Increasing Water Availability Within the SSRB

History has taught Albertans that striking a balance between storing too little and too much water, and managing an unpredictable supply, is not simple. The droughts of 1992 and 2001 combined with the floods of 2005 and 2013 have made it clear that we cannot rely solely on current modeling and management approaches to meet the goals of Alberta’s Water for Life strategy (i.e., a safe, secure drinking water supply; healthy aquatic ecosystems; and reliable, quality water supplies for a sustainable economy). This article explores how we might increase water availability in the South Saskatchewan River Basin (SSRB), and how current work in the Oldman Basin lays the groundwork for a new approach across the SSRB.

Background

In 2006, the Government of Alberta released the Approved Water Management Plan for the SSRB. The plan was intended to “provide guidance to decision makers and act as a foundation for future water management planning of sub-basins in the SSRB.” Moreover, it “recognizes and accepts that limits for water allocations have been reached or exceeded” in three of the SSRB’s four sub-basins, and that the limit of the water resource “will be reached” in the fourth sub-basin.

In 2018, the Bow River Basin Council led a review of the implementation of the SSRB plan. The final report, prepared by Basin Advisory Committees from all four sub-basins using the most recent data available at the time, showed that approximately 62% of the SSRB’s surface water is allocated (based on mean natural annual flow and cumulative volume of active allocations). While this does not represent actual water use in the SSRB, it does reflect the maximum volume of licenced water that could be used by various sectors (e.g., municipal, irrigation, industrial), and that maximum volume represents a significant portion of the SSRB’s surface water.

In July of this year, the Auditor General of Alberta prepared an audit on whether the Department of Environment and Protected Areas has effective systems to manage water resources in Alberta. The audit noted that “Alberta could face more severe and frequent droughts.” The audit concluded that the department “lacks effective processes to manage surface water allocation and use. And public reporting on surface water and the outcomes of surface water management is lacking.”

Going forward, how can we meet the goals of Water for Life? What ‘effective processes’ are needed? And, while there is a need to develop additional reservoir capacity, how might we better leverage existing reservoirs, particularly in times of drought and given expectations that allocations in the SSRB will be more fully utilized in the future?

Modeling

Current models, e.g., SSROM and WRMM, are useful tools in managing water demand and supply. These tools, for example, can support decision making regarding water allocation transfers, water conservation holdbacks, and water sharing agreements. There are, however, limitations with these models when it comes to optimizing available reservoir storage (i.e., deciding when, where, and how much water to hold and release in the system). Moreover, the current decision-making processes associated with modeling input and output warrant review. We need, in short, robust supply- and demand-side management processes and tools; processes that are more transparent, and tools that are updated, validated, and more useful for managing storage and release volumes.

Uncertainty and Assumptions

Many variables are considered in reservoir operations. These may include, for example, prediction of runoff conditions over the next few months when managing droughts, or for the next few days when managing floods. Predictions require data, algorithms, and software. However, due to uncertainties related to required inputs (e.g., weather forecasts and related soil conditions), predictions are not 100% accurate. Hence, we can say that managing water resources on a basin level also has a stochastic component that relies on utilizing random variables that are often defined by probability distribution functions. Put simply, we can guess the likely range of incoming variables, but we cannot pinpoint their single values with certainty.

To address this uncertainty, our traditional approach in Alberta has been to explore options based on selected cases of past conditions. One or more ‘representative years’ from historical records are chosen, and computer models are then used to explore the effects of different river diversions and reservoir operations on water availability during those years. Decision makers can then discuss the results and collectively decide on the best options. This is a useful, evidence-based approach for decision making and building consensus. It also informed the recent drought responses in Alberta, where modeling assumed runoff conditions of a previous dry year (2001) with small modifications (i.e., delayed onset of runoff and a 10% runoff reduction in runoff in the Bow River Basin).

Yet, importantly, how do we know that a single historical dry year represents our current conditions? Each year is different, and it seems those assumptions may have been too conservative when irrigators were instructed to expect only 50% of their allocation in 2024; after a wet June this year, it appears more water-intensive crops might have succeeded after all.

While hindsight is 20/20, the point made here is this: assumptions used in modeling, and the decisions that are based on that modeling, can critically impact users. This past spring and summer, decisions related to water availability critically impacted many farmers. In a billion-dollar industry supplying a nutrient-deficient world, what could have been done differently?

In addition to improving modeling capabilities to optimize water availability (i.e., storage and release volumes), modeling results need to be transparent, and subjected to checks and balances – a best practice currently promoted in academic research. Analyzing model output is not trivial and requires expertise, and decisions based on this output can have serious implications for municipalities, irrigators, businesses and ecosystems. Open and timely sharing of modeling assumptions and results, and a system of checks and balances, are therefore essential to a more collaborative – and possibly more effective – approach.

The Importance of Rule Curves

Reservoirs are built to help combat droughts and floods. For droughts, excess runoff during wet seasons is stored and released later in the year when natural runoff is insufficient to meet the downstream water requirements.

The first attempts to simulate the operation of reservoirs using a computer involved the application of the so-called “Standard Operating Policy (SOP),” which releases water in dry seasons as needed to help meet downstream water demands and address adequate river maintenance flow requirements.

The thin blue line in Figure 1(a) shows the results for three consecutive years simulated using the SOP. Note that this “storage drawdown curve” has a similar shape in years when storage was either full or half-full at the start of the dry season.

Notably, it is much easier to manage storage that is full at the start of a dry season. When it is not full, as shown in the second simulated year in Figure 1(a), the storage can run out prematurely, as shown in Figure 1(b), where water demands for irrigation (dashed line) and the supplied amount (solid line) show 100% deficits (in other words, zero water supply) over more than 10 consecutive weeks later in the season. A crop failure would result under such conditions.

The concept of the rule curve was introduced to prevent this from happening. The rule curve is defined as a target trajectory of storage levels that should not be violated, shown as the black dashed line in Figure 1(a) that distributes the available storage in the dry year evenly over the entire irrigation season. In practice, its shape is defined by the starting storage at the beginning of the irrigation season, so with more storage, the red dashed line may be more suitable instead of than the black line.

This traditional approach does not respond well under highly variable water supply conditions. In India, for example, the co-authors are solving the problem with advanced modeling that accommodates both extreme flood conditions and prepares the system for not one but two growing seasons. This is accomplished by modifying the rule curves depending upon the severity of the drought. In addition, the response of the river system to water releases is more closely modeled to mimic the dynamic response of the river to the influx of more water, particularly within the shallow river systems seen on the prairies.

A New Approach

Optimal Solutions Ltd. and the University of Alberta are currently working on optimizing reservoir operations in the Oldman River Basin. The project is jointly funded by Alberta Innovates, and the ministries of Environment & Protected Areas and Irrigation & Agriculture, and is scheduled to end in the spring of 2025. The research uses a mathematical optimization approach (rather than the simulation approach currently used) to find the best way to operate reservoirs for a range of hydrological conditions. Previous published results promise significant improvements over the simulation approach used in the current model.

The difference between the two approaches is significant. While simulation models require a user-specified rule curve as part of the model input, optimization models produce the best operating rule for each simulated year as part of the model output. Essentially, the simulation model requires the user to identify the operation rule while the mathematical optimization model identifies the operating rule.

We should expect the optimization algorithms used in river basin models to give us the best water allocation for any assumed hydrological conditions. Computers can quickly examine millions of combinations of possible reservoir releases and find the best solution, similar to how they identify the best chess moves.

The use of historical (known) runoff conditions as model input should produce better reservoir operations than the actual historical operation. However, this is not always the case. For example, the simulation results in Figure 2 show the Oldman Reservoir, which remained empty for more than six months in 2001. Downstream water supply was prioritized until the reservoir was completely dry. Here, the historical operation (dashed line) seems to have managed the storage much better than the model. Kudos to the dam operators, but then the logical question is “How do operators benefit from relying on model results that do not improve on the historical operations”? Optimization offers such improvement, and it is the subject of the joint Optimal Solutions and University of Alberta project previously mentioned.

The near-drought of 2024 has taught us we need to move more quickly to the application of advanced modeling using optimization algorithms across the entire SSRB. The prospect of a large irrigation project within the Special Areas combined with announced capital investments in downstream irrigation infrastructure in Alberta will increase pressure on our limited water supply.

While we continue to model scenarios based on the past to help us improve water sharing, we need to increasingly focus on the application of more powerful tools and more thoughtful, thorough, and transparent approaches to their application to optimize the management of our storage system.