Forecasting models for surface water quality using predictive analytics

Modeling surface water quality has become crucial in providing better strategies for managing surface water resources, and adequate findings need accurate and geographically dispersed data. Hydrogeological modeling of these data sets is possible using empirically-based models. The other statistical models are also an alternative approach. In this study, a process with maximum probability is considered with the help of machine learning tools (MLT) to have optimized and valid output. The proposed method combines remote sensing and geographic information systems (RS and GIS) and MLT, which are appropriate for the predicament of neither small, large scale, nor long-term simulations. MLT methods such as VAR and ARIMA are developed in the Python programming with Jupyter notebook and tested according to the data in the spatial prediction for surface water quality parameters such as Tr, pH, Ec, TDS, AL, Ca++, NO−3, So, Cl, F−, Fe, and Mg2+ in the Krishna District, Andhra Pradesh, India—lower delta part. The delta with susceptible zones was identified using RS and GIS as those areas are prone to direct exposure to surface water contaminants from aquaculture, agricultural runoff, small- and medium-scale businesses, and household trash. Achieving effective surface water management for this ecosystem is critical for regional water management. The geographical information about the concentrations acquired via the RS and GIS was compared to the statistical modeling findings and verified using real-time measurements. MLT modeling seems more realistic than the experimental setting; data from the previous 20 years (2000–2020) were used for modeling, and the predicted values presented in the paper are predicted for the year 2021. The computed R2 value of ranges between 0.75 and 0.96% is recorded with ARIMA, and VAR posted range between 0.56 and 0.75% with the trained and tested data. The findings show the potential for MLT of geographically dispersed hydrogeological data to be used for pollution-free surface water management. From the surface water management perspective, combining RS and GIS and MLT offers an alternate data analysis approach for obtaining quick results utilizing a less laborious process that produces acceptable results.