Probabilistic approach of reservoir level depletion induced by drought

Droughts are one of the most complicated natural disasters on earth. The repetitive occurrence of droughts has enormous adverse impacts on different aspects of human lives and natural environment. Careful monitoring and early warning systems can assist in the development of effective drought management strategies. Therefore, it is of immense significance to have a full understanding of the characteristics of a developing drought (severity, frequency etc.) before planning any drought response measures. The main aim of this research is to develop a methodology to evaluate reservoir storage levels during drought periods in a probabilistic way. In doing so, a case study was conducted of the Upper Yarra reservoir, which is located in the upper part of the Yarra River catchment in Australia. In order to identify the impacts of drought on this reservoir, it is important to have detailed knowledge of the general drought conditions surrounding this reservoir, as major portions of its inflow are harvested from neighbouring areas. Therefore, a comprehensive investigation of drought characteristics over this area is essential. Six rainfall and six streamflow stations near the Upper Yarra reservoir were selected for evaluating meteorological and hydrological drought events using the Standardized Precipitation Index (SPI) and the Standardized Hydrological Drought Index (SHDI), respectively. Both of these indices detected drought events successfully when applied to the data. Univariate and bivariate frequency analysis of drought duration and severity were carried out using the Gumbel-Hougaard copula. A probabilistic assessment of the reservoir storage condition was carried out by joint consideration of probability of initial storage volume and probability of drought events affecting inflow to the reservoir. Therefore, frequency analysis of drought events of inflow to the reservoir with particular severity and duration were conducted before applying them to the reservoir system model with specific initial water levels. The quantitative exploration of trends of drought characteristics (e.g. severity, frequency) provides meaningful insight to water authorities for developing of drought management plans. This study employed basic and modified Mann-Kendall tests to detect monotonic trends in drought characteristics. Both tests identified significant decreasing trends for four stations in the study area. More specific results of trends were reported by Innovative Trend Analysis (ITA) method. The results indicate that extreme drought situations are more likely to appear at the Reefton, Warburton, Alderman Creek, Little Yarra and McMahons Creek stations. Using the Sequential Mann-Kendall test, it was observed that the starts of the abrupt change points for most stations were found during the Millennium Drought (1996 to 2009) in Victoria. The changing patterns of drought frequencies were also investigated using the Poisson regression method. All stations exhibited decreasing trends in inter-arrival times between successive drought events, indicating that droughts are becoming more frequent in this catchment. The integrated modelling software Source is used to construct a reservoir system model. The development of water demand function is an essential requirement for building of the reservoir system model by Source software. Multiple regression analysis (MRA) and principal component analysis (PCA) are used and, finally, PCA was selected for development of water demand function because PCA gives better results than MRA. This study determines a risk assessment of storage condition of the Upper Yarra reservoir due to impacts of drought events. A probabilistic approach is proposed, taking into account the variability of reservoir storage volume prior to a drought event and different drought scenarios. Both drought severity and durations are included in developing drought scenarios. All required inputs are used in Source software to determine the reservoir storage volume at the end of a drought event. The analysis is performed for Period 3 (June to August, the most critical time of a year in terms of availability of water in the reservoir) and Period 1 (December to February, the least critical time). Three prespecified storage conditions are studied: (1) when storage drops < 50% of its full supply volume (FSV) (CC1); (2) when storage drops < 40% of FSV (CC2); and (3) when storage drops < 30% of FSV (CC3). The main conclusions of these analyses are summarized as follows: 1) the probability of storage reduction below the prespecified conditions is higher in Period 3 than in Period 1; 2) the risk of storage reduction can be successfully evaluated based on two uncertain parameters (initial storage volume and drought severity) and the results show that the initial storage volume is a more dominant uncertain parameter in probability calculation than drought severity for long as well as short-duration droughts; 3) several drought zones are successfully constructed for each condition on plots of initial storage vs. drought severity. It should be noted that each zone is constructed for a specific drought duration and period. If needed, other zones can be developed for other periods and drought durations following the same approach; 4) the constructed zones will give indications to water authorities about the reduction of storage due to long- and short-duration drought events; 5) finally, the general form of the relationship between initial storage volume and probability of storage reduction below any particular level for any drought event of known duration and severity is developed. Results of this study provide a technical reference for the risk assessment of reservoirs due to drought events and will assist in the development of appropriate action plans