<p>Reliable monitoring of reservoir water quality is essential for effective water resources management, particularly in regions facing water scarcity and increasing anthropogenic pressures. Conventional field-based monitoring is often constrained by limited spatial coverage, high operational costs, and low temporal resolution, which restricts comprehensive assessment of water quality dynamics. In this study, Sentinel-2 satellite imagery and in situ observations were integrated with machine learning techniques to model key water quality parameters in the Zayandehroud Reservoir, Iran. An artificial neural network (ANN) was developed to establish nonlinear relationships between spectral reflectance and seven water quality parameters, including chlorophyll-a, turbidity, Secchi disk depth, electrical conductivity (EC), total dissolved solids (TDS), nitrate (NO₃⁻), and pH. The model demonstrated strong predictive performance across both optically active and inactive parameters. For optically active variables, Nash–Sutcliffe efficiency (NSE) values ranged from 0.949 to 0.956, with coefficients of determination (R²) between 0.900 and 0.914. For optically inactive parameters, NSE values ranged from 0.924 to 0.949, and R² values varied between 0.854 and 0.901. Spatial distribution maps derived from the ANN outputs revealed pronounced spatial heterogeneity and seasonal variability across the reservoir, reflecting the combined influence of hydrological processes and external pollution sources. The results highlight the effectiveness of integrating Sentinel-2 imagery with machine learning models for high-resolution water quality monitoring. This approach provides a reliable and cost-effective framework for supporting data-driven decision-making in reservoir management, particularly in data-scarce regions.</p>