Post-processing of reservoir releases to support real-time reservoir operation and its effects on downstream hydrological alterations

Abstract

Reservoir operation relies on two sets of flow data to conduct error analysis: inflow data and release data. Previous studies applied only the inflow data to reduce errors and improve the reservoir’s operational effectiveness. Little attention has been paid to the accuracy of the release data. However, release data provides more accurate information than inflow data for the protection of downstream river ecosystems. In this paper, we aim to illustrate the necessity of fully using the release data to improve a reservoir’s operational performance in a real-time reservoir operation system. Firstly, we designed a hypothetical reservoir system, performed five numerical experiments with different ranges of model parameters, and found that the error variances of reservoir releases were always larger than those of inflows. This indicates that the post-processing of reservoir releases is required to improve the accuracy of release data and support real-time reservoir operation. Then, we used a Bayesian joint probability (BJP) model for post-processing release data to reduce errors and quantify the uncertainty. We also applied the ecodeficit and ecosurplus based on flow duration curves to evaluate downstream hydrological alterations and further demonstrate the necessity of post-processing of release data. Results showed that the BJP model led to considerable improvement in the root-mean-square error and the continuous ranked-probability score. This demonstrates that the BJP model was effective in reducing errors. Our results highlight the importance of applying release data and improving their accuracy to support reservoir operation. If the release data is neglected, operational performance will degrade.