Sustainable reservoir management is essential to ensure the productivity of agriculture and to adapt to a changing climate . With the global population projected to reach about 9.5 billion by 2050 , irrigated agriculture will be even more essential to meet food requirements, since irrigation enhances agricultural reliability and food productivity. In southern Alberta, numerous dams are operated for irrigation, which accounts for 60-65% of the province's annual water consumption . However, growing water conflicts with a rising population and increasing economic development, changing supply and demand with climate change, and an increasing emphasis on ecological concerns all threaten water availability in the South Saskatchewan River Basin . To assess the effects of changes to current reservoir management practices, numerous water resources modeling studies have been conducted. Their aim is to quantify spatially and temporally the best allocation of available water under different optimization methods and management policies in order to minimize drought impacts on crop yields and other water use sectors [2, 5, 10]. Despite progress in reservoir modelling and management, academia still faces the challenge of applying the output of optimization models to real-world reservoir operations [6, 8, 12]. Until solutions are identified to bridge the gap between theory and practice, innovative advances in water resources management are limited to theoretical applications. To improve applicability of reservoir management theory, this research analyzes reservoir managers' perspectives in Alberta's most allocated river basin: the South Saskatchewan River.
Our research approach aims to improve understanding of the behaviour of reservoir operators under different climatic and hydrological conditions. The method involves the collection of interview data as opposed to the more common approach, which analyzes historical hydrological data. For example, many studies have attempted to understand reservoir operators' decisions by analysing the relationships of historical releases to different factors such as current inflow, previous releases and previous storage using statistical tools . In contrast, this research project has investigated water managers' practices regarding reservoir operations through in-person interviews with Irrigation Districts from all over Southern Alberta. The interview questions were oriented to connect reservoir operators' practice with optimization model development.
In optimization models for reservoir operations, the costs and benefits related to optimal reservoir operating rules, the water supply allocations and the physical characteristics of the water network are defined mathematically with an objective function, decision variables, and constraints . In western Canada, many optimization models such as the Water Resources Management Model (WRMM), developed by the Government of Alberta, are based on the water licenses system of "first-in-time, first-in-right (FITFIR)" . However, the data collected suggest that the seniority-based allocation priorities of senior versus junior water users are generally not applied in real-world operations of reservoirs. Instead, cooperation between districts and between irrigators within a district indicates that optimization model's water allocations should be driven principally by the infrastructure capacity on a river basin-scale basis. Moreover, water managers do not seem to apply annual or inter-annual water deficit-distribution strategies, but instead impose variable water rationing for all irrigators at the beginning of a growing season. These findings contrast with theoretical optimization methods which often aim to equalize the water supply deficit over the total simulation length without considering irrigator's actual strategies at the beginning of the growing season. (It is probable that the actual practice described above is non-optimal, compared with modelling results obtained without irrigators' adaptation strategies; however, an understanding of irrigators' and district water managers' perspectives is nonetheless useful and informative.) Therefore, the real benefits obtained from approaches like "multiple timestep optimisation (MTO)", which aim to derive optimal rules curves of reservoirs based on foreknowledge of water supply and water demand, should assess early season adaptations as well. Of additional importance is recognition of the "day-by-day" approach adopted by all water managers interviewed who will "never sacrifice today for tomorrow" (Philips, 2014, personal communication). Their release decisions are a function of today's water availability and today's water demand, and do not consider possible risks for "tomorrow". This approach correlates well with the "single timestep optimisation (STO)" method, which optimizes water allocation at each timestep. The innovative contribution of our research findings is to provide real-world data and a better understanding of water managers' perspectives.
In conclusion, sustainable irrigation management is essential to face the challenges that a growing population and climate change pose for food security. To be effective, the accuracy and application of optimization models for irrigation management have to be improved through incorporation of water managers' experience. Key results from interviews with water managers of southern Alberta suggest that the rules behind water allocations modelling should be oriented toward basin scale cooperation and accounting for the effect of early season rationing in water uses as well as day-by-day releases strategy. As a result, more valuable outcomes could be obtained from modelling studies, with operating rules that could be more easily applied and results that could be more readily accepted by water managers.
Keywords: Irrigation, Operators' decision-making, Optimization models, Reservoir operations, Stakeholder interviews