Long-term ensemble forecast of snowmelt inflow into the Cheboksary Reservoir under two different weather scenarios

Hydrol. Earth Syst. Sci., 22, 2073–2089, 2018 https://doi.org/10.5194/hess-22-2073-2018 © Author(s) 2018. This work is distributed under the Creative Commons Attribution 4.0 License. Long-term ensemble forecast of snowmelt inflow into the Cheboksary Reservoir under two different weather scenarios Alexander Gelfan1,2 , Vsevolod Moreydo1 , Yury Motovilov1 , and Dimitri P. Solomatine1,3,4 1 Water Problems Institute of Russian Academy of Sciences, Watershed Hydrology Lab., Moscow, Russia 2 Moscow State University, Geographical Department, Moscow, Russia 3 IHE Delft Institute for Water Education, Chair of Hydroinformatics, Delft, the Netherlands 4 Delft University of Technology, Water Resources Section, Delft, the Netherlands Correspondence: Alexander Gelfan () Received: 30 June 2017 – Discussion started: 14 July 2017 Revised: 14 February 2018 – Accepted: 25 February 2018 – Published: 4 April 2018 Abstract. A long-term forecasting ensemble methodology, applied to water inflows into the Cheboksary Reservoir (Russia), is presented. The methodology is based on a version of the semi-distributed hydrological model ECOMAG (ECOlogical Model for Applied Geophysics) that allows for the calculation of an ensemble of inflow hydrographs using two different sets of weather ensembles for the lead time period: observed weather data, constructed on the basis of the Ensemble Streamflow Prediction methodology (ESP-based forecast), and synthetic weather data, simulated by a multisite weather generator (WG-based forecast). We have studied the following: (1) whether there is any advantage of the developed ensemble forecasts in comparison with the currently issued operational forecasts of water inflow into the Cheboksary Reservoir, and (2) whether there is any noticeable improvement in probabilistic forecasts when using the WGsimulated ensemble compared to the ESP-based ensemble. We have found that for a 35-year period beginning from the reservoir filling in 1982, both continuous and binary modelbased ensemble forecasts (issued in the deterministic form) outperform the operational forecasts of the April–June inflow volume actually used and, additionally, provide acceptable forecasts of additional water regime characteristics besides the inflow volume. We have also demonstrated that the model performance measures (in the verification period) obtained from the WG-based probabilistic forecasts, which are based on a large number of possible weather scenarios, appeared to be more statistically reliable than the corresponding measures calculated from the ESP-based forecasts based on the observed weather scenarios. 1 Introduction Spring freshets are a hydrological phenomenon of which magnitude is highly dependent on the amount of water accumulated on the surface and in subsurface storages of the river basin during several months prior to the snowmelt. This dependency serves as a physical basis for the predictability of spring runoff (Li et al., 2009). As stated by Lettenmaier and Waddle (1978, p. 1), “snowmelt runoff is one of the few natural phenomena for which relatively accurate long-term forecasts can be made”. Implementation of this opportunity is crucial for the water reservoirs of the Volga-Kama reservoir cascade (VKRC) in Russia – one of the world’s largest multi-purpose water management systems. The VKRC is located within the largest European river basin, the Volga River basin (area of 1 350 000 km2 ), and consists of 11 reservoirs that hold from 1 to 58 km3 of water. It is used to conduct seasonal and multi-year flow regulation. The VKRC was designed to redistribute the highly uneven runoff of the Volga River, with two-thirds of the annual runoff volume occurring during the 2–4 months of the spring–summer freshet. This task, aimed at optimizing reservoir management for power production, navigation and flood protection, is even more complex due to the requirement of annual spring water release to Lower Volga aimed at allowing for sturgeon spawning. Such release that is regulated over several weeks with a predefined amount and temperature of water during the spring freshet is an extremely complex task for water management (Avakyan, 1998). Hence, a reliable and firsthand forecast of snowmelt Published by Copernicus Publications on behalf of the European Geosciences Union. 2074 A. Gelfan et al.: Long-term ensemble forecast of snowmelt inflow into the Cheboksary Reservoir inflow into the VKRC reservoirs is crucial for decision makers. By the mid-1960s, the specific methods were developed which underlie the contemporary operational forecast for VKRC management (water supply forecast). For different reservoirs, the produced forecasts are based on two primary techniques: the index methods and the so-called physical– statistical methods (Gelfan and Motovilov, 2009; Borsch and Simonov, 2016). Both methods produce deterministic (despite the term “physical–statistical”), purely data-driven forecasts and relate the predictors (such as initial snow water equivalent, soil freezing and soil moisture indices, precipitation amount for the forecast period) to the main predictand – the spring inflow into a reservoir. The initial basin characteristics are derived from observations; yet the precipitation amount is typically set to the climatic mean. The operational water supply forecasts’ methodology is used in real practice by water managers and has remained unchanged over the past half-century. While the utility of data-driven flow forecasts (which currently may be based on advanced statistical and machine learning techniques) has been demonstrated through various examples (see e.g. Abrahart et al., 2012), their skill and reliability depend on the amount and stationarity of available data and they are not always adequate. It would be difficult to expect a forecast improvement within the existing framework of the purely data-driven approach because of the reduction of the observational network in the Volga basin (estimated at 30 % in Borsch and Simonov, 2016), the non-homogeneity of the observations caused by changes in the measurement techniques and changes in climate, land use and so on. An opportunity to improve the operational water supply forecasts of water inflow into the VKRC lies in shifting from the traditional exclusively data-driven forecasts towards hydrological model-based forecasts, and from a deterministic methodology to one using ensembles with a possibility of characterizing forecast uncertainty. During the last 20– 30 years there has been a general understanding of the necessity of such a shift to Ensemble Streamflow Prediction (ESP) systems (e.g. Day, 1985) and a considerable research effort in this direction (Franz et al., 2003; Wood and Lettenmaier, 2006; Li et al., 2009; Shukla and Lettenmaier, 2011; Yossef et al., 2013; Najafi and Moradkhani, 2016; Demirel et al., 2015; Beckers et al., 2016; Arnal et al., 2017; Mendoza et al., 2017). Such systems are currently used more and more (...truncated)