of a novel groundwater quality index

The assessment and prediction of the groundwater resource quality are required for the sustainable management of this crucial resource. This study develops a new index for assessing and a model for predicting the quality of groundwater resources. The groundwater quality index (GWQI), the Shannon entropy method, was used to determine the weight of parameters, and the complex proportional assessment multi-decision criteria method was used to score the GWQI. Water quality parameters, including TDS, EC, TH, Mg+2, Na+, SO42-, pH, K+, Cl-, HCO3- and Ca+2, were used as decision criteria. The support vector regression-particle swarm optimization )SVR-PSO( simulation-optimization model is developed to predict new GWQI (C-GWQI) of the aquifer. The development of this new index called C-GWQI is one of the innovations of this article. Based on these approaches, the index is used to determine three water quality classes (optimum, permissible, and impermissible) for drinking water following World Health Organization (WHO) criteria. The distribution of C-GWQI shows that groundwater quality in most of the Zanjan aquifer of Iran was in the optimum range. Still, it is deteriorating into the permissible range due to pollution from urban areas during some periods. The hybrid SVR-PSO model can predict the groundwater quality with sufficient accuracy with a Mean Absolute Relative Error (MARE) of 1.5% and 0.88% in training and testing phases, respectively. Results show that temperature, precipitation, evaporation, returned water and groundwater level did not significantly affect groundwater quality prediction. In contrast, the previous month's C-GWQI, recharge, and discharge were most influential in predicting groundwater quality.

Automated summary

This study develops a new index and model for assessing and predicting the quality of groundwater resources. The index is used to determine three water quality classes for drinking water and the model predicts the groundwater quality. The results show that factors such as temperature and precipitation have minimal impact on groundwater quality prediction, while the previous month's groundwater quality index, recharge, and discharge have the most influence.