Reducing Predictive Uncertainty in Real-Time Reservoir Operations by Coupling LARS-WG with ARNO Continuous Rainfall–Runoff Model

Abstract

In reservoirs with relatively low capacity-inflow ($\mathrm{C}/I$ ) ratios, real-time daily operations provide the possibility of minimizing the risk of failure and meeting downstream demands sustainably even during extreme droughts. In the present study, it has been attempted to generate synthetic daily precipitation and maximum and minimum temperature data using the widely applied weather generator Long Ashton Research Station weather generator (LARS-WG). The synthetic meteorological data were then inserted into the Arno River (ARNO) daily rainfall–runoff (R-R) model to reproduce long-term daily streamflow scenarios and subsequently used to determine optimal daily releases and operating policies. The daily R-R model can also produce appropriate predictions of future inflows to the reservoir, which can be applied to implementing real-time operations in daily time steps. The methodology used in this study provides two main advantages: (1) the capability to determine optimal daily releases for streamflow sequences, which can be applied to make optimal real-time decisions even by considering the unregulated downstream flow regime, and (2) the ability to predict future inflows and determine real-time decisions, which can be used to decrease predictive uncertainties of future releases.