Jet Stream – Sea Ice + Polar Vortex = ???

A fascinating idea has emerged, that when heat waves or cold waves happen, global warming might make them last longer. Some think that this has already begun, an idea suggested by the recent long-lasting cold wave to hit the eastern U.S. It has also been implicated in some recent extreme hot times, such as the Moscow heat wave in 2010, notable not just for its extremity but for its long duration.

The gist of the idea is this: we already know that the Arctic is warming much faster than the globe as a whole, in fact in general high latitides tend to warm faster than the tropics. Arctic temperatures are getting closer to their more southerly cousins, and this means that the temperature difference as you travel toward the pole (the gradient) is getting smaller. The temperature gradient is what drives the jet stream, which has a big influence on how weather patterns tend to move about. When the gradient is weaker the jet stream is more meandering, which can make weather systems move more slowly, perhaps even get “stuck” in place for longer times. That means that whatever the weather brings — including hot spells and cold snaps — can last longer.

The effect may be especially pronounced in the Arctic because the reduction of sea ice cover can also profoundly affect how wind patterns behave.

It sounds to me like sound theory, but the proof is in the pudding. Whether the effect is real or not remains to be seen, but I wonder, is there evidence that it has begun already?

I retrieved daily temperature data for New York City (Central Park), in order to see how long cold and hot spells last when they do occur. It seems a good choice for a first look, because it has such a long record and it is feeling the cold that has gripped the U.S. east coast.

The question itself opens a “can of worms” because now we have to decide what makes a cold or hot spell, how to quantify it. It even requires a definition of what’s “hot” and “cold,” and I chose to define them relative to what’s “normal” for the given time of year, so a temperature of 60°F (about 15.6°C) could be part of a cold spell if it happens in July, or part of a hot spell if in January. So the first order of business is to transform temperature to temperature anomaly, the difference between a given day’s temperature and the “normal” for that time of year.

That of course raises the question, “What’s normal?” I wanted to avoid getting more/fewer/longer/shorter cold/hot spells simply because what’s “normal” changed via global warming. So I fit a lowess smooth to the temperature anomalies in order that the “normal” could change to keep up with the times. That means rather than just look at temperature anomaly (which really needs an unchanging baseline), I’m looking at what I’ll call adaptive anomaly, in which the baseline follows the estimated global warming signal.

And what, you may be wondering, makes a “cold spell” or “hot spell”? The first measure I’ve tried is days when temperature is as much as 5°C hotter or colder than “normal.”

Then, each time daily temperature went into the specified range, I counted how many days that condition persisted. This gives me a series of times at which such “excursions” happened and how long they last.

Of course there are many different ways to define such conditions, but this occurred to me as a good starting point. If there is already an obvious change in the duration of excursions, it might show in such data.

Allow me to illustrate. Here’s the daily mean temperature in New York City (Central Park) for the last 149 years:

Subtracting away the average seasonal cycle gives us the usual form of temperature anomaly:

The long-term trend is clearly visible and easily established statistically. It can be estimated with a lowess smooth, which is shown as a red line in the above graph.

I can now subtract the smoothed value from the anomalies themselves to generate adaptive anomaly:

This is, I believe, a pretty good estimate of how daily temperature has fluctuated, apart from the seasonal cycle and the global warming trend. It is in the sequence of adaptive anomalies that I will search for changes in the typical duration of hot/cold times.

What did I find? Here’s the length (in days) of each excursion of daily mean temperature by 5°C or more, either hotter or colder than normal, in the 149 years of data for New York City:

The longest-duration event was a 16-day cold spell which finally came to a close on February 3rd, 1961. The next-longest has just happened, a 14-day cold spell.

Both are in the 2nd half of the data set, one is extremely recent — does that mean hot/cold spells are getting longer? A few extreme events can be suggestive but rarely demonstrate anything other than the fact that extreme events happen, but they’re rare enough that “statistical significance” eludes us.

I’ve looked for a trend in these data, and found none. The average duration of cold/hot excursions (beyond the 5°C range) hasn’t changed at all; when the “p-value” gets up to about 0.98, any claim of statistical significance is ludicrous. The number of excursions which exceed some limit, like those 3 days long or longer, or 7 days long or longer, likewise shows no significant trend.

I also looked for any change at all in the lengths of excursions, by dividing the data into two different time spans and comparing the probability distributions of excursion duration for each. I compared pre-1945 to post-1945, pre-1980 to post-1980, and pre-2000 to post-2000. Here are histograms of pre- and post-2000 durations, with the pre-2000 distribution in blue and post- in red:

It certainly seems that excursions lasting only 1 day are less frequent while those 2 days or longer are on the rise, but is that a real difference or just a result of random fluctuation? I compared the distributions with the Kolmogorov-Smirnov test, and all three tested splits gave the same answer: no, the distribution isn’t significantly different. When the p-value gets up around 0.7, claiming statistical significance is just nonsense.

What does it mean? The most important conclusion is that we need more analysis to know. A single location is a paltry sample. New York City is not the world (despite the perspective of many New Yorkers). And we already know that climate change affects different areas differently, particularly with regard to heat waves, which are definitely on the rise in Europe, almost certainly in Asia, but have not yet shown in the United States.

But it is an important result, I think, that New York City, in the midst of the east coast cold wave, shows no sign that this latest is a change in climate conditions. It’s certainly unusual, but not unheard-of, we’ve seen similarly unusual excursions before. Maybe the take-home message is the now-known expression that “Just because climate changes, that doesn’t mean we won’t still have weather.” From what I’ve looked at the recent bitter cold is just that: weather. Extreme weather, yes, but out of line with what extreme weather has been like in the past, no.

I think analysis like this deserves a more far-reaching and systematic approach. When we look at more than just one location, using more than just one analytical approach, we’re bound to learn more. And I emphasize again that this investigation, with its limited scope, is really just a first exploratory step.

So much data, so little time …

This blog is made possible by readers like you; join others by donating at My Wee Dragon.

16 responses to “Jet Stream – Sea Ice + Polar Vortex = ???

  1. You need this:
    And this: .

    As to the latter, a catalog of North Atlantic/Euro circulation patterns: on average duration of any pattern went from 3-4 to 6-8 days.
    In Western Europe, I speak for the Netherlands, weather has become outright boring. Generally spoken instead of some 20 different patterns we have had over the past year a time of anticyclone (drought in spring and summer, record hot June) followed by west, west, west, only west, and more west, and still no end in sight.

    My head is an archive of circulation patterns per about 1880. I have thrown it away. Starting this century and as of this decennium in earnest, everything has completely changed.

  2. An interesting post, thank you. It has also been suggested that a wandering jet stream causes more frequent extreme temperatures. If it does and if the jet stream is wandering more now than in the past, should the variance of the adaptive anomaly have increased? Has the variance increased?

  3. > The temperature gradient is what drives the jet stream… When the gradient is weaker the jet stream is more meandering, which can make weather systems move more slowly…

    Lots of steps there, and no citations for any of them. I don’t think these are common knowledge.

    [Response: I’m trying to give some perspective about a complicated subject which is accessible to the lay reader, so of course I don’t want to make the writing pedantic. This is a blog post, not a submission to GRL.]

  4. This remark made me wonder: “…heat waves… are definitely on the rise in Europe, almost certainly in Asia, but have not yet shown in the United States.”

    I thought they’d been shown for the US, too. The recent special report has some things to say on the topic:

    I quote Key Finding #2 (Chapter 6, US temperature change):

    “There have been marked changes in temperature extremes across the contiguous United States. The frequency of cold waves has decreased since the early 1900s, and the frequency of heat waves has increased since the mid-1960s. The Dust Bowl era of the 1930s remains the peak period for extreme heat. The number of high temperature records set in the past two decades far exceeds the number of low temperature records. (Very high confidence)”

    Is this mostly a matter of how, specifically, one characterizes ‘on the rise?’ It appears that in the US the Dust Bowl years have a big impact on the time series, and that one must therefore be very careful in handling the record, both in statistical analyses and verbal descriptions.

    [Response: The extremity of the dust bowl years is a huge part of it. I’ll also point out that “record hot days” isn’t the same as heat *waves*, which by definition last longer than a day (usually 3 days or more of high temperatures, sometimes longer). Then there’s the fact that when determining record heat you must *not* remove the global warming signal as was done here. Also, “heat wave” usually (but not always!) is defined relative to an absolute temperature threshold, so in most northern-hemisphere locations they just don’t happen in January, they’re confined to hot-weather months. But I’m looking for persistence so I’m defining it by departure from the seasonal norm.

    It emphasizes the need, not only to be clear about what you’re defining, but how to interpret it as well.

    When it comes to the U.S., I’m not sufficiently current on the literature to know the status of modern research. But I’ve looked at some of the data, and in Europe I see undeniable unambiguous evidence, in the U.S. it’s not so clear.]

    • Tamino, I’ll take that explanation as a ‘yes’ to my question regarding the need for care in specification! ;-)

      Thanks for expanding.

  5. Susan Anderson

    Lay comment here. Frances/Vavrus has been around a while, and from the northeast of the US, this seems to be a done deal, much in evidence at a real-world level for a decade or more. I appreciate the straight-up language and lack of references (which F/V have done extensively over the years, including long presentations with plenty of data at the AGU). But for a person inhabiting the real world and living in these stuck patterns (fire season in the west this year too) it feels to me like it should be a no-brainer. In casual conversations about the weather with people who are concerned and have less scientific background, meteorological knowledge, and fascination with the subject, this idea is readily understood, because we’ve all experienced it.

    The disconnect between reality is well to the fore here, as fake skeptics promote detailed information that contradicts most people’s direct experience, particularly those who are no longer young. We here in flood-recovering Boston (yes, it’s still causing problems) just had a powerful illustration in the long cold spell before hot met cold and full moon and hit us where it hurts.

    [Response: Perhaps I should have emphasized more strongly the limitations of this “first look.” That includes considering *only* persistence of temperature excursions. One of my goals was to address the specific claim that global warming had enhanced the chances of the recent U.S. cold wave.]

  6. michael sweet

    I think weather is normally more stable in summer than in winter. The standard deviation of the anomaly is greater in winter than summer. That might be relevant to your analysis since you treat the summer and winter the same. Perhaps using your definition of heat (or cold) wave you would see the effect first in winter.

    [Response: I had thought of that, and I agree it needs to be considered. That’s yet another reason this analysis is as “preliminary and exploratory” as it gets. I strongly encourage readers with skillz and interest to look into this.]

  7. Meteorologists tackle the issue of persistence by examining “blocking” charactertistics, defined by properties of the flow at the 500mb level. Hoskins and Woollings reviewed recent work on climate extremes associated with blocking and other atmospheric circulation regimes (Curr Clim Change Rep (2015) 1:115-124).

    One might conclude from this work that (1) persistent patterns are influenced by topology and coastlines (land/ocean contrast), so location matters, (2) there are seasonal effects, and (3) there are a lot of inconclusive results regarding trends, so it sure would be nice if some straightforward regional analyses along the lines of the one presented here by tamino were shown to produce robust results!

    (I’m having trouble linking to online pdfs; the usual html protocol doesn’t seem to work. I use Realclimate’s preview, where the link looks ok, but when I try it I get “page not found”. Any suggestions?)

    • I’m particularly interested in the “blocking pattern” issue as blocking patterns are implicated in several recent intense rainfall events in western North America…Montana 2011, Alberta and Colorado 2013 and also to some extent Hurricane Harvey in Houston. These precipitation event have huge implications for the mining industry, which is what I’ve focused on. Alas I’m inclined, so far, to agree with Tamino, these are relatively rare events and we just can not say with certainty there is a linkage. What this really means is that we’re doing the experiment now.

  8. Another way to look would be to look at dry/wet excursions and to look at the extremes (maybe events in top/bottom 1% 5% 10%)?

  9. Re Susan’s Francis/Vavrus and “the “sticky” jet stream, there’s lot’s more about all that at:,750

    See also the work of James Screen:

  10. Thank you Tamino, I appreciate the clear, unbiased look at the data. I was hoping it would show something, but I agree more analysis is need.

  11. you said “in fact in general high latitides tend to warm faster than the tropics”. And that was what I understood to be the case until recently, when I actually looked after being challenged. It seems to me that it’s only really the case in the northern hemisphere? ‘High latitudes in general’ would suggest that the Antarctic is warming more quickly than average, as well as the arctic but the data I have seen suggests that this is not the case. The southern ocean is one of the few places showing relative cooling. Antarctica itself seems to have a flat temp. This map doesn’t show the Antarctic continent itself very well, but I think you can see what I mean.

    Can someone who actually knows clarify this point. It is ‘high latitudes in general’, or only northern high latitiudes (presumably due to the influence of land masses)?

    [Response: As far as I know (an important qualifier), it has so far only manifested in the northern hemisphere. But also (as far as I know) we expect that in the future Antarctica will join the “club” and warm faster than the tropics. Anyone know more about this?]

    • Like Susan, I’m a layman. I’ve got nothing profound.

      But–maybe it’s worth mentioning three well-known but pretty clearly relevant things. One, warming in the Southern hemisphere generally is more muted, presumably because the proportion of land to water is much lower. Two, the Southern polar circulation is much stronger (presumably because there’s a circumglobal ‘fetch’ completely unimpeded by land at about 60 degrees South.)*

      And three, there is an exception to the muted warming of Antarctica: the Peninsula, and adjacent areas of West Antarctica have shown some rapid rates of warming. (Not sure what the current picture in that regard is; I seem to recall that in recent years there had been–dare I say it?–a regional ‘slowdown’, but I’m not sure about that.) And more generally to that point, if you look at the GISTEMP zonal timeseries plot, you can see that higher Southern lattitudes show more warming than Southern midlatitudes.

      *Reality check on that statement (ie., point #2): I’ve understood it to be the case that the southern circumpolar circulation is the stronger, but a quick peek at the climate reanalyser for today, at least, shows stronger jetstrean winds in the *Northern hemisphere*, particularly evident in a powerful northeasterly circulation above the North Atlantic. Not sure what is typically the case… and not sure about seasonal effects in that regard. Anyone with actual background on this?

  12. Thanks Tamino, interesting post..
    I have a certain feeling that the weather has become more persistent lately (mostly in the boring way though), but I am likely biased by reading the theories of Jennifer Francis, etc.

    Well, I thought that autocorrelation of a weather variable could be a useful indicator of “persistence”.
    So I tried with two stations, New York Central Park like you, and Stockholm, Sweden, across the Atlantic, using KNMI Climate Explorer. I examined the autocorrelation of daily temperature data (anomalies) in 30 year periods 100 years apart, 1988-2017 vs 1888-1917. Here is the result:

    In New York there were only small increases of the autocorrelation (1-3 days lag), over 100 years. The autocorrelation in Stockholm has increased more clearly, for lags up to seven days. Technically, I think that the change is statistically significant. The KNMI tool says 2/sqrt(N) is 0.0192 for each 30 year interval, so I guess that an increase of the correlation coefficient larger than 0.03 is significant with these large N. There might be biases though, but for instance detrending data within the 30 year periods doesn’t change the outcome very much.

    Maybe the day by day change (autocorrelation with one day lag) is the best measure of perceived weather persistence?
    A persons (my) weather experience usually doesn’t span 100-130 years years, so I also examined 2008-2017 vs 1978-1987 in Stockholm. Here, the one day lag autocorrelation coefficient increased from 0.7773 to 0.8022. I believe that these small changes are too subtle to be genuinely percieved as “more persistent weather”, so I guess I have been biased..