Arctic Heating

News stories about the Arctic always seem to say either that the Arctic is warming twice as fast as the global average, or that it’s warming nearly twice as fast as the global average. That’s not correct.

Arctic warming is more like three to four times as fast as global warming.

Some global temperature estimates (e.g. that from HadCRU) simply omit much of the Arctic, which makes it difficult to deduce how the Arctic has been warming. Others (specifically, NASA data, the extended HadCRU data from Cowtan & Way, and the data from Berkeley Earth) deduce pan-Arctic temperature from nearby measurements. This is based on the fact (not supposition but fact) that temperature change is strongly correlated in nearby areas, the correlation extending as far as 1200 km.

Let’s take a look at the data from NASA, and from Cowtan & Way, for the region from latitude 64°N to the pole, which we’ll call the Arctic. We’ll also look at their global data, in order to compare the Arctic warming rate to the global warming rate. In order to treat the two data sets on an equal footing, I’ll restrict attention to the time period from 1880 through 2017, when both data sets have coverage.

Here’s how Arctic warming (in red) compares to global warming (in black) according to the NASA data:

The greater warming of the Arctic is obvious. While the globe as a whole has warmed about 1.1°C (2°F), Arctic temperature has gone up 3.2°C (5.8°F). That’s 2.9 times as much.

To compare the present rates, I’ll use just the data since 1985 because there are signs (especially in the data from Cowtan & Way) that the modern rate for the Arctic starts a bit later than it does for the planet (which begins its modern rate around 1975). That gives me this:

The warming rate since 1985 in the Arctic, at 6.48 °C/century (11.7 °F/century), is fully 3.4 times as fast as the global rate since 1985, 1.90 °C/century (3.4 °F/century).

What about the data from Cowtan & Way?

According to these data, while the globe has warmed 1.1°C the Arctic has gone up by 3.5°C, a total rise 3.2 times as high. Using data since 1985 to estimate the modern rate we get this:

The warming rate since 1985 in the Arctic, at 8.25 °C/century (14.9 °F/century), is fully 4 times as fast as the global rate since 1985, 2.06 °C/century (3.7 °F/century).

It’s interesting to examine the warming patterns for different seasons of the year. Here’s the full time span for NASA data (1880 through 2017):

Winter is warming fastest in the Arctic, while during summer the Arctic is warming only a little faster than the globe as a whole, and much more slowly than during other seasons. This is likely due to the fact that there is still ice in much of the Arctic during summer, so temperatures have a hard time getting much above 0°C (32°F); added heat, rather than raising temperature, is consumed by the melting of ice.

Looking at data since 1985 we have this:

Again its clear that the Artic, while still warming faster than the globe, is doing so much more slowly during summer than other seasons. While winter Arctic warming is 5.8 times as fast as the global rate, summer Arctic warming is only 1.4 times as fast.

For the Cowtan & Way data since 1880 we have this:

Again summer is warming considerably more slowly than other seasons, but not nearly as much so as according to the NASA data. The rates since 1985 are these:

These data say that winter Arctic warming is 6.2 times as fast as winter global warming, while summer Arctic warming is only (!) 2.3 times as fast.

The faster Arctic rates from the Cowtan & Way data rather than the NASA data are due to the different ways they interpolate to cover the Arctic. I have more confidence in the Cowtan & Way data because they use Kriging to do so; it’s an ingenious interpolation method which is far superior to others. Still, both data sets have their advantages and disadvantages; I’d say it’s premature to say that one is definitely preferable to the other.

The bottom line is that any way you look at it, the Arctic is warming faster than the globe as a whole, and saying it’s “twice as fast” is quite an understatement.

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


51 responses to “Arctic Heating

  1. Some recent “anecdotal” evidence for the Arctic heating you refer to.

    The research icebreaker Polarstern spent quite some time last week in the wide open waters off Kap Morris Jesup, the most northerly point of Greenland:

    If I can manage to embed an image:

    And in conclusion, Hi Geoff!

  2. Wow, Thank You Tamino. I can’t understand why nobody did this analysis before. It will give new Death Spirals I guess.
    If you have interest and time… why do “global ice” graphs always “add” North and South, without moving one of them by 6 months? Right now they rather average a maximum and a minimum… Adjusting a set of data by 6 months would allow to compare or to add similar periods, ie Northern September and Southern March, and might prove interesting.

    • David B. Benson

      Fredt34 — It is a mistake to in any way combine data for the Arctic and the Antarctic.
      The geography is “opposite” for the two polar regions and the heat flows are entirely different.

      • David B. Benson,
        We need to combine Arctic and Antarctic sea ice data to observe global sea ice trends (especially relevant WRT Earth’s albedo). It’s as simple as that.

        Not unlike combining Northern and Southern Hemisphere temperature data, even though the geography of the two are very different.

      • If we’re trying to figure albedo effects, then we definitely need to adjust for seasonality, as Fred suggests. (Or possibly in some more sophisticated fashion that combines all relevant factors.)

        For instance, AFAIK, the expansion of Antarctic sea ice observed over most of the last couple of decades was mostly observed over winter, while in summer the ice mostly melted back to the continental margins. Which suggests that said expansion may not have had much albedo effect, since much of the ‘delta’ in extent occurred during the polar night.

  3. Sorry if I confused you Geoff. I’m back to my more usual appearance!

    As luck would have it The Indy recently published this:

    You’ve guessed it I expect?

    “The Arctic is warming at twice the rate of the global average”!

    However the article also mentions that:

    “A warm region of water trapped deep below the surface of the Arctic seas north of Canada has the potential to leave the entire area devoid of ice.

    Scientists have discovered warmer water that originated hundreds of miles away has penetrated deep below the ice pack’s surface.”

    • Doc,
      Great points regarding albedo.

    • The paper discussed in the Independent article is here, Timmermans et al (2018). I picked up a quote from it somewhere which talked of the doubling of heat content, all sounding a little too dramatic for my liking.
      The paper does indeed describe the heat content as having “nearly doubled” but this is the increase with the zero-point being water-at-freezing, so such ‘dramatic’ doublings should not be so unexpected for a freezy part of the world.
      In £sd, the extra warmth amounts to +0.16Zj, enough to melt 530 cu km of ice which is rather a significant amount, although the drama is reduced a little as this heat accumulatied over two decades.
      In the context of the whole Arctic, PIOMAS show annual average Arctic Sea Ice Volume has halved over the 40-year satellite record, with the September minimum losing 75%. The lost volume over the four decades is some 12,000 cu km, a loss seen over all months.

  4. Timothy (likes zebras)

    My understanding was that some Arctic researchers use different season definitions for summer, etc, than the meteorological standard because of the greater thermal inertia of the Arctic.

    Standard winter = December, January, February
    Arctic winter = January, February, March

    Which are you using here?

    Whichever you have used is probably the right one to use because I’d guess shifting the season one month in either direction would reduce the difference that you’ve found between summer and the other seasons, so I’m just being curious.

    [Response: I used Dec-Jan-Feb for winter, Jun-Jul-Aug for summer.]

  5. David B. Benson

    You ain’t seen nothin’ yet.

    ‘Archived’ heat has reached deep into the Arctic interior, researchers say
    2018 Aug 29

  6. Good post.
    One detail, however, I don’t think that C&W shows a faster Arctic warming rate than Gistemp because kriging is more sophisticated than the Gistemp infill.
    The main reason is the Arctic cooling bias in adjusted GHCNv3. Arctic stations are warming so rapidly, and are relatively few, so the PHA-algorithm finds them suspicious and adjusts them down.
    Kevin Cowtan has a piece on the Arctic bias:

    Click to access update.140404.pdf

    The remedy for this bias is GHCNv4, bringing more Arctic stations, which “convinces” the PHA that the Arctic warming is real. As I remember, Kevin C made a quick check of this when GHCN v4 came out as beta, and the arctic cooling bias was gone..

  7. So the summer, when it’s hotter has less heat increase than the winter when it’s colder?

    • angech – Yes. The Arctic in summer is like the ice in your (assumed!) Martini. The temperature doesn’t rise until the ice has all melted.

      The high Arctic hasn’t reached that point yet. In fact, since sea ice on the surface acts as an insulator, lower temperatures in summer indicate MORE open water, not less.

      • “:..since sea ice on the surface acts as an insulator, lower temperatures in summer indicate MORE open water, not less.”

        Can’t be true as a blanket statement, since clearly that depends on water temperature. And in the Arctic summer, that can vary quite a bit. Near the pack edge, it’s safe to assume that the water is near freezing, but farther away–but still well within normal winter/spring extents–it may be as much as 10 C warmer, or more:

        Care to elaborate a bit? My assumption would be that the more open water, the more of it would be likely to be well above freezing. And certainly the less the ice extent, the less it acts to sink atmospheric warmth via phase change. But all this is OTTMH; I’m not claiming any expertise here.

      • Hi Doc,

        When I spoke of the “high Arctic” I was attempting to specifically exclude the marginal ice zone. These days even “North of 80 degrees” includes a fair bit of that!

        Up in the areas away from the ice edge I think you’ll find that the pockets of open water won’t get above zero, whereas the air over sea ice can get a lot hotter than that.

        See this comment by Axel Schweiger when my alter ego made the point on Twitter:

      • Thanks for clarifying, Jim. So in this definitional frame, the High Arctic is shrinking over time, I guess!

        I’m still a bit bothered, though, by the notion that the heat exchange is so much better with the water than with the ice. Can you point to some more information on that? What you’re saying is contrary to Tamino’s posited mechanism, and with some quick sanity checking I don’t really see the pattern that would result. Rather, the lowest temperature anomalies, by and large, seem to be over the ice:

        Admittedly, the resolutions aren’t great on any of those plots, but still…

      • Your DMI “imagecontainer” link is broken Doc.

        Have you ever looked at “measured” data from buoys rather than “modelled” data from satellites, and preferably in “midsummer”? If not here’s some suggestions for you:

      • Not really broken, Jim, you just have to click again on the link (lower on the ‘error page’) to reload today’s image. But no matter, really.

        Thanks for the links. May not have a chance soon to check them out, but hopefully I’ll be able to get back to them later. It’s an interesting question.

      • Yeah but that effect works both ways – latent heat coming back OUT when it freezes back up going into winter.

      • Actually, ice decreases the humidity of the air above the ocean and since water vapor is the primary greenhouse gas on Earth (less humidity = less insulation – just ask a desert, which is what an ice-covered ocean resembles), ice actually acts as a thermal radiator! Kinda strange how physics often acts in the opposite direction as one would think, eh?

      • Richard,

        Assuming this works, please see this animation:

        That’s the temperature profile of an ice floe during the Arctic summer.

        On the left is air. On the right is sea water. In the middle is sea ice.

        Please explain the physics to me.

  8. If the amount of variation in the arctic to the global is at least 5 times larger, eyeballing the shifts in the graphs does this make the trend 5 times larger?
    Silly question I guess.

  9. Interesting analysis. I wonder if the more rapid warming in the Arctic winter vs. Arctic summer could be, at least in part, mechanistically related to the observations elsewhere (e.g., that daily low temps (typically night-time temps) are warming faster than daily high temps (typically daytime temps). Winter in the Arctic can be a like very long night.

  10. DMI have shown similar plots for their 80+N temperature.

    • While the addition of that page you link-to showing seasonal temperatures (Winter/Spring/Summer/Autumn) for each year, the real drama, certainly over the last two decades is best captured by looking at average temperatures through the year for each decade, as per this plot here (usually 2 clicks to ‘download your attachment’)

    • Strangely, the DMI flat (or slightly negative) trend for summer 80-90 N can’t be reproduced with any reanalysis.
      I have tried NCEP/NCAR, ERA-interim, JRA-55, MERRA2 and they all have a slightly positive trend for summer in the polar region (80-90 N).
      Here is one example, Era-interim 2 m SAT (via KNMI climate explorer):

      • The DMI numbers are heavily weighted towards the Pole. From the DMI docs:

        However, since the model is gridded in a regular 0.5 degree grid, the mean temperature values are strongly biased towards the temperature in the most northern part of the Arctic! Therefore, do NOT use this measure as an actual physical mean temperature of the arctic. The ‘plus 80 North mean temperature’ graphs can be used for comparing one year to an other.

      • +1 to what Jim said–I think it will show up below this, if I’m remembering correctly how this site threads things. (If not, it’ll be above, and you’ll probably already be chuckling at me!)

        However, there’s actually more on this topic, as:

        1) The DMI data we’re talking about is also reanalysis-based–actually, not even quite that, it’s based on the operational analysis (ie., the numerically-modeled everyday ‘weather map’)–and
        2) The caution they provide is quite clear about drawing climatic conclusions, such as ‘no warming in summer’:

        As described in the data information sheet, the +80N mean temperature index is not a climate data record. Since 2002, the daily mean temperatures are calculated from the operational atmosphere model at ECMWF, and changes in the operational model over time may affect the resulting temperature trends. The effect of this should be considered before making firm conclusions on basis of trends in the +80N climate indices.

        This is the same page that Phil had linked, but for convenience:

        For generalized statements about ‘the Arctic’, much better to use Tamino’s data analyses.

        If we want to talk just about the extreme high Arctic–essentially, the central Arctic Ocean–then there’s not much alternative to reanalysis data, as there are otherwise only satellite retrievals, which I think have their own polar issues, and sporadic short-term in situ measurements from ships or drift stations. (But note Olof’s cross-checks!)

        If we want to talk about sea ice albedo, the ‘lower Arctic’–call it ~67-80N–is much more important because that’s where most of the change in ice conditions is concentrated, and because it’s a considerably bigger area to start with.

      • Why on earth are they not area-weighting the gridcells?? It feels like they are making a less useful product on purpose..
        However, the poleward bias is not the whole explanation. If I pick out 89-90 N only, the graph looks quite similar.
        There’s obviously a cool bias after the introduction of ECMWF weather model data (and its successive versions) in 2002:

        ” As described in the data information sheet, the +80N mean temperature index is not a climate data record. Since 2002, the daily mean temperatures are calculated from the operational atmosphere model at ECMWF, and changes in the operational model over time may affect the resulting temperature trends. The effect of this should be considered before making firm conclusions on basis of trends in the +80N climate indices.”

        Its well known that weather model data must be bias corrected to be comparable with a reanalysis reference (eg CCI or Karsten Haustein’s).
        I guess that skeptics never read the disclaimer above, but they like to cite the DMI product when they claim that ” the Arctic is colder than normal”, which they commonly do in the summer..

  11. The Very Reverend Jebediah Hypotenuse
    • I came across this a few days a ago, really interesting:

      “About 14,000 years ago, the southwest United States was lush and green, home to saber-toothed cats and mammoths. Meanwhile, the Pacific Northwest was mostly grassland.

      That all changed as the last ice age was ending. Climate changes might be expected with the melt of a global freeze, but what’s surprising is how quickly climate and rainfall patterns changed. According to research published Nov. 22, the collapse of an ice sheet in what is now western Canada triggered a reorganization of the jet stream over the course of about 100 years — a blink of an eye in geological time.”

  12. This is a nice piece of work. I wonder if you could do something like it using charts that compare the record of total global warming to those of the average warming data for land and ocean surface air taken separately? The latter can best be found on James Hansen’s website at both of these two links: and
    It looks to me like land air has been rising twice as fast as ocean since about 1980, with oceans having twice the weight of land on the global average because of geographical area. At that rate, when the global average reaches 1.5C, in 20-25 years, land, now already at around1.6C, could be as high as 2.3C. (I believe the difference between the two occurs mainly because the source of deep ocean heat storage is removed directly from ocean surface absorption, but not much if anything from land surfaces. There are also probably some differences due to ice melting effects and relatively different rates of evaporation, and maybe more.) Any thoughts?

    • The rate of increase of Land surface air temperatures is actually a little more than double Ocean SST rates.
      I did think to see if the Wood for Trees graphical engine gave the rate of increase when it provided a linear regression. But all you find is the line through the data as here showing CRUTEM4 & HadSST3.. (Actually this is pretty close to double but CRUTEM is not fully global land.) The NOAA Climate-At-A-Glance site provides seperate graphs + data for Ocean & Land & using the data 1975-to-date Ocean has a linear rate of +0.13°C/decade while Land yields +0.29°C/decade which is more than double.

  13. Last year, I published an analysis in which I compared the three main satellite “temperature” data sets over the Arctic. I presented a poster paper update using the more recent versions of the satellite data at the Fall AGU Meeting in December. The poster paper, which is not peer reviewed, is available as an open source PDF file, which includes references to the earlier paper, may be found at:

    At the end of the poster paper, I showed a comparison of the change in the annual cycle over high latitude land vs. over the Arctic Ocean, using RSS data. The land only data indicated little change in the annual cycle, but over the Arctic Ocean, there was greater warming during the Winter months than during the Summer, as Tamino has also demonstrated. I also suggested that the satellite data may be under reporting the warming during the melt season, since the surface melt ponds exhibit a lower microwave emissivity than ice, thus the measured brightness temperature from orbit will be reduced, as I understand things. If my conjecture is correct, all the satellite data sets based on the MSU/AMSU will under report the AGW temperature change over the Arctic Ocean.

  14. Belatedly–I find it pretty intriguing that the ’30s’ bump is so evident in the Arctic graph. It’s prominent in the US record (a fact sometimes exploited for rhetorical purposes), but not so much in the global record. So it’s interesting to be reminded that it’s also a big feature in the Arctic record. (Where it has also been exploited rhetorically by those who wish to suggest that modern levels of sea ice melt are ‘nothing new’–a claim not supported by actual research.)

    Not sure what to make of it, frankly. But it makes me wonder what circulatory maps would reveal, if you did a climatology of those years? It could be done, I expect, by constraining models with observed met data. Did a little quick searching, and this seems on beam:

    “Recently, Tokinaga et al (2017) were able to reproduce the large-scale ETCAW [“Early Twentieth-Century Arctic Warming”] by introducing historical SST forcing in an atmosphere–ocean coupled global climate models (AOGCM), underlining the importance of SSTs in the onset and evolution of the ETCAW.”

  15. According to DMI arctic data it isn’t warming at all in the summer time – only in the winter (in the dark so NO amplification; less ice is therefore a NEGATIVE feedback)

    • You seem to have accepted my explanation above, in which case it should now be clear to you that ice albedo feedback is POSITIVE, not negative.

      • This paper is partly on point:

        “Sea ice loss has driven increased energy transfer from the ocean to the atmosphere, enhanced warming and moistening of the lower troposphere, decreased the strength of the surface temperature inversion, and increased lower-tropospheric thickness; all of these changes are most pronounced in autumn and early winter (September–December).”

        That first clause–‘increased energy transfer from the ocean to the atmosphere’–is very much in line with what I suggested in my previous comment, and suggests the negative feedback discussed. On the other hand, ‘enhanced warming and moistening of the lower troposphere’ should be a positive feedback. I’d guess that ‘decreased… strength of the surface temperature inversion’ would be weakly negative, as it would presumably tend to increase convective mixing and hence heat transport upward. About the last, I’m not even going to hazard a guess. I do guess that all of the above net out as a positive feedback, but I wouldn’t put money on it.

        Also, all of that is based on modeling, whereas I was looking for empirical observations.

        However, a 2010 paper by the same authors (Screen & Simmonds) is directly on point–albeit analysing the surface, not TOA fluxes (and using partly reanalysis data tin combination with in situ data.)

        “…previous studies have hypothesized that fall/winter Arctic warming has been enhanced by increased oceanic heat loss but have not presented quantitative evidence. Here we show increases in heat transfer from the Arctic Ocean to the overlying atmosphere during October–January, 1989–2009. The trends in surface air temperature, sea ice concentration and the surface heat fluxes display remarkable spatial correspondence.”

        More recently (2018), the topic was revisited, with the conclusion that the really important factor in warming was really the downward IR flux:

        Still nothing on TOA fluxes, though. I did find some work using satellite retrievals to study the surface balance, but it appears that the satellite data are not yet, as they put it in one study, ‘homogenous.’

    • Mike, first, I think you are misinterpreting the DMI data, which is based on reanalysis, and explicitly cautioned against for use in long-term climatic comparisons. What Tamino has done in this post is much more accurate and precise.

      Second, albedo isn’t driven by warming directly, so the DMI winter warming you cite is not directly relevant to determining the sign of ice albedo feedback. That is, ice albedo is a function of ice area and state.

      The winter warming does, however, help drive ice mass loss by slowing ice formation during the dark months, because heat flux from ocean to atmosphere depends significantly on temperature differential–a warmer atmosphere means a smaller differential, less heat loss at the base of the sea ice, and less new ice formation. So there is less ice to melt when the sun arrives in March. Or so goes the only physically consistent account of it that I’m aware of.

      Regardless of that account, though, the observed reality is that there is a whole lot less ice during the sunlit half of the year than there used to be. Hence, more sunlight absorbed, and that means more warming of the surface. Now, if we presume (as described above) that Arctic warming in whatever season indeed means less ice–and that’s kinda hard to argue with, IMO–then the lowering of albedo is reinforcing that same warming, which is a POSITIVE feedback by definition.

      Which, of course, doesn’t mean it’s a ‘net positive’ for the environment, or for us.

      NB–there is a (presumed) negative feedback, though, for autumn. As ice area decreases, more open water is exposed to the cool atmosphere, increasing heat loss from the ocean, and probably to space, not just the atmosphere.

      (Anybody have a reference on that last notion? I don’t think I’ve seen regional Arctic TOA fluxes specifically studied, but the data has got to exist for the satellite era, and someone may well have analyzed it to see if there’s a discernible signal of increasing OLR flux that correlates with ice area.)

      • Doc,
        “NB–there is a (presumed) negative feedback, though, for autumn. As ice area decreases, more open water is exposed to the cool atmosphere, increasing heat loss from the ocean, and probably to space, not just the atmosphere.”

        In theory, more open water also means more water vapor, clouds and snow (good insulator)
        …….factors that slow the rate of heat lost to space. So there’s that to consider as well.

      • Clouds are tricky things. See for example:

        “Our results indicate a potentially significant amplifying sea ice-cloud feedback, under certain meteorological conditions, that could delay the fall freeze-up and influence the variability in sea ice extent and volume”

      • Tricky indeed.
        There is an argument that says the sea-ice-loss is amplified by positive feedbacks through the summer but leads to massive negative feedbacks through the winter as the low-ice condition boosts heat loss to the atmosphere and thence to space. I’m not sure how true this is & I don’t see any literature covering the mechanisms. This difficulty arises because the warming resulting from sea-ice albedo is a feedback and not a forcing. As a feedback, it doesn’t get the same attention from the modellers.
        What can be said is that the long-term global effect of a Black-Carbon-on-snow/ice forcing is 170% (“efficacy”) of the same level of global CO2 forcing, as per Hansen et al (2005). Yet BC-on-snow is not equivalent to ice-melt albedo as GC-on-snow is not necessarily opening up water with the resulting increase in water vapour. But with efficacy boosting the impact of Arctic albedo changes, it would be a surprise to find that the impact of positive summer feedback was reduced by more than it was boosted by “efficacy”.

    • Mike if you go to the DMI page I referenced you’ll see that temperature is increasing in winter, spring and fall, only in summer does it not increase.

  16. Off-topic …

    Tamino, I would like to make a reply in comment thread I have been participating in for your article “Global Warming: How Long Do We Have Left?”, but it seems you have closed comments there? There is no comments box anymore, and there are no “reply” anymore on any of the comments. Can you fix that please?

  17. Ahh, so what you’re saying is that the areas where most people live are experiencing LESS warming than the global average. (though that’s a fishy claim because land is warming faster than ocean)

    • Richard,
      You don’t think it’s possible, given the Earth surface is ~70% water, that both the arctic AND the areas where most people live could be warming faster than the global average?

    • Richard Caldwell,
      I don’t see where or how here you are finding an argument presented that suggests “the areas where most people live are experiencing LESS warming than the global average.” Global land temperatures are increasing at roughly twice the rate of global ocean temperatures, so increasing 50% faster than the average for the full globe.
      But while this may suggest the opposite to the “LESS warming” assertion, bear in mind that while humanity is land-dwelling, as of 1994 a large minority lived not-so-far from a coast (see here) suggesting that a large minority of humanity are potentially inhabiting a climate where AGW is mitigated by ocean temperatures. It would therefore not be such a big ask to find a bit of populous continental climate without too much AGW to overturn the “most people” assertion; that assuming the proportion of coastal population has not grown above the 50% during the last two decades.