Water is essential for life, as it supports bodily functions, nourishes crops, and maintains ecosystems. Drinking water is crucial for maintaining good health and can also contribute to economic development by reducing health-care costs and improving productivity. The present study evaluated the surface water quality of Mahanadi River (Odisha, India). Hence, to evaluate the hydro-chemical processes, sources of contamination, and water quality, a methodical examination was conducted using an integrated approach, namely Bayesian Approximation (BA), Analytical Hierarchy Process (AHP)-Technique of Order of Preference by Similarity to Ideal Solution (TOPSIS), Fuzzy Logic (FL), Multi-Objective Optimization on the Basis of Ratio Analysis (MOORA), and Random Forest (RF) method. For this, water samples from 16 locations were taken for a period of 2018–2024, to test 21 physicochemical parameters in the selected sampling sites. From the assessment of parameters, with respect to WHO standards, pH indicates alkaline, TKN, and TC in all samples surpassed the prescribed drinking water limit. However, major ion and hardness spatial interpolation maps typically show that the quality of the water declines from upstream to downstream, with extreme values found in the downstream. The index for BA-WQI value revealed that 50% of samples belong to unsatisfactory water quality. This was also accompanied by several parameter's high values, namely TDS, NO3−, Cl−, and SO42−, which were also highest among all the locations. Again, it is noticed that 12.50% of sites come under the zone of excellent water. However, 37.50% of samples indicated a good water class. As a result, a renowned MCDM model, such as AHP-TOPSIS, was presented, which makes use of rough set theory and Bayesian weights to provide a trustworthy and objective assessment of the total pollution levels at each sample site. Hence, this innovative technique depicted that W-(9) was the most polluted region if compared to other places, followed by W-(8), (16), (2), and (7), respectively. Based on FL-DWQI values, 12.5% of monitored specimens point towards excellent category, and rest 18.75% indicated as good quality. The remaining samples, or 68.75%, consist of 'poor, very poor, and unsuitable qualities'. However, it was relevant that the degree of pollution at these stations was more closely linked to a variety of expanding human activities, such as excessive water use, fertilizer effects, agricultural runoff, and industrial activity in and around the river corridor. Additionally, MOORA has been conducted and performance scores were extracted. These four polluted sites such as W-(9), (8), (16), and (4), which contain higher performance scores, were 0.89, 0.093, 0.06, and 0.04. The major four places containing variables that exceeded the WHO limits, which account for TKN, coliform, and EC properties, were named accordingly. It was discovered that the main causes of the river's water quality adulteration were agricultural runoff and home waste water. Furthermore, a RF analysis of the 16 sites was carried out and five critical variables such as Cl−, TH, TDS, EC, and TC were obtained on the basis of R2 and RMSE score. Here, the first four RF factors were sufficient to explain 83.86%, 84.27%, 84.14%, and 85% of the model accuracy for correlation matrix. In the end, the target TC suggests about 89% of the total accuracy. Afterwards, water quality at all sampling locations was expressed in terms of RF-WQI. The value obtained varied between 15 and 243, denoting good to poor water. The main finding of the investigation was that at eight inadequate water sites, the main sources of adulteration of the river's water quality were agricultural runoff, illegally deposited municipal solid waste, and deteriorating household water supplies. This work highlights the viability and dependability of integrating these techniques for monitoring and evaluating river water quality. Hence, these findings are essential for comprehending surface water sustainability, for human consumption in the research area.
