<p>Addressing uncertainty related to reservoirs' evaporation estimation is vital for making informed decisions such as investment in reducing evaporation from reservoirs or distribution of water to consumers. The present study aims to assess the uncertainty of evaporation estimation models by applying the Bayesian model averaging (BMA) approach and examine their impact on the operation of the Zayandeh-Rud dam, in central Iran. In this regard, the fourteen empirical evaporation estimation models from the Temperature group, Dalton group, Solar Radiation-Temperature group, and Temperature-Day length group were chosen to estimate evaporation, and the uncertainty of these models was handled by the BMA. Class A pan evaporation was used as a reference method for estimating evaporation from the Zayandeh-Rud reservoir. For comparing the effects of models on the accuracy of BMA, five different sets of ensembles for the combination of models’ prediction are considered (BMA (1), BMA (2), BMA (3), BMA (4), and BMA (5)). The results indicate the Hamon model had lower Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) in almost all months than individual models. The BMA ensembles outperform all individual models; in particular, BMA (3) in July improves the RMSE, MAE, Nash–Sutcliffe Efficiency (NSE), and Index of Agreement (IOA) of the Hamon model by 13.36%, 40.6%, 200%, and 78.8%. It is worth noting that the behavior of the system was influenced by the model used; models which underestimated evaporation indicate a more resilient reservoir, whereas models that overestimated evaporation indicate a more vulnerable reservoir. The performance of the reservoir after employing the BMA schemes was closer to the real state.</p>