Mid-February 2021 is a time that many in the central USA won’t forget in a hurry. They experienced an Arctic air mass of a cold that some parts of the country had never reliably recorded before.
Accompanying snow and ice led to widespread power outages, affecting millions simultaneously in areas such as Houston, for example. Unable to heat their homes, people without alternative sources of warmth struggled to withstand the freeze.
Tragically, the exceptionally low temperatures proved lethal for some. At least 70 fatalities have been attributed to this spell of weather as of 25th Feb 2021.
Against a background of overall climate warming, this extreme cold has come as a surprise for many. It demonstrates the important difference between weather (short-term variations on a daily or weekly basis) and climate (multi-decadal trends). There may be less cold air coverage overall, but under the right weather patterns, sections of it can still become as cold as we’ve ever seen.
Widespread Snow… in Texas!
The cold air was not only unusually intense, but it also reached incredibly far south. Texas was struck by snow and ice storms bringing widespread accumulations. Such weather is so rare there that the state holds few resources (e.g. snowploughs) for dealing with such events.
With little to keep the transport network running, there was widespread disruption for an extended period. Combined with the power outages, there was a severe restriction to business activities.
These were among the most substantial impacts of those experienced across a vast stretch of the USA. You can get a sense of this scale from satellite analysis of snow cover on 15th Feb:
Only very occasionally will you ever see much in the way of white shading across lowland Texas!
Could We Have Seen It Coming Months in Advance?
Believe it or not, the answer is ‘yes, to some extent’!
While such an extreme weather event could never be predicted outright, anticipating an increased risk of some very cold days occurring is certainly possible.
At MetSwift, we do this by identifying a set of ‘most representative historical years’ (analogue years) and comparing conditions during those with a climatological baseline (e.g. all years 1981-2010). What makes a most representative year depends on the behaviour of numerous ‘key climate drivers’ (teleconnections) and how they compare with what’s been observed in the runup to the forecast period. The more similar that behaviour, the more likely a historical year is to be selected.
In the maps below, you can see how the average number of very cold days (at least 1 sigma* below average) in February compares between MetSwift analogues and climatology, for analogue issues made in early Dec 2020 (left) and early Jan 2021 (right).
* A measure of standard deviation, explained more in my mid-Dec blog .
A temperature that’s one sigma (σ) below the mean (average) is in the lowest 15.9% of historical observations.
For example, this equates to daily mean temperature of about 7°C at Brenham Municipal Airport, Texas, over 5°C colder than normal! The overnight low on such days tends to be near freezing – cold enough for snow.
Click on an image to view in full resolution.
The key thing to observe here is the signal for more very cold days than usual in the central USA.
In the early Dec 2020 issue, this was relatively slight but still enough to catch the eye when compared to the eastern USA. The implied risk of a disruptive spell of cold weather is generally near or above normal in central and western parts. Leaning the right way, 2 months in advance.
New observations then led to a strengthened signal, in the early Jan issue, for more very cold days than usual in these regions. The map shows that in the average of the MetSwift analogue years, a wide area has 1 or 2 more days at least 1 sigma below average compared to the average of 1981-2010.
With this information, risk assessment identifies an increased risk of a disruptive spell of cold weather in many states, including Texas.
So, it is possible to see an increased likelihood of such severe weather events coming at least a month in advance, using the MetSwift analogues.
In this instance the signals strengthened as the time approached and the conditions became very unusual, extreme in many locations. At 1-2 month’s range, indications will never be as striking but nevertheless, with a combination of careful data selection and predictive techniques into mid-long range weather drivers it is possible to identify key trends amongst the noise.
As of writing this, winter’s still hanging on in the northwest of the USA, but it looks like things will warm up by early March. In fact, widespread warmer than average weather is suggested by most forecast models. Central parts may see some of the most anomalous warmth – a stark contrast to mid-February!
The MetSwift analogues advise caution, however. They predict a colder than usual March in western Canada, which is where the mid-February conditions originated.
James Peacock MSc
Head Meteorologist at MetSwift
Cover Photo by Unknown Author is licensed under CC BY-SA