Water quality management is a critical aspect of environmental sustainability, particularly in arid and semi-arid regions such as Iran where water scarcity is compounded by quality degradation. This study delves into the causal relationships influencing water quality, focusing on Total Dissolved Solids (TDS) as a primary indicator in the Karkheh River, southwest Iran. Utilizing a comprehensive dataset spanning 50 years (1968–2018), this research integrates Machine Learning (ML) techniques to examine correlations and infer causality among multiple parameters, including flow rate (Q), Sodium (Na+), Magnesium (Mg2+), Calcium (Ca2+), Chloride (Cl−), Sulfate (SO42−), Bicarbonates (HCO3−), and pH. For modeling the causation, the “Back door linear regression” approach has been considered which establishes a stable and interpretable framework in causal inference by focusing on clear assumptions. Predictive modeling was used to show the difference between correlation and causation along with interpretability modeling to make the predictive model transparent. Predictive modeling does not report the causality among the variables as it showed Mg is not contributing to the target (TDS) while the findings reveal that TDS is predominantly positive influenced by Mg, Na, Cl, Ca and SO4, with HCO3 and pH exerting negative (inverse) effects. Unlike correlations, causal relationships demonstrate directional and often unequal influences, highlighting Mg as a critical driver of TDS levels. This novel application of ML-based causal inference in water quality research provides a cost-effective and time-efficient alternative to traditional experimental methods. The results underscore the potential of ML-driven causal analysis to guide water resource management and policy-making. By identifying the key drivers of TDS, this study proposes targeted interventions to mitigate water quality deterioration. Moreover, the insights gained lay the foundation for developing early warning systems, ensuring proactive and sustainable water quality management in similar hydrological contexts.
1 September 2023 in Environmental Science and Pollution Research
Water quality assessment is definitely important, as the available water resources are highly stressed by population growth, climate change due to anthropogenic activities, and a significant change in consumption patterns. This study aims an innovative framework to predict the total dissolved solids (TDS) with more accuracy in the rivers, case study: “Karkheh River”, Iran, with the integration of signal analysis with machine learning algorithms. First, continuous wavelet transform (CWT) was applied to decompose the time series of water quality variables (e.g., Ca, HCO3, SO4, and Cl) into their trends, seasonality, and residuals, extracting features that capture temporal dynamics. These features served as input for non-linear machine learning models (XGBoost, Random Forest, Decision Tree) in differenct scenarios to compare which way of adding new feature would improve the model performance in terms of the TDS predictions. Adding new features characterized by only TDS signal analysis improved the TDS predictions and was compared with adding all variables signal characterization and compared with only using raw data to predict TDS level. Using a 50-year dataset from three different hydrometric stations, the models could achieve over 95% accuracy, and XGBoost outperformed others in terms of taking the advantage of the new extracted features from signals. The results indicates that signal-driven features significantly contribute to ccurately TDS prediction by 30% improvement in RMSE, and it can offer a scalable approach for real-time water quality monitoring in semi-arid river systems, which leads to a better early warning system for designing future mitigation strategies.
Meandering rivers are complex geomorphic systems that play an important role in the environment. They provide habitat for a variety of plants and animals, help to filter water, and reduce flooding. However, meandering rivers are also susceptible to changes in flow, sediment transport, and erosion. These changes can be caused by natural factors such as climate change and human activities such as dam construction and agriculture. Studying meandering rivers is important for understanding their dynamics and developing effective management strategies. However, traditional methods such as numerical and analytical modeling for studying meandering rivers are time-consuming and/or expensive. Machine learning algorithms can be used to overcome these challenges and provide a more efficient and comprehensive way to study meandering rivers. In this study, we used machine learning algorithms to study the migration rate of simulated meandering rivers using semi-analytical model to investigate the feasibility of employing this new method. We then used machine learning algorithms such as multi-layer perceptron, eXtreme Gradient Boost, gradient boosting regressor, and decision tree to predict the migration rate. The results show ML algorithms can be used for prediction of migration rate, which in turn can predict the planform position.
Managing salinity in arid rivers is impeded by sparse monitoring, relying on low-frequency grab samples that miss hydrological event dynamics. Here, interpretable machine learning is applied to a 50-year monthly archive (1968–2018; Discharge, major ions, pH) from three stations on Iran’s Karkheh River. Gradient Boosting Regression achieves high predictive skill for Total Dissolved Solids (TDS)/Electrical Conductivity (EC) (test-set R 2 = 0.94/0.97; RMSE = 55 mg L −1 /56 µS cm −1 ), validated via time-aware cross-validation. SHAP-based feature attribution reveals that Na + and SO 4–2 are the strongest contributors to TDS, while Na + and Cl − dominate EC, consistent with conservative salinity sources under baseflow conditions. A reduced-input decision tree (four predictors) retains R 2 = 0.81–0.87, enabling minimal-sensor monitoring. Flow-regime partitioning and STL (Seasonal-Trend decomposition using Locally estimated scatterplot smoothing)-detrended event composites reveal low-flow salinization and ion-specific post-flood recovery (Cl − : 1–2 months; Na + : 2–3 months), guiding targeted sampling. This framework extracts predictive power, process associations, and operational guidance from legacy grab-sample archives, scalable to data-limited basins worldwide.