Water quality index modelling and its application on artificial intelligence (AI) in conjunction with machine learning (ML) methodologies for mapping surface water potential zones for drinking activities
Abstract The deterioration of water quality has made monitoring condition and quality indicators in river water systems an urgent issue. The main causes of this decrease include inappropriate home waste disposal practices, the release of partially or completely untreated sewage, and industrial efflu...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-08-01
|
| Series: | Discover Civil Engineering |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44290-025-00298-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The deterioration of water quality has made monitoring condition and quality indicators in river water systems an urgent issue. The main causes of this decrease include inappropriate home waste disposal practices, the release of partially or completely untreated sewage, and industrial effluents into nearby bodies of water that are connected to river systems. However, traditional methods of evaluating the quality of water are expensive, labour-intensive, and complicated. For the foreseeable future, using affordable imaging offers tremendous potential in place of conventional field methods. To evaluate the contamination level, a basic standard reference i.e., World Health Organization guidelines is implanted to decipher the values ranging from natural to anthropogenic contribution.In the Mahanadi River Basin, Odisha, however, this study has highlighted the evaluation of surface water quality (WQ) for drinking reasons by the combined use of Machine Learning (ML) methodologies like Genetic Algorithm Particle Swarm Optimization-based WQI (GAPSO-WQI), with dependability-oriented decision-making approaches such as Firefly Algorithm (FA) and Algorithm of Weeds (AW), that have been used for river water quality monitoring and assessment due to their dependability and feasibility. The geographic variability of point properties is interpolated using geospatial techniques like Inverted Distance Weighted (IDW) and further, forecast for an unobserved site using neighbouring, well-known variables, and it is visualized as maps utilizing ArcGIS 10.5 software. For this purpose, water samples were collected from 16 different locations across the stretch and 21 parameters were analysed for a time frame of 2017 to 2024 respectively. The study area predominantly comprises of slightly alkaline in nature, and WQ parameters like TKN and coliform exhibits higher values and crosses the acceptable limits. Despite being an essential factor for rating of under exploitation water stations, WQI entails conflicting issues. The GAPSO-WQI method has been adapted and its value at all sampling sites, varied between 15 and 256. Thus, water comes under low to high class of water quality. At 8 survey sites of accounting 50%, it suggests that the primary reason for the adulteration of the river's water quality may be the degradation of home water supplies. Illegally disposed of municipal solid waste and agricultural runoff were the main contributors.As per AW framework, the study analysis revealed that 50% (n = 8) and 31.25% (5 locations) of the tested sites fall under low-medium water quality while, 18.75% of examined places confer in poor water category. However, except for 3 samples, all 13 samples were categorized as fair for human consumption. It was also confirmed that the primary cause of these two indicators is the influx of saltwater, the anthropogenic source of fluoride, and the combination of geogenic and human activity-derived origins for other metrics. On the other hand, the degree of FA, were implemented to reduce inconsistencies, involving WQI index. The factor weights and field data were taken into account while calculating the final ranking, and ultimately, that leads to compute a FA score.The degree of FA ranking were in the order of sites (Q) as: (9) >(8) >(16) >(2) >(7). Out of 16 water samples, Q-(9) was signified and referred as the most polluted location. Based on the ranking order, rest contaminated sites, following the site Q-(9) are: Q-(8), (16) and (2) respectively. The scores of FA were 8.75, 8.45 and 7.68, respectively. The result affirms chiefly about geogenic, anthropogenic, dissolution, precipitation reaction and concentration effects. Again, in this appraisal, we utilized three models- Cat Boost (Cat B), AdaBoost (AB), and Gradient Boosting (GB)- to estimate the specified river catchment’s suitability for surface water irrigation. According to our findings, the surface water is appropriate for irrigation because of the irrigation indices like, sodium percentage (%Na), sodium adsorption ratio (SAR), soluble sodium percentage (SSP), and potential salinity (PS), Kelly’s ratio (KR), and magnesium adsorption ratio (MAR). The irrigation parameter’s regional distribution revealed the direction of North East-South East for SAR, MAR, %Na, KR, SSP and PS respectively. Subsequently, irrigation parameters showed that the model as a whole did a good job of forecasting the suitability of irrigation. Based on the outcomes of RMSE, R2, Adjusted R2, MAPE, AIC and BIC, the GB and Cat B outperformed the AB model.For the purpose to achieve consistent and favourable outcomes in irrigation forecasting with surface water suitability, the current research suggests merging modelling approaches from Cat B, GB, and AB. The study suggests incorporating these modelling techniques because of their dependable and consistent outcomes, which can cut down on analysis time and expenses and be used both locally and internationally with comparable techniques. |
|---|---|
| ISSN: | 2948-1546 |