What caused the record-breaking cold air outbreak of February 2021?
If you don’t test it, you don’t know it.
You’ve probably heard that the February 2021 cold air outbreak that blanketed the United States in snow and led to a power crisis in Texas was caused by the polar vortex. It’s an interesting story, but what if we developed a way to test whether it’s true?
It’s the extremes that matter.
We almost always experience average weather. We expect it to be colder in winter and warmer in summer, and we expect it to rain and snow, but not too much. We build our lives and infrastructure around these expectations. It’s the extreme weather, the once-in-a-lifetime events that are difficult to forecast and prepare for, that impact our lives and cost us billions of dollars in damages. In February of 2021, the United States suffered a severe cold air outbreak that killed hundreds, broke the power grid, and disrupted the economy for weeks. Scientists and forecasters scrambled for an explanation.
Was it the polar vortex?
Researchers have hypothesized that cold air outbreaks in the United States may be caused by a “stretching” of the stratospheric polar vortex. While the physics are complicated, the idea is simple. Large-scale waves constantly travel up to the stratosphere. When the vortex becomes stretched, it reflects these waves back down toward the surface, where they can invigorate weather systems and produce extreme cold. While it’s true that the vortex was stretched in the weeks before the February 2021 cold air outbreak, how would you actually test whether the stretching and wave reflection caused it?
Use a forecast model - cleverly.
The atmosphere in a forecast model is initialized as close to observations as possible so that the forecast most represents reality. But what if you wanted to test which part of the atmosphere was most important for the future evolution of the weather? You could choose to not initialize, or to “scramble”, different parts of the atmospheric initial conditions, and see how the forecast changes. When the stratospheric initial conditions are scrambled, the forecast still predicts extreme cold. On the other hand, when the initial conditions in the troposphere, or the weather layer of the atmosphere, are scrambled, the forecast predicts extreme warmth. How is it that the stratosphere had no impact on the extreme cold, while the lowest layer of the atmosphere had full control?
Check the mechanism.
In both observations and the standard forecast, wave activity travels upward into the stratosphere over Eurasia and is reflected off of a stretched vortex and back down over North America. This process is completely absent in the forecast with scrambled stratospheric initial conditions. How is it possible that the vortex stretching and wave reflection had no impact on the extreme cold?
Examine the physics with great care.
It is true that the wave activity that originated over Eurasia was reflected downward across the Pacific - but a careful analysis of the flux streamlines shows that these waves never reached the troposphere over North America. All of the wave forcing in the troposphere over North America originated from North America. In other words, the reason that the forecast with scrambled stratospheric initial conditions - and no vortex stretching or wave reflection - still predicted the extreme cold over North America, is because the waves never reached the troposphere, anyway.
Why is this research so decisive?
Most (really, nearly all) experiments with numerical models test how the entire atmosphere or Earth system respond to an external or internal forcing. But if we want to understand the internal processes within these systems - whether the stratosphere drives the variability in the troposphere, for example - we need to shift the experimental scope from the whole system to these subsystems. This is difficult, because many of the processes of interest to us are directly resolved by the model. How do we control these independent, resolved variables so that we can align our experimental scope with our hypotheses?
For individual events, the initial condition scrambling procedure is one possible method to directly test causality. And a direct test will always be more certain, and less ambiguous, than statistical or machine learning approaches. By construction, initial condition scrambling uncouples the intertwined dynamics of the Earth system that requires a supercomputer to simulate.
If it takes a supercomputer to simulate, we might as well leverage that supercomputer to answer our questions.
Want to read more?
This research is published in Nature: Communications.
Davis, N. A., Richter, J. H., Glanville, A. A., Edwards, J., and LaJoie, E. (2022), Limited surface impacts of the January 2021 sudden stratospheric warming. Nat Commun., 13, 1136, https://doi.org/10.1038/s41467-022-28836-1.