From InkStain (John Fleck):
The conventional calculation of Colorado River shortage risk, which people like me frequently report, shows a 51 percent chance of Lake Mead dropping into “shortage”, below the magic trip line of elevation 1,075 at which mandatory cutbacks kick in, in 2020. But a new approach to modeling risk, which lots of folks (*cough* me *cough*) think more accurately represents the changing climate, shows a significant risk of a much quicker drop in Lake Mead’s levels, blowing quickly past 1,075, with a greater than 50 percent chance of dropping below 1,050 sometime in 2020. Absent actions to reduce water use, Lake Powell has a greater than one in four chance of dropping near power pool (the level at which it could no longer generate electricity) by the mid-2020s.
It is to the Bureau of Reclamation’s credit that they’re not only running the new modeling methodology in parallel with the more traditional approach, but that they’re doing this in a very public way, presenting both last week at the Basin States principals meeting in Santa Fe as part of the federal effort to get negotiations over water use cutbacks back on track.
Here’s the Mead graph showing the traditional modeling approach (the blue) and the newer “stress test” (shown in red):
I like this graph (kudos to Carly Jerla and the other wizards at the Bureau for coming up with such a clear visualization) because of how intuitively it shows how different the answers are when you use the two different approaches. Especially visually striking is the probability “cloud” showing the statistical range of possibilities. The old blue approach shows a decent chance of wetness refilling reservoirs. The newer red approach not only offers no such blue bits of optimism, but shows some statistically credible chances of Lake Mead seriously tanking very soon.
The difference between the two modeling approaches gets to the heart of the importance of the “period of record” you use to analyze river flows in support of decision-making.
The traditional (blue) approach uses historic river flows from 1906 to 2015 as the basis for a statistical simulation of the range of possible futures given current reservoir levels. This is classic “stationarity” – assume the future will be like the past. The specific model then uses what’s called the “index-sequential method” (ISM), which is a bog standard way of modeling the statistics of river flows.
The new approach (red), which has come to be known as the “stress test”, is based on the view of some water managers and scientists that the old full-record approach is understating risk, because the past climate – especially in the case of the Colorado River that exceptionally wet first quarter century of the record – is putting more water in the river in your simulations than we can realistically expect any more because of a warming climate. The stress test still uses the ISM, but instead of the full record, it uses 1988-2015, which seems more like a modern climate than the full thing back to 1906.