A comparative analysis of statistical, MCDM and machine learning based modification strategies to reduce subjective errors of DRASTIC models

• Published in: Environmental earth sciences Vol. 83; no. 7; p. 211
• Main Authors: Dasgupta, Rijurekha, Banerjee, Gourab, Hidayetullah, Sekh Mohammad, Saha, Nilanjan, Das, Subhasish, Mazumdar, Asis
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2024; Springer Nature B.V
• Subjects: Aquifers; Arsenic; Biogeosciences; Comparative analysis; Contamination; Correlation coefficient; Correlation coefficients; Crop production; Developing countries; Drinking water; Earth and Environmental Science; Earth Sciences; Environmental Science and Engineering; Errors; Geochemistry; Geology; Groundwater; groundwater contamination; Groundwater pollution; Groundwater quality; Hydrology/Water Resources; India; Irrigation water; LDCs; Machine learning; Mitigation; Modelling; Original Article; Pollutants; Pollution index; Public health; risk; Statistical analysis; Statistical methods; Statistics; Subjectivity; Sustainability; Sustainability management; Terrestrial Pollution; Vulnerability; Water pollution monitoring; Water purification; Water quality; India; Aquifer vulnerability; Multi criteria decision making (MCDM); Convolutional neural network (CNN); Modified DRASTIC model
• DOI: 10.1007/s12665-024-11515-3
• ISSN: 1866-6280

Groundwater, the main source of drinking and irrigation water in developing countries, plays an important role in maintaining public health and crop production. However, groundwater quality is often compromised and geological aquifers that contain this valuable resource are at risk of contamination such as arsenic and emerging pollutants in the North 24-Parganas district of India. Monitoring of aquifer vulnerability is generally done by modeling which subjects to errors that must be taken into account. For an emerging vulnerability model DRASTIC, prone to subjectivity errors, three modification methods: statistical modification, DRASTIC-AHP, and DRASTIC-CNN, are used to identify the most effective strategy for reducing such subjectivity errors in this study. The model results are validated against the composite groundwater pollution index. The DRASTIC-CNN method outperforms other modifications, with the correlation coefficient increasing from 0.226 to 0.9 compared to the conventional DRASTIC model. The coefficient of determination for DRASTIC-CNN improves from 0.05 to 0.81, while the kappa statistic underscores its superiority in identifying vulnerable regions. Although statistical modification and DRASTIC-AHP demonstrate respectable performances, they are lower than the accuracy achieved by DRASTIC-CNN. DRASTIC-CNN emerges as the most efficient approach to reduce subjective errors in modeling aquifer vulnerability. This study accurately identifies groundwater vulnerability, helping policymakers to formulate sustainable management plans and mitigation measures for water purification. It importantly addresses environmental problems such as arsenic contamination and emerging pollutants. Integrating the DRASTIC model with other factors allows us to assess vulnerability and pave the way for sustainable treatment technologies.