Artificial Neural Networks Used to Assess Groundwater Quality and Health Risks for Schoolchildren in Bangladesh

Strategic Insights -

Researchers at Islamic University in Kushtia, Bangladesh, have used artificial neural networks (ANN) to assess the quality of groundwater in the southwest region of the country, specifically in schools. Their study, published in the Environmental Geochemistry and Health journal, focused on the presence of iron (Fe), arsenic (As), pH, electrical conductivity (EC), and total dissolved solids (TDS) in tubewell water. The researchers, led by Md Anisul Kabir, found that 68% and 48% of the water samples exceeded World Health Organization (WHO) and United States Environmental Protection Agency (USEPA) limits for Fe and As, respectively.

Key Takeaways:

  • The study assessed the quality of tubewell water in the southwest region of Bangladesh, focusing on Fe, As, pH, EC, and TDS.
  • Groundwater is a vital source of drinking water in Bangladesh, with tubewells commonly used, particularly in schools.
  • Financial support for this research came from The University of Newcastle.
  • Using ANN modeling, daily safe water intake limits for children were estimated based on Fe and As levels.
  • A total of 75 school-based water samples were collected, with 68% and 48% exceeding WHO and USEPA limits for Fe and As, respectively.
  • Spatial mapping identified Mahajanpur, Bagoan, and Dariapur as hotspots for contamination.
  • TDS showed a strong positive correlation with conductivity (r = 0.92) and a moderate one with salinity (r = 0.7), indicating their interdependence.
  • ANN and Decision Tree Regression models showed 87% accuracy in estimating safe water intake, highlighting Monakhali Union as the most vulnerable area.
  • High health risks were identified for Fe in Mahajanpur and As in Dariapur.
  • To avoid health impacts in children, safe consumption levels in Baguan and Monakhali were notably low at 0.59 L/day and 0.39 L/day, respectively.

Statistics:

  • 68% of water samples exceeded WHO limits for Fe.
  • 48% of water samples exceeded USEPA limits for As.
  • 56.74% of variance was explained by PC1 and PC2, with Fe and As weakly correlated in PC1 but negatively loaded in PC2.
  • Linear regression showed Fe (r = -0.34) and As (r = -0.71) decreased with depth.
  • ANN models showed 87% accuracy in estimating safe water intake.
  • Fe and As levels in water samples from Mahajanpur and Dariapur were significantly higher than the safe consumption limits.

Sources:

  • Groundwater quality assessment and health risk evaluation for schoolchildren in Mujibnagar, Bangladesh: safe consumption guidelines using artificial neural network modeling. Environmental Geochemistry and Health, 2025;47(8):324.
  • Islamic University Reports Findings in Artificial Neural Networks (Groundwater quality assessment and health risk evaluation for schoolchildren in Mujibnagar, Bangladesh: safe consumption guidelines using artificial neural network modeling). Journal of Engineering. August 4, 2025; p 1992.