Monitoring and predicting river water quality is crucial for urban water management, agriculture, and environmental sustainability, especially in hot and arid regions. The presence of dams along the river course can significantly alter water quality by affecting flow regimes and salt accumulation, potentially leading to increased salinity and other related problems. However, such changes can be effectively managed through accurate modeling and forecasting. This study assesses the performance of two machine learning models, Random Forest (RF) and Long Short-Term Memory (LSTM), in predicting electrical conductivity (EC) and sodium absorption ratio (SAR) within the Maroon River Basin, Iran. Principal Component Analysis (PCA) and stepwise regression were employed to reduce input dimensionality and enhance model efficiency. Results indicate that the LSTM model consistently outperforms RF at both Idank station (upstream of Maroon Dam) and Tang-e-Tekab station (downstream of Maroon Dam) for both parameters, particularly in SAR prediction. At Idank station, the LSTM model combined with stepwise regression achieved the highest accuracy for EC prediction, with an R² of 0.96, RMSE of 61.56, and KGE of 0.96 on the test dataset. For SAR at the same station, LSTM again demonstrated exceptional performance, attaining an R² of 0.99, RMSE of 0.08, and KGE of 0.99. At Tang-e-Tekab station, LSTM with PCA yielded the most precise EC predictions (R² = 0.96, RMSE = 76.60, KGE = 0.97). Similarly, the best SAR predictions at this station were obtained using LSTM with PCA (R² = 0.96, RMSE = 0.18, KGE = 0.95). These findings underscore the effectiveness of combining LSTM networks with tailored input selection techniques based on site-specific conditions, highlighting their potential application in water resource decision support systems. Overall, this study demonstrates that although dam operations influence water quality, such impacts can be successfully managed through advanced predictive modeling to facilitate sustainable water resources management.