Click the link to read the article on the InkStain website (John Fleck, Anne Castle, Eric Kuhn, Jack Schmidt, Kathryn Sorensen, and Katherine Tara):
As we await Friday’s (Aug. 15, 2025) release of the Bureau of Reclamation’s Colorado River 24-Month Study, we need to remember a painful lesson of the last five years of crisis management: whatever you see in Reclamation’s report of the “Most Probable” reservoir levels for the next two years, we must prepare for things to be much worse.
A year ago, Reclamation’s “Most Probable” forecast told us to expect Lake Powell to hold 10.36 million acre feet of water at the end of July 2025, with a surface elevation 3,593 feet above sea level. Actual storage in Powell at the end of July was 7.46 maf, 2.9 million acre feet less, and the reservoir is 38 feet lower, than the “Most Probable” forecast.
Four years ago, one of us (Eric Kuhn) wrote this, which is helpful in understanding what is happening:
“The problem: the assumptions underlying the study do not fully capture the climate-change driven aridification of the Colorado River Basin.”
In 2022, a Utah State Center for Colorado River Studies team led by Jian Wang (including one of us, Schmidt) took this on in more technical detail – Evaluating the Accuracy of Reclamation’s 24-Month Study of Lake Powell Projections. The finding provided technical support for an intuition water managers already had: the 24-Month Study has an optimistic bias.
It is a practical demonstration of the problem U.S. Geological Survey scientist Paul Milly and colleagues famously warned us about nearly two decades ago – in water management, climate change means the past is increasingly unhelpful in projecting the future. [ed. Also: Stationarity Is Dead: Whither Water Management?]
The 24-Month Study: A Brief Primer
Produced monthly, Reclamation’s 24-Month Study includes three scenarios: Most Probable, Minimum Probable, and Maximum Probable. The Study includes 18 pages of data and forecasts for twelve Colorado River system reservoirs, from Fontenelle and Flaming Gorge in the north to Mohave and Havasu in the south, projecting things like elevation, storage, inflows, releases, evaporation, and hydropower production each month for the next two years.
Here is Wang et al’s explanation of how it works:
“Projections for reservoir elevations during the next few months are based on predictions of reservoir inflow using a widely accepted watershed hydrologic model run by the Colorado Basin River Forecast Center. The input data for that model are observed snowpack in the watershed, soil moisture, and anticipated precipitation and temperature. Projections for reservoir elevations beyond the immediately proximate winter, a year or more in the future (‘second year projections’), are based on statistical probabilities calculated using analyses of past inflows during a 30-year reference period.”
The resulting model runs represent a wide range of uncertainties, which are captured in three resulting scenarios:
- Most Probable: the middle of the range
- Maximum Probable: the 90th percentile scenario, meaning that 10% of the model runs predict even wetter hydrology and 90% predict drier.
- Minimum Probable: the 10th percentile scenario, meaning that 10% of the model runs predict even drier hydrology and 90% predict wetter.
The problem, implicit in the argument Milly et al. made nearly two decades ago, is that a 30-year reference period is no longer a reliable indicator of what we should expect in the future. It represents a river we no longer have. This is not to suggest any bias or partiality on the part of Reclamation, but merely that the algorithms and modeling used to produce the 24-Month Study have proven in recent years to be skewed more toward the the past than the true-to-life. Our response needs to reflect that reality.
Because of the changing conditions in the Colorado River Basin, the Minimum Probable scenario has become the most valuable in providing a reliable indicator of the future. Actual flows and reservoir levels have been tracking the minimum probable forecast since March of this year. As we enter the fall of 2025, with the weak summer monsoon for most of the Upper Basin coupled with weak La Niña conditions persisting through the fall and early winter, and NOAA’s seasonal outlook pointing to a warmer and drier than average fall, it’s a good bet that this trend will continue at least through mid-winter. The Basin should be prepared for minimum probable conditions, with a clear possibility that actual conditions could be worse than the 10th percentile scenario. The basin community needs to be ready to respond with the necessary water use reductions now to protect the Colorado River system on which we all depend.
Sources:
- August 2024 24-month studies: Most probable, Minimum probable
- July 31, 2025 reservoir levels: USBR hydro data
