The Week The South Froze: Meteorological Context

Disclaimer: the views expressed in this post are mine and mine alone and do not represent the views of the company that I work for, IBM, or its subsidiary, The Weather Company. By reading this, you the reader assume full and sole responsibility for the outcome of any action that you take on the content provided herein. You indemnify and hold me, the writer, and my employer, IBM, blameless.

Event Summary and Statistics

The southern U.S. just endured the coldest 10 day stretch in modern memory. The statistics below come from my own munging of the Integrated Surface Data downloaded from the National Center for Environmental Information. Countless temperature records were approached or broken:

  • Oklahoma City (KOKC) set a new record for the longest consecutive run of temperatures at or below 20F, edging the previous modern record set in the great 1983 freeze: an INCREDIBLE 210 hours from February 9 to February 17, 2021!
  • Houston’s Bush International Airport (KIAH), with observations going back to 1973, experienced its 7th longest run of temperatures at or below freezing (43 hours; first place remains with 1983 at a whopping 129 hours) and its 3rd longest run of temperatures at or below 20F, which occurred in the 14 hours from 7 pm Feb 15 to 9 am Feb 16.
  • Houston (KIAH) recorded its 6th coldest temperature of record, reaching 13F early in the morning of Feb 16.
  • Dallas (KDFW) experienced its 2nd longest run of temperatures both at or below freezing (140 hours) and at or below 20F (78 hours). It hit its 3rd coldest temperature (0F; ISD shows 0F but there are other reports citing -2F).
  • About 20% of U.S. stations set record low temperatures and 30% set record low MAXIMUM temperatures between Valentine’s Day and February 16:

In addition new all-time record cold temperatures were set or equaled at 80 U.S. stations during the week! Correspondingly, weekly average temperature departures for the entire region were incredible, with negative departures from the 30-year normal of at least THIRTY DEGREES Fahrenheit across a huge chunk of the central U.S., maximized in Oklahoma and northern Texas:

That’s the kind of cold that can dominate an entire seasonal outcome! Prior to this event, 90-day average temperatures were running 2-3 degrees above normal across Texas and Oklahoma. As of February 19, the 90-day average is now 1 to 2 degrees below normal! In other words, the ~10 days of 30 deg F below normal wiped out the prior 80 days of 3 deg F above. Basic math.

Accompanying the arctic cold were numerous bouts of frozen precipitation with snow of nearly a foot occurring in some areas, including Nashville. Serious ice storms hit a large swath of territory from west Texas through the Deep South. Numerous accidents resulted, including one particular horrific event in the Ft. Worth area on Feb 11.

One thing that is really worth underscoring is the incredible event duration. Look at this meteogram (time series of meteorological information) from Yukon, Oklahoma – just west of the OKC metro:

That’s nearly TWO WEEKS of temperatures below freezing!


I cannot begin to do justice to a comprehensive accounting of human and animal impacts from this event. What we can be sure of is that this will be a multi-billion dollar disaster. In my own neighborhood alone (several thousand homes) on the northwest side of the Houston metro, there were many more flooded homes during and immediately after this arctic cold outbreak than during Hurricane Harvey in 2017! This flooding occurred as pipes broke in thousands of homes, sending water cascading through attic and attached garage roofs and into homes. Even those nimble enough to shut off water to the home still suffered damage. Waiting lists for plumbers and renovation companies are miles long, and many many people are still without running water in their homes. DIY plumbing expertise is trending and there has been an epic run on PVC and fittings. Of course, businesses, schools, places of worship and residences are not generally built for this kind of cold down here; rather they’re built with endless summer heat in mind.

Property damage owing to busted pipes only scratches the surface of the damage. Far worse is the loss of life, whose tally is still being counted. There are reports of people that literally froze to death in their homes, millions of which were without power for days leaving many unable to heat their homes (and also accelerating pipe freezes). There are still countless folks in the nation’s second most populous state without WATER! I have no idea what the toll on ranching and agriculture is but it must simply be devastating. There was literally nothing that anybody around here could do to adequately protect flora and fauna from the wrath of one of the most intense and durable cold waves in history.

Much as already been written about why the electrical infrastructure broke down in ERCOT, the state’s electric grid operator. What I’ll say about the matter is that, in a nutshell, electric generators are strongly disincentivized to gird their infrastructure against this kind of event, which comes along once per generation on average. To do so would incur substantial capital costs that would have to be offset via rate increases, making those that took such actions uncompetitive in a market that only awards the right to generate power on a given hour to the lowest bidder. Perhaps government subsidies, ultimately backed by tax dollars of course, could be used to help foot the bill for such improvements. Freeze offs were by far the largest contributor to the severe imbalance in supply and demand that led to sustained power outages for millions during the heart of the cold. Roughly one third of the largely gas-fired power was offline for days as a result. The supply-demand imbalances were plainly evident in day-ahead ERCOT forecasts during the peak of the event:

I want to note that this was not about the failure of the renewables stack. There is more installed capacity for wind generation in the Lone Star State than in the next 4 states combined (almost) and about THIRTEEN TIMES the installed capacity in New York State. Yes, the turbines froze, but there are fixes for that and I think the next time around they’ll be ready for that.

There are many good articles explaining the various complexities of the situation. For example: here and here. Generator installers are going to be busy for a year servicing the fresh demand of thousands of Texans that now have little faith in the grid operator. Despite the recent surge in stock price, Generac holdings may look even more attractive to investors!

Anatomy of historic southern U.S. cold air outbreaks:

It is instructive to illustrate the kind of meteorological pattern that leads to the biggest cold air outbreaks for the south-central U.S. Warning: this section may get a little into the meteorological weeds so if you’re not up for a meteorology lecture, you won’t hurt my feelings if you skip ahead!

Let’s examine the pattern at two atmospheric “slices” for a few of the big cold air outbreaks that have impacted the south and the Lone Star State in particular. These two atmospheric levels will be in the middle of the troposphere about 18,000 feet (roughly 5600 meters) above our heads, and also mean sea level (near Earth’s surface) so that we can visualize the spatial relationships between large north-south displacements of the former and the position and evolution of intense surface high pressure systems (anticyclones) near Earth’s surface.

The big chills of December 1983 (left column), December 1989 (middle column) and February 2021 (right column). Top row: 500 millibar heights (like the topography of the atmospheric mass field). Bottom row: mean sea level pressure. The latter is displayed near or just prior to record-breaking cold entering the southern Plains and the former is the ‘set up’ pattern aloft a couple of days prior.

One thing that this “synoptic” comparison with the other two big chill events in the modern era (December 1983/1989) reveals are some common factors: just prior to the cold air outbreak, a large amplitude upper-level ridge of high pressure is almost always situated over far northwestern North America and/or Alaska (top row; indicated by the black “H” symbols on the maps above) with a downstream low pressure trough aloft. Winds through a deep layer of the atmosphere tend to blow from a northerly direction (white arrows). At the surface (bottom row), intense surface anticyclones (high pressure systems, around which winds blow anticyclonically in the Northern Hemisphere) build over the cold source region of far northwestern North America and are then driven southward. The reason why these surface features are high pressure systems (also shown with black “H” symbols in bottom row) is that the arctic cold air masses are very dense and thus the weight of the atmospheric column is heavier above them, exerting higher pressures on Earth’s surface. I have plotted the recently expired 2021 Arctic chill in the right column. You can see that this shares the common features shown in the other two events (and many others that I could show).

Because these patterns take time to establish, there is almost always several days’ lead time in predicting them. In fact, these Arctic outbreaks are often the culmination of a persistent pattern called ‘blocking’, a situation where transient upper-level disturbances are re-directed around the periphery of a stationary upper-level high pressure ridge. Think of a large boulder in stream that re-directs the water around its mass. That’s actually not what’s happening physically with these atmospheric blocks, but the effect is the same. These same disturbances actually continually re-energize the block by pulling momentum equatorward (southward in our hemisphere; sometimes this happens in reverse in certain kinds of blocking patterns). Because the pattern AROUND the blocking high pressure cell reinforces it, these patterns become durable. And because they are persistent, they tend to be especially predictable until they break down. That’s one reason why the 2021 Arctic cold blast was fairly predictable pretty far in advance. We’ll explore some of the issues related to event predictability below.

Early Indications of winter’s attack:

The Stratosphere

This full magnitude of the event was predictable over a week in advance. In fact, the potential for a vicious cold outbreak was potentially signaled over a month earlier. The meteorological world was watching a large-scale disturbance develop over the stratosphere in early January. This is a well-known phenomenon caused by wave energy that evolves upward from the troposphere (the lowest slab of Earth’s atmosphere where we live) for the largest atmospheric “Rossby” waves (eponymously named for the great Carl Rossby). As the largest of these waves propagate upward, they decelerate the westerly winds that flow around the periphery of the stratospheric polar vortex and simultaneous warm temperatures very significantly in its core, leading to its disruption. Sparing the details, this breakdown of the westerly wind flow at high altitudes is accompanied by an expansion of these winds, including the jet stream itself, into lower latitudes. The aforementioned warming at the core of the stratospheric polar vortex is why these events are called “Sudden Stratospheric Warming” episodes (SSW for short). It turns out that the occurrence of SSW’s can sometimes set the stage for significant and durable incursions of arctic air into the lower latitudes. Of course, this is not true of all SSWs and I am waiting to see if somebody can establish causal linkages connecting this year’s SSW to the recent impressive cold air outbreak. Nonetheless, there is a statistically elevated risk for severe cold outbreaks within 2-6 weeks of an SSW. Here was how this year’s event looked in a time-vertical cross section:

A time-vertical section of temperatures averaged along latitude circles and between 60 and 90 degrees of latitude north. Time is on the x-axis, running from left to right and atmospheric pressure on y-axis. Since pressure decreases upward, altitude increases upward as well on this plot. The top of the plot extends to the top of the stratosphere and the bottom near the mean sea level on Earth.

You can see the warm temperatures (appropriately shaded with the warmer colors) developing high up in the stratosphere in early January and descending towards Earth until late month. Some meteorologists keyed in on this as a forecast factor more than a month prior. Once such a disturbance begins descending, probability for intense mid-latitude cold outbreaks increase within weeks and catching one becomes a matter of daily monitoring of analyses and predictions. One issue for North America forecasting is that one cannot be sure that the cold air, which almost always builds up over Siberia, will find its way over to our hemisphere at any point. That DID happen this year.

Subseasonal Predictions:

It is noteworthy that even simple statistical models were predicting this over 3 weeks in advance. One such model that had success is called “Linear Inverse Modeling”, or LIM for short. The LIM makes predictions by essentially asking what kind of future temperature (or precipitation or any other predictand of interest) outcomes are most strongly correlated with slowly varying signals in the oceans and atmosphere. Here is Sam Lillo’s LIM prediction from January 26, made a solid 3-4 weeks prior to the event:


Here was the average of a number of numerical weather prediction models, initialized on 1/21 or about a MONTH prior to the cold snap:

So you can see that even weeks in advance, there were indications of what was to come. Of course, the problem is that predictions at this time scale are not generally consistently skillful and thus are not particularly trustworthy and few stakeholders will take action on them. One area of research that has become very active in recent years is the examination of ‘forecasts of opportunity’: extreme outcomes (e.g., tornado outbreaks, active hurricane periods, arctic outbreaks, etc…) that sometimes (but not often) have predictable long-range precursors. The better we get at figuring out when the signals for these events can be trusted – and when they cannot – the more confidence we will have in forecasts like the above.

Medium-to-short range predictions:

By 7-10 days prior to the event, it was possible to see that a truly extraordinary event was in store for many in the southern states. This was also the time range when we could anticipate some of the regional details of the atmospheric pattern that was likely to unfold. There are many nuances to the event lead-up that I could totally geek out on, but I don’t want to make eyes glaze over so I’ll try for an overview. Here is the progression of 500 mb heights (this time shown as anomalies – or departures from a 30-year climatology of the same) in the top row and mean sea level pressure and ‘thickness’ in the bottom row. So this nearly the same display as I provided above in the context of historical cold outbreaks, but for two additional differences (besides the fact that I’m displaying 500 mb height anomalies here): (1) the maps I’m showing below are the actual forecast for the 2021 event (from the GFS) and (2) an extra field, thickness, is displayed below. Thickness is literally just that: the vertical height difference between two atmospheric pressure surfaces, in this case the 500 mb surface aloft and the 1000 mb surface near Earth’s surface. Thickness is proportional to the temperature of the layer: low thicknesses = cold and high thickness values = warm. Thus, the closer together the two pressure surfaces are, the colder the average temperature of the air column between them.

Top row: 500 mb height anomalies. Warm colors = high pressure and cold colors = low pressure. I’ve marked the approximate centers of the Alaskan upper-level high pressure and downstream Canadian low pressure at 500 mb on the top left panel. Bottom row: surface mean sea level pressure (black contours, every 2 millibars; color shading = 1000-500 mb thickness; cold colors = lower thickness values and cold air and warmer colors = higher thicknesses and warm air). The left plot is the actual weather pattern as of the evening of Feb 7, the middle panels are the forecast for 72 hours (3 days) into the future and right column the forecast for 144 hours (6 days) into the future, or near the peak of the event for Oklahoma.

The forecast progression reveals the reason why forecasters were issuing strongly worded warnings about the impact of the cold. Some things to note from the above:

  • The arctic cold air mass (the whites and purply colors in the bottom row) was slow moving and very intense, with low thicknesses literally “off the scale” for days (the white colors in the bottom row).
  • The upper level pattern (again, top row) associated with the cold air dome featured a very strong low pressure system aloft over Canada. This vortex sat in place for many days, facilitating the maintenance/build up of the extremely cold air mass over northwestern North America. This pattern was not forecast to – and in reality did not – change much during the first half of the month, leading to a prediction for an unusually durable event.
  • Once locales were behind the Arctic front, they stayed in the cold air for a long time. The bottom row shows how slowly the surface anticyclone and associated very cold air evolved. Refer back to the beginning of this post to review some of the statistics.

The foregoing provided a fair amount of forecast confidence in the severity and durability of the event and meteorologists amplified their messaging accordingly. Here are some excerpts from my own Facebook feed, where I progressively strengthen my wording as the event draws closer:

Feb 3: “The weather pattern is poised to facilitate the build up and transport of severe arctic air into the CONUS by later next week. The details are not predictable yet but the intensity of the arctic anticyclone and amplitude of the jet stream are likely to favor a very cold and wintery outcome for much of the CONUS east of the Rockies. This risk includes the Gulf Coast and Texas where setups like this can certainly produce unusually severe ice and snow accumulations. Best case our area will see a few days of very cold temps. Still monitoring but the risks warrant some attention, even 7-10 days in advance.

Feb 9: “The possibilities mentioned over the past couple of weeks are materializing: a real blast of winter is coming, rivaling anything we’ve seen in several years in Texas and in the greater Houston metro.”

Feb 10: “*** Near record-breaking arctic cold increasingly likely for Texas just after Valentine’s Day *** …preparations should be made to protect exposed infrastructure/plants/pools, keep pets inside and plan for the inability to travel from midday Monday through Tuesday.

As the event approached, temperature forecasts from a large suite of available model guidance were very impressive:

While these temperatures, predicted by the GFS about 3 days prior to the coldest day of the event – February 16 – were a bit too cold, they didn’t verify too far off and underscored the threat. In fact, these kinds of forecasts prompted forecasters to really amp up the wording to ensure that followers were prepared. Here is an excerpt from my own Facebook feed, part of a Feb 12 post:

Conservatively I’d expect temperatures to bottom out not far from 10F at places like Spring, Tomball and The Woodlands! This will be more than cold enough to solidify any ice on surfaces, destroy exposed and poorly insulated pipes and kill pets. So preparation is absolutely essential.

Snow and ice:

One of the things that made this event remarkable was that there was an active storm track across the southern U.S. This set the stage for truly impressive snow and ice accumulations. The latter also helped perpetuate the event:

Snow pack to the Gulf Coast is certainly not something you see very often! In fact, for a “last hurrah” the system dumped several inches of snow in southwest Texas and adjacent Mexico!

Here’s a picture out my front door. Although I only measured about an inch, that’s still a rare occurrence in these parts!

This post is now long enough. There are additional angles that I intended to cover that I don’t have time or inclination to at this time. This includes the climate change context (spoiler: winter can still happen under global warming; keep the difference between climate and weather in mind), a detailed weather-climate analysis and a more in-depth examination of the human impact side of the story. I’ll conclude by empathizing with the many people that have suffered – and continue to suffer – as a result of poor planning and a lack of readiness by the organizations that they depend on to keep the lights on. I am glad that my own little humble efforts to give my followers as much actionable advanced notice as possible helped them prepare as best they could.

14 thoughts on “The Week The South Froze: Meteorological Context

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s