Colorado River Basin Decision Making Under Deep Uncertainty

aerial view of lake mead and hoover dam

Funding Agency
Bureau of Reclamation

Research Team

  • Elliot Alexander, MS student, Reclamation engineer in training
  • Joe Kasprzyk and Edie Zagona, principal investigators
  • Carly Jerla, Jim Prairie and Alan Butler, Bureau of Reclamation

Classical decision-making frameworks have aided water managers in selecting an optimal water resources plan. These methods are applicable when dealing with simple systems and well-understood uncertainties impacting the system. As water resources systems and models become more complex and the future climatic conditions are poorly known, the classical decision-making frameworks don’t hold up.

The Colorado River spans seven U.S. states and Mexico and is an important cultural, economic and natural resource for 35–40 million people. Its complex operating policy is based on the “Law of the River,” which has evolved since the Colorado River Compact in 1922. Recent (2007) refinements to address shortage reductions and coordinated operations of Lakes Powell and Mead were negotiated with stakeholders in which hundreds of scenarios were explored to identify operating guidelines that could ultimately be agreed on. The Colorado River Basin’s water delivery reliability has continued to degrade since deciding these 2007 operational guidelines due to the imbalance of growing demand and dwindling supply. The magnitude of this imbalance is challenging to predict since the most likely realization of future water demand and hydrology are unknown, thus describing conditions of deep uncertainty. Moreover, the demand-supply imbalance may stress the previously identified operational guidelines, motivating new methodology to supplement the existing planning to create new alternatives.

This research explores innovative planning approaches that are appropriate for conditions of deep uncertainty and then demonstrates an application of a method called Many Objective Robust Decision Making (MORDM). This MORDM application couples a multiobjective evolutionary algorithm (MOEA) with the Colorado River Simulation System (CRSS) model to generate and evaluate thousands of new operating policies for Lake Mead. The MOEA-generated policies are then resimulated across multiple future water supply and water use scenarios, testing each policy’s performance across a wide range of plausible future uncertainty. This analysis uncovered multiple candidate robust operating policies from the set of MOEA-generated policies. The operational similarities between the robust policies reveal that Lake Mead's current operation should consider significantly higher shortage volumes at lower pool elevations to be more robust under conditions of deep uncertainty. This research has laid the groundwork for practitioners and academics to use MOEAs and the MORDM framework to support future complex water management studies with real-world implications.

Alexander, Elliot A. (2018). Searching for a Robust Operation of Lake Mead, Civil, Environmental and Architectural Engineering MS Thesis, University of Colorado, Boulder, CO.