Monitoring the water depth of agricultural ponds is crucial for the optimal management of water resources and irrigation planning, despite the implementation challenges and numerous limitations involved. This research aimed to develop a cost-effective remote sensing-based model to estimate water depth in geomembrane-lined ponds. The study utilized 190 field measurements of water depth from 14 ponds over a five-month period (from Azar 1403 to Farvardin 1404) in the Majan region (Shahroud County), along with concurrent Sentinel-2 satellite imagery. A comprehensive set of 96 independent variables (12 spectral bands and 84 spectral indices) was extracted. After data refinement, a stepwise linear regression algorithm was employed to automatically select the most effective variables and optimally model the relationship between the spectral variables and water depth. Analysis of the spectral reflectance variation curve revealed a stable decreasing trend with increasing water depth. Examination of correlation matrices uncovered distinct patterns in the relationships between the bands and spectral indices. Using stepwise regression, the final model was developed with four selected spectral indices (NGBDI, B1_7, B1_9, and B6_9). This model demonstrated appropriate accuracy and reliability, with a coefficient of correlation (R) of 0.75, a root mean square error (RMSE) of 70 cm, and a mean absolute percentage error (MAPE) of 9.41%. Statistical evaluations, including analysis of variance (ANOVA) and diagnostic plots, confirmed the model's acceptable performance. The findings affirm the effectiveness of remote sensing technology as a precise, cost-effective, efficient, and scalable tool for monitoring the water depth of agricultural ponds, aiding water resource managers in informed decision-making for agricultural water security.