Groundwater quality parameters prediction based on data-driven models

Mohammed Falah Allawi, Yasir Al-Ani, Arkan Dhari Jalal, Zainab Malik Ismael, Mohsen Sherif, Ahmed El-Shafie

Research output: Contribution to journalArticlepeer-review


Groundwater quality assessment is essential for achieving safe and sustainable water resources, specifically in regions that rely mainly on groundwater. This study focuses on evaluating groundwater quality metrics in the Alnekheeb basin located in Iraq to obtain a more suitable and sustainable water source, which plays a pivotal role in policy development and strategies for more efficient utilization of groundwater. In this regard, three groundwater water quality metrics presented in hardness, sodium absorption ratio (SAR), and salinity are purportedly predicted using two AI-driven models, namely the Radial Basis Neural Network (RBF-NN) and the Probabilistic Neural Network (PNN). Furthermore, this study investigates the influence of input parameters on the performance of the proposed models. Several water quality parameters, including SO4, Cl, NO3, Ca, Mg, Na, HCO3, and CO3, are used for the development modelling. The effectiveness of the proposed models is assessed using various statistical indicators and graphical presentations. According to the evaluation results, adding more input variables can sometimes increase the efficacy of the proposed models with regard to prediction accuracy. Moreover, the findings show that the PNN model provides a promising performance in predicting the groundwater’s water quality (WQ) matrices, showing superior performance compared to the RBFNN model.

Original languageEnglish
Article number2364749
JournalEngineering Applications of Computational Fluid Mechanics
Issue number1
Publication statusPublished - 2024


  • Artificial intelligence
  • Groundwater
  • Prediction
  • Water quality

ASJC Scopus subject areas

  • General Computer Science
  • Modelling and Simulation


Dive into the research topics of 'Groundwater quality parameters prediction based on data-driven models'. Together they form a unique fingerprint.

Cite this