250 thoughts on “Curry’s criticisms of Cowtan and Way”

1. So these cold and snowy Northern hemisphere winters we’ve had over the last 5 or 6 years? They never actually happened.

2. Its interesting that apparently GISTEMP+SST adjusted has a similar trend as the CW method. Hope somebody out there will make something out of Judy Curry’s brainstorming. I too am interested in finding out more about whether the cross-validation aspects that the authors emphasise is meaningful.

3. Hi Lucia,

   Regarding #4, it seems to me at first glance that simply having delta_T (T_sat – T_surf ) differ at the poles from what its krigged value predicts would not be enough to introduce a bias in the trend, as a constant offset (even if in error) would not create a trend bias. This would only matter if there was a lot of temporal interpolation going on at those high latitudes using delta_T (in which case grid cells go from having “actual values” to predicted values, so there wouldn’t be a constant offset). But my understanding is that there is generally *no* data for these gridcells, and it is primarily spatial interpolation (but I could be wrong on that).

   Rather, the bias would occur if the *trends* of delta_T were substantially different in the interpolated regions versus those cells used in the prediction of those regions. Here is my preliminary analysis of the trends in delta_T for GFDL CM2.1 (run 1) from 1996-2005, using the temperature at 850kPa as a proxy for T_sat:

   https://dl.dropboxusercontent.com/u/9160367/Climate/CW13_GFDL_deltaTTrend_11-15-2013.png

   There is evidence there of some heterogeneity there in the Arctic, but without diving further (which I doubt I will have time to do anytime soon) it is not clear how this would actually the methodology, or even if this single run of the model represents the relationship between T_sat and T_sfc in the real world.

4. With regard to the question of validation, an elementary test would be to see how well the method works with a simulated climate, no?

   After all, the claim seems to be this reduces disagreement between models and data, but the same procedure was not used on model output and observations and, as others have pointed out, masking model data to the observed domain does not have that result.

5. Lucia, like you, I appreciate the contribution made by this paper and, and I admire the amount of diligence that was applied to its preparation.

   Rather than trying to directly address your concerns, let me just raise my related concerns, which also present in the use of kriging in BEST.

   The issue brought up by Judith here:

   First, Kriging. Kriging across land/ocean/sea ice boundaries makes no physical sense. While the paper cites Rigor et al. (2000) that shows ‘some’ correlation in winter between land and sea ice temps at up to 1000 km, I would expect no correlation in other seasons.

   Relates in part to kriging across the land-marine boundary. Of course this boundary is absent in the winter time, so you’d expect these issues to disappear.

   If you look at surface meteorology, you can divide it up into these very rough categories (there are more):

   1. Marine
   2. Coastal
   3. Inland

   Marine and inland communicate with each other via costal, but the meteorologies in two domains are very different, enough so they are considered different sub-specializations. Marine and costal meteorologists are often found in marine science departments whereas land meteorologists are often located in atmospheric groups.

   Anyway, the dynamics for land versus marine are enough different that you’d expect any correlation to quickly vanish between inland points and marine points, when the spatial separation gets very large.

   More generally, this brings up the question of whether one can really treat the kriging (spatial correlation) function as cylindrically symmetric, as is usually assumed. I would suspect the assumption of cylindrical symmetry is very poor in the north Arctic in the summertime.

   While this issue was raised with BEST, as far as I know, they never addressed the question of the degree of symmetry. Things got extremely busy at work about the time BEST came out, so while I wanted to look at this question myself, I’ve never had time.

   Regarding Paul’s comment:

   Q5: What are we to make of the fact that UAH and RSS both show “the pause” and the HadCrut4 shows “the pause” but Cowtan and Way’s do not?

   I don’t think you can really assert that Cowtan and Way aren’t showing a pause. There is definitely additional structure in Cowtan and Way (which is probably real, even if not accurately depicted in their reconstruction), but it is virtually identical in slope to GISTEMP.

   At some level, the assumption that Cowtan and Way that Judith takes issue with above is similar to the assumptions used in GISTEMP for the high Arctic (to extend the land temperatures north). So maybe it’s not surprising you end up with nearly identical slopes.

6. By the way, regarding this statement:

   Anyway, the dynamics for land versus marine are enough different that you’d expect any correlation to quickly vanish between inland points and marine points, when the spatial separation gets very large.

   If you make the assumption of correlation, where none is present, and missing data is all in the marine sector, then what you do is smear the land temperatures into marine ones. This is important to note, because while land temperatures show significant polar amplification, sea temperatures do not.

   See e.g. this.

   You do run the risk of overstating the amount of warming with this method. I’d consider the assumptions made by GISTEMP to be in that sense, an extreme. Filling in the missing cells in the Arctic with the the average (anomalized) temperature for the oceans is another extreme. I don’t know anybody who does exactly that, but it’d be interesting to see what happens to the series if you do. Obviously, I’d expect the slope to be smaller with this assumption than with the assumption made by GISTEMP.

7. Troy_CA,

   as a constant offset (even if in error) would not create a trend bias….
   Rather, the bias would occur if the *trends* of delta_T were substantially different in the interpolated regions versus those cells used in the prediction of those regions

   I was pondering this last night and same thing has occurred to me. I agree we need the spacial variation in δT itself evolved over time. That said: That would result in yet a second layer in the hypothetical chain. So it’s not something one should assume does happen. But if those who doubt the reconstruction are going to bring up the “inversions” issue, it might be something that needs to be looked at.

   it is not clear how this would actually the methodology, or even if this single run of the model represents the relationship between T_sat and T_sfc in the real world

   I wasn’t thinking of a single run resolving it. It would definitely be something that would require people to plow through multiple runs. Generally speaking, once your set up to do 1 run, one can do serveral. But it is time consuming, and the question is: Does the scenario of spacial variations of trends δT seem sufficiently plausible to warrant such a thing?

   Carrick

   You do run the risk of overstating the amount of warming with this method. I’d consider the assumptions made by GISTEMP to be in that sense, an extreme.

   If so, one might suggest HadCrut which simply omits the area at the poles (which is in some ways just treating it as having the same trend as the average of the measured part) and Cowtan&Way represent ‘bounding’ calculations to some extent. I suspect they won’t be viewed that way (and of course, in some sense they aren’t because we still have other sources of error which are shared.)

8. It would appear that a critical issue arising with Cowtan and Way could be the polar amplification of global warming. I have not yet read their paper but a change in a global trend, even if relatively small, has to require a large change in trend for the polar regions from consideration of area weighting alone. The global percentage of the areas north and south of 60 , 70 and 80 degrees are by my calculations 13.5%, 6.0% and 1.5%, respectively. I had thought that the polar amplification in the Antarctica, outside the small portion of that region that includes the Peninsula, was significantly less than the Arctic due to isolation of that region by prevailing and circular wind patterns. I would be curious to see what Cowtan and Way have found in these two regions with regards to polar amplification over the past 15 years or so.

   The other region of differing coverage between Cowtan/Way and HADCRU4 is the continent of Africa. The whole of Africa is around 7% percent of the global area and it might have more leverage on global trends than the polar regions with smaller changes. Curious to whether Cowtan and Way attribute more complete Africa coverage to a large portion of the changes they found in global trends.

   I would think that a better comparison of the trends and with fewer caveats between MSU satellite and Cowan/Way would be from Africa. For the Antarctica we have some overlap between Cowtan/Way and O’Donnell, et al 2011, who used AVHRR satellite data and ground stations, and with Steig et al 2009 who also used AVHRR and ground stations but with a different methodology.

9. Kenneth Fritsch (Comment #121203)-“Curious to whether Cowtan and Way attribute more complete Africa coverage to a large portion of the changes they found in global trends.”

   Seems doubtful, given that the general change in trend is to more warming but there isn’t much of that in Africa over the study period as I recall (from the atmospheric data from satellites, at least).

   Far more likely the most important factor is the Arctic. The answer itself isn’t all that unreasonable, seems roughly in line with the the GISS extrapolation as far as I can tell. But I am *very* dubious of the methodology itself as an *improvement* over GISS or not extrapolating at all.

10. From what I could piece together, I like the work. Even though I was looking at it from a distance over the paywall it has much more of a geostatistical feel than BEST which somehow comes across as a little idiosyncratic. However that may reflect personal taste and personal experience. As with BEST I found the Cowtan and Way material frustrating as a result of its not incorporating kriging error analyses as part of the presentation [Maybe it is in the paper and/or supplemental material]. The approach to error analysis is a big asset that geostatistical methods bring to the table–both from the point of view of interpretation and utilization of the analyses, but also diagnostics, and the planning of future work. It is that kind of tool–or more accurately set of tools. Uncertainty here is a big deal, yet neither Cowtan/Way ro BEST seem to have employed geostatistics strongest asset–that raises a caution flag but isn’t the kiss of death. (Aside: the interactions between Cowtan/Way and BEST that Steven Mosher has mentioned is a good thing for both parties and I hope everyone in the community is going to profit.)

    Now, about the first item on whether kriging of surface temperatures (across boundaries, etc.) is physically unrealistic: I find that position very puzzling and a little aggravating. Maybe it is as you suggest and is just the wording. Here is why I disagree rather strongly with JC’s position. At the risk of bringing the hounds of hell down on me, I do not see physical reality as a constraint in exploratory data analysis such as this work. We are often measuring quantities in an effort to found out something about circumstances where we do not yet understand things fully–in particular we seek out hidden relationships in the face of incomplete knowledge. I do not want to limit possibilities in the initial exploratory phase. Here JC said, ‘no it can can not be done because….’ Cowtan and Way said, ‘but we tried it and we really think we saw something. Please scrutinize our work. Look.’ Also in the present case the material presented is perceived by many to be good quality, and it seems reasonable to follow-up on the work. I appreciate that JC made the comment based on some very specific experience, but then so did Cowtan and Way.

    Anyway, to sum it up, I give the work a definite thumbs up in the modern sense of the term. :O)

    At times CE was a zoo and that is really a shame given this particular paper. Still they’ll probably be better off having the shake-down cruise.

11. Andrew_FL (Comment #121206)
    November 15th, 2013 at 2:35 pm

    I am not at all sure all the difference comes from the Arctic. If it does my calculations show the following:

    “Cowtan & Way investigate the claim of a global surface warming ‘pause’ over the past 16 years by examining the trends from 1997 through 2012. While HadCRUT4 only estimates the surface warming trend at 0.046°C per decade during that time, and NASA puts it at 0.080°C per decade, the new kriging and hybrid data sets estimate the trend during this time at 0.11 and 0.12°C per decade, respectively.”

    Using the 0.12 figure for CW, those trend differences between CW and HadCRU4 and CW and GISS translate approximately into the 1997-2012 temperature difference globally of 0.064 degree C and 0.118 degree C, respectively and if those differences came entirely from the Arctic at latitudes 70N and higher, or 3% of the global area, those differences for that area would need to be 2.1 and 4.0 degrees, respectively.

12. Lucia,

    I was pondering this last night and same thing has occurred to me. I agree we need the spacial variation in δT itself evolved over time. That said: That would result in yet a second layer in the hypothetical chain. So it’s not something one should assume does happen. But if those who doubt the reconstruction are going to bring up the “inversions” issue, it might be something that needs to be looked at.

    Right, I do not mean to suggest it does happen, but rather to point out that simply noting a global variation in T_sat -T_sfc is not sufficient to show bias in the methodology…one would additionally need to show global variations in the trend of T_sat – T_sfc (as you say, it would require another link in the chain). I think the figure I put up shows such variations could theoretically exist…however, if I am reading it correctly they also include a non-hybrid kriging reconstruction that doesn’t rely on T_sat at all, which appears to match up pretty well with the hybrid reconstruction. So while I will still want to try the methodology on some GCM outputs if I get a chance, at this point I wouldn’t expect to find any large bias due to the aforementioned inversion issue.

    While the paper is new and it will likely require time for people to process and experiment with the results, I think a convincing argument for bias in the methodology would have to explain:

    a) Why this bias would affect the kriging vs. hybrid approach in such a similar way, and

    b) Why such a bias would not appear in their validation methods.

13. Kenneth Fritsch (#121203) –
    Regarding polar amplification – UAH trends for the period Dec78-Sep13 inclusive:
    Global (85N-85S): 0.0139 K/yr
    NoPol (85N-60N): 0.0468 K/yr
    SoPol (60S-85S): -0.0015 K/yr
    Excluding poles (60N-60S): 0.0126 K/yr

14. mwgrant (Comment #121211)
    November 15th, 2013 at 3:10 pm

    Maybe JC would have been more predisposed to kriging of surface temperatures across boundaries even if it were physically unrealistic if she could have put some error limits on that.

15. Troy_CA

    While the paper is new and it will likely require time for people to process and experiment with the results, I think a convincing argument for bias in the methodology would have to explain:

    a) Why this bias would affect the kriging vs. hybrid approach in such a similar way, and

    b) Why such a bias would not appear in their validation methods.

    Agreed. And these are very strong counter arguments in light of
    (a) The validation being pretty thorough and
    (b) the hypothesized “problem” requiring a series of things to be true. (That is: it’s not sufficient to for inversions to be different at the poles but for the differences in rates of inversions to be changing at different rates and global warming evolves. Both could happen, but when you need two things to happen simultaneously, the speculation that they do cause a problem because more and more speculative.)

16. On whether kriging of surface temperatures (across boundaries, etc.) is physically unrealistic:

    I find JC’s criticism entirely valid. Speaking as a geologist, and lest we forget kriging was “invented” by a geologist interested in how you extrapolate assay values from drill holes into a surrounding ore body, it does make sense that some consideration of physical boundaries is relevant. As I understand the problem at hand, there is a sparse network of temperature stations around the Arctic, some on land, some on water. In the interest of understanding whether temperatures in polar regions have increased or not, C&W use data infilling. It seems self-evident to me that the degree you can extrapolate between stations on or over land and between stations on or over land and sea could be different.
    I would be interested in Steve McIntyre’s opinion on this…

17. Kenneth Fritsch (Comment #121215)-Interesting. I thought I saw a map of what their interpolated data look like somewhere, which might give us a better idea where the increased warming is coming from. If it is coming from Africa I am going to be pretty skeptical of it. And actually I think there is probably a lot of data from Africa not presently digitized and used in these analyses which might allow the gaps there to *actually* be filled in-unlike in the Arctic.

18. Andrew_FL (#121222) –
    Look here for map comparisons between C&W and other methods.

19. nvw

    It seems self-evident to me that the degree you can extrapolate between stations on or over land and between stations on or over land and sea could be different.

    I would tend agree if one was kriging a surface property like surface temperature. But they aren’t doing that. They are krigging somethign that depends more on some vertical temperature profile. I realize that the vertical temperature profile is influenced by the surface below them, but the connection is not as direct as between the surface temperature itself and the boundaries.

    I know that when one sees (C=A-B) one is tempted to think that if “B” can’t be krigged, then C can’t. But that might not be so. It may be that C varies smoothly in a way that is not affected by the boundaries even to B is. To take an extreme: it could be that C is an absolute constant and so trivially krigable! (Note in this case, we would expect that even though C is trivially krigable, then either both B and A will be unkrigable or both will be krigable. )

20. Lucia, I may have missed in the comments, but sudden stratospheric warming event are basically inversions of the entire Arctic air mass where the polar vortex breaks down. Then using UAH would be like kriging to the stratosphere. Since SSW events have pseudo cycles, small during the 90s and shifting to larger after 1998, you can have 2 to 10 C jumps in lower troposphere temperatures over areas that are normally treated like ocean surface even though it is ice.

    There is nothing wrong with kriging or using satellites, but a little curiosity might be required when you suddenly find the answers you are looking for as if by magic.

21. HaroldW (Comment #121223)-Eyeballing, the maps seem to confirm my suspicions, the trends filled in in Africa seem mostly neutral but those added at the poles are strongly positive.

22. The problem with this paper is obvious.
    It relates to the massive increase in temperature trend (0.11 and 0.12 up from 0.04 and 0.08 approx as specified by the authors)
    This is a massive increase for changing (making up) new figures for the modelled sea/land surface temperatures.
    Massive.
    It means perhaps one sixteenth of the coldest parts of the world has anomalously warmed two to three times as much as the rest of the world combined.
    At both poles.
    W heras we know that there has been arctic loss of ice there has been an equivalent Antarctic gain of ice.
    The total sea ice has been negative for the last 15 years.
    IMO Kevin has purely assumed that ice loss equates to increased SST and written a programme that incorporates the assumed heat gain for the missing areas.
    He can call it Kriging or hybrid reconstruction but the fact remains that he just plugged this figure into the methods and came up the increase.
    LUcia could do it in 10 minutes with any of the temperature data bases available.
    It does not make it real.
    Just one opinion.
    If one plugged in the upside down version of the sea ice extent you could claime global cooling.
    One comment , ice melts from the bottom and is loss reflects the the water temperature which has been warmer at the North Pole but obviously colder at the South Pole
    SST/land ST at the poles where it is frozen on top are not reflective of the underlying SST.
    Finally my comment could be disproved by finding just one example of the Kriging or Hybrid model showing a cooling area in the missing areas. It would not be possible as the magnitude of the assumed arctic amplification (which has probably been spuriously applied to the Antarctic sea areas as well) is so large it would not allow a cooling area.
    Why is this so hard to see?

23. Kenneth Fritsch (Comment #121218)
    November 15th, 2013 at 3:52 pm

    mwgrant (Comment #121211)
    November 15th, 2013 at 3:10 pm

    “Maybe JC would have been more predisposed to kriging of surface temperatures across boundaries even if it were physically unrealistic if she could have put some error limits on that.”

    You are having to speculate, but that might be–she didn’t say. [There were other JC comments throughout the page, but this my short answer :O) ]

    BTW we really should be saying ‘interpolating across boundaries.’ It is not a kriging-specific issue.

24. If you want to open each of these two links in a new tab, and then click back and forth between each, you will see how C&W2013 differs by gridcell from Hadcrut4.

    http://www-users.york.ac.uk/~kdc3/papers/coverage2013/media_had4.png

    http://www-users.york.ac.uk/~kdc3/papers/coverage2013/media_hybrid.png

    No matter what the methodology is or how justified it might sound, this is NOT what has happened to temperatures in the regions of difference. Its a methodology which produces inaccurate results.

25. lucia says

    “They are krigging somethign that depends more on some vertical temperature profile. I realize that the vertical temperature profile is influenced by the surface below them, but the connection is not as direct as between the surface temperature itself and the boundaries. ”

    Now that comment makes a connection.

26. Andrew_FL (Comment #121232)
    November 15th, 2013 at 5:39 pm

    “HaroldW (Comment #121223)-Eyeballing, the maps seem to confirm my suspicions, the trends filled in in Africa seem mostly neutral but those added at the poles are strongly positive.”

    You two worked an eyeballing exercise on my old eyes before with the pi that I took as an n, but I am going say those maps with trends are tricky even for young eyes to interpret. First of all the areas are not to earthly proportions and further I am not sure the resolution allows a decent comparison. Actually maps showing differences in trends from series to series would be much better for making comparisons. Can anyone, who has read the paper and SI, tell us if the authors breakdown the differences in series by global regions -or guide me to the data that would allow me to do that?

    Since BEST and CW both use kriging it might be interesting to compare those series trends even if limited to land by the BEST series.

    Also I note the authors and others have pretty much confined the discussion of trends to the 16 year period ending in Dec 2012. What about trend comparisons from the entire length of the CW series? I have done selected CMIP5 models to the observed series of HadCRU4, GISS and NCDC from 1964-2013 Sept and found trend differences that are significantly higher for the models. It would take a large increase in the observed trends to make that significance go away.

    I believe it was the Cowtan and Way authors who noted that 16 year trends are simply too short for making a valid comparison. I see this statement a lot lately but I do not see comparisons at longer time periods (and with more degrees of freedom from which to obtain significance) other than what Lucia has done and my results.

27. Kenneth,

    Can anyone, who has read the paper and SI, tell us if the authors breakdown the differences in series by global regions -or guide me to the data that would allow me to do that?

    Figure 6 in the paper shows a breakdown by regional contribution to their overall diagnosed global bias. By my eyes, since 1997, the Arctic (>60N) contributes about 3/4 of the trend, the Antarctic contributes about 1/4, and the mid latitudes almost nothing. Given the small areas involved, this means they diagnose very strong underestimation in these regions (my BOE calc suggests ~0.6 K/decade for the Arctic).

28. Kenneth Fritsch (Comment #121238)-Well, it is possible I am not assessing it correctly, but I would be happy to see a numerical analysis of where the change in trends comes from geographically.

    Looking at it again, *North* Africa may play a significant role in adding to the trend.

29. Bill Illis –
    Thanks…that back-and-forth method is helpful to me.

    I thought it odd that the Antarctic was so red when the trend is negative (cf.#121217), but the negative trend in that comment was 1979-2013. The maps cover the 16-year interval Jan1997-Dec2012 inclusive, and over that period the UAH trends are:
    Global (85N-85S): 0.0094 K/yr
    NoPol (85N-60N): 0.0553 K/yr
    SoPol (60S-85S): 0.0319 K/yr
    Excluding poles (60N-60S): 0.0043 K/yr
    So both polar regions were warming significantly faster then the average over the period in question.

    I do note, however, that the bulk of Antarctica is the reddest, not the peninsula. (Shades of Steig!)

30. HaroldW (Comment #121242)-It’s not impossible that the structure of short term trends is very different from long term ones, so I can see where there could be a strong interior warming over that period of Antarctica-keep in mind it’s only since 1997.

    I always like to show two maps to show some interesting things about the structure of the long term trends:

    http://s23.photobucket.com/user/gatemaster99/media/GISSNHWARM.png.html

    http://i23.photobucket.com/albums/b370/gatemaster99/GISSNHCOLD.png

    Which I would like to update at some point, incidentally, which shows both the geographic distribution of warming and it’s seasonal structure. Notably the strongest trend is toward warming of cold season anticyclonic air masses. On the other hand this structure is not quite so distinct even over just thirty year periods.

31. So… What does the Fox say!!!

32. I know the dog goes Woof
    and the cat goes Meow
    …
    Perhaps the cow goes mow?

33. Kenneth Fritsch (Comment #121215)
    November 15th, 2013 at 3:34 pm

    Using the 0.12 figure for CW, those trend differences between CW and HadCRU4 and CW and GISS translate approximately into the 1997-2012 temperature difference globally of 0.064 degree C and 0.118 degree C, respectively and if those differences came entirely from the Arctic at latitudes 70N and higher, or 3% of the global area, those differences for that area would need to be 2.1 and 4.0 degrees, respectively.

    This was my very first thought about their paper—we’re only talking about a few percent of the planet in the Arctic. Thanks for doing the calculations, Kenneth. I see your results as a “best case scenario”, since some of the area further north than 70° does have station data.

    As to whether the difference is the interior of Africa, the tropics in general have warmed very little, particularly over that period, compared to the Arctic. So the African claim seems doubtful.

    The second issue that occurred to me was one Lucia listed above—since the pause is present in all of the datasets, HadCRUT4 and UAH and RSS and GHCN and Reynolds SST … how could a combination of them give a dataset without a pause?

    As usual, Lucia, you see the important questions.

    w.

34. Generally, by Kriging people mean fitting data with a particular type of basis function. My understanding is that its like an inverse tangent. This method has some good statistical properties but makes some strong assumptions about the data that may not be justified here. Some analysis of these error terms might go a long way to determing if this method has large errors here.

35. I’ve read the paper now. A few comments:
    1. Their hybrid method is actually (Eq 1):
    T_hy=krig(T_surf-s*T_sat)+s*T_sat
    The parameter s is determined by their various cross-validations. They seem to like a value of about 0.6.

    2. Can you krige across boundaries? It seems to me the critical one is open sea/not, and that is quite variable through the year. They test kriging with cross-validation, with varying result. Ocean good, land not so much (good for 1 or 2 cells). Overall, their Fig 3 shows that hybrid does very much better in cross-validation than just null(BAU) or just kriging.

    3. Although hybrid looked better in cross-validation, the end results in Fig 5 are very similar.

    4. They also test against buoy readings (IABP). That seems quite good.

    5. On why do they not have a pause when others do – I think the facts need checking. C&W do have a pause; and UAH say, has less pause than people often think. If you look at their fig 5b, C&W still follows the same main pattern as HAD4; it just deviates upward near the end to a fairly minor extent.

36. Kenneth Fritsch (Comment #121215)
    “if those differences came entirely from the Arctic at latitudes 70N and higher”

    But they don’t. Their Fig 1 shows it fairly well. Had 4 is very poorly covered below 60S. It shows the UAH trend there, and although it is not as high as the Arctic, it is well above the global average. As Mosh says, if you leave cells out, you are in effect treating them as global average.

    Their Fig 6 shows the region attributions. Currently, it looks like global and Arctic are the same for actual temp, but if you look for trends, Antarctic makes a substantial contribution.

37. Following up on that last issue – could a small region make that difference without huge trends – I calculated latitude average trends for their 5 deg hybrid grid, for 1997-2012 (16 yr). The numbers, in C/decade starting at N Pole, were:
    [1] 1.509 1.692 1.452 1.196 0.825 0.353 0.088 -0.035 -0.034 0.102
    [11] 0.066 0.074 0.086 0.091 0.090 0.093 0.059 -0.011 -0.044 -0.040
    [21] -0.004 0.010 0.063 0.093 0.141 0.094 0.074 -0.048 -0.039 -0.041
    [31] -0.009 0.099 0.566 0.646 0.685 0.901
    So yes, big implied trends in the Arctic, and fairly big in Antarctic, but not impossible.

    As a check, my global av was .1187; they gave .119.

38. I’ve read the paper now. A few comments:
    1. Their hybrid method is actually (Eq 1):
    T_hy=krig(T_surf-s*T_sat)+s*T_sat
    The parameter s is determined by their various cross-validations. They seem to like a value of about 0.6.

    2. Can you krige across boundaries?

    One of my points is that it might be possible to krig(T_surf-s*T_sat)
    even if you can’t krig(T_surf). C&W did krig(T_surf-s*T_sat), not krig(T_surf). And — as far as I can tell– the argument that one can’t “krig” is really an explanation why you can’t krig(T_surf). But for this criticism to apply to C&W, they need to go further and explain why that means you can’t krig krig(T_surf-s*T_sat) wish is a different thing.

    Maybe you can krig both, maybe neither, maybe one or the other. But it’s silly to just say that if you can’t krig one that automatically implies you can’t krig the other. It’s worse to just forget that these are two different things– because they are two different things.

39. Nick, “So yes, big implied trends in the Arctic, and fairly big in Antarctic, but not impossible.”

    It they define a new “surface” called ice phase transition or some such, it is not impossible but to have DJF sea ice covered SST jump 4C is not realistic. Of the 3% of the globe in the 70-90N area the vast majority is ocean which in winter can be reasonably assumed to be ~-1.5C +/-0.5C. As it is their kriging is not related to any current global mean surface area other than Pseudo-Land.

    Or should we adjust the freezing point of water next?

40. dallas –
    “in the 70-90N area the vast majority is ocean which in winter can be reasonably assumed to be ~-1.5C +/-0.5C.”
    In the winter the majority of the Arctic is ice-covered. While the ocean temperature near the surface will be close to freezing as you say, the surface temperature (above the ice) is well below that.

41. HaroldW, “In the winter the majority of the Arctic is ice-covered. While the ocean temperature near the surface will be close to freezing as you say, the surface temperature (above the ice) is well below that.”

    No doubt, so when you append surface air temperature to water temperature you get a up spike in temperature. Apples and oranges. If they are kriging SST then there are physical limits, but stick air temp on sst you get some exotic fruit salad.

    Now if they would like to krig the marine surface air temperature and create a new metric, go for it.

42. The appropriate value for s should not be the same everywhere and constant in time.

    I also think .6 is too high for the short term global average but that is another story.

43. Pardon, too *low*. I was think of the inverse.

44. Nick Stokes (Comment #121250)
    November 16th, 2013 at 7:21 am

    Thanks, Nick, for the trend data in latitudinal zones from CW. As I stated in my initial post I believe the critical issue of CW could be polar amplification – which is also present in GISS and UAH over the 16 year period as you note- but perhaps not to the same extent as CW.

    I guess another interesting look at these data would be these same zonal trends over the entire time period covered by CW.

    HaroldW’s trend date from UAH for the period 1979-present showed a very much amplified Arctic polar trend and none for the Antarctica. I am curious whether we are seeing an increase in the Antarctica amplification and slow down in the Arctic amplification over the past 16 years or perhaps simply looking at “weather” noise in both regions that is just that, i.e. noise.

    I am probably having selective memory here but I recall thinking after the analyses of Steig et al and O’Donnell et al that using the MSU like AVHRR (as done by O’Donnell et al) might be an interesting project. It would be nice for me to find a post were I stated that. I think it might have been to an author of O’Donnell et al.

    I am even more interested in the CW and BEST comparisons now because, of course, BEST does not use MSU in their kriging. We can have a three way comparison of global land temperatures: CW krige, CW krige and MSU and BEST krige. I continue to have not read the CW paper so I do not know how they used the raw station data and whether it was similar to the weighting that allowed BEST to use all the data and even data over short time periods.

45. Here are some actual long-term high latitude stations (direct from the sources of UK Met Reader Antarctica and Environment Canada rather than the adjusted GHCN or GISS data). Should be clear that starting in 1979 or 1997 provides a false signal.

    In order of Amundsen Scott 90S, Mawson 70S, Faraday 65S and Eureka Nunuvut Canada 80N – all world-class weather stations staffed by several researchers.

    http://s13.postimg.org/6w98pvd8n/Amund_Scott_90_S.png
    http://s8.postimg.org/7l41rf5b9/Mawson_70_S.png
    http://s11.postimg.org/q80w8vwv7/Faraday_65_S.png
    http://s7.postimg.org/6y6u8vpuz/Eureka_80_N.png

    If someone knows where the raw data for Greenland from DMI is, I’d appreciate it.

46. I see I should have gone further back in this thread to look for replies as I now see some of the questions I had have received replies. I need to mull those replies over. It would appear though that the Arctic polar amplification from HaroldW’s UAH data has not slowed down in recent times.

47. I am attempting to put Nick’s trends for CW together with HaroldW’s for UAH polar trends and with Troy’s comment on where the warming went and with Andrews eyeballing and my first impression is that these impressions and data are rather consistent and that CW’s series puts at lot more warming into the Arctic than UAH and also more into the Antarctica – although one has to be careful when weighting the areas that Nick provided in 5 degrees of latitude in the polar regions. The area from 85-90 is a diminishing small portion of the global area and much smaller than the 60-65 area.

48. Bill Illis,
    Is the station for Eureka the homogenized version from Environment Canada or the non-homogenized (raw)?

49. I did a simple minded area weighted trend calculation for the polar regions from 60-85 from Nick’s data for CW for the period 1997-2012 and compared it below to the HaroldW data from UAH for the same approximate period and the 60-85 zones.

    For the zone 60N-85N:

    CW trend 12.3 degrees C per century
    UAH trend 5.5 degrees C per century

    For the zone 60S-85S:

    CW trend 4.8 degrees C per century
    UAH trend 3.2 degrees C per century

    I am definitely recommending to my grand children to tell their children to plan vacations at the North pole and not the South.

50. Kennth Frtisch, “CW trend 12.3 degrees C per century”

    I am sure that is sustainable 🙂

51. Kenneth Fritsch (Comment #121268)
    I’ve plotted the latitude trends, with some comparisons,
    here. The C&W hybrid trend is globally not that much higher than UAH, although the polar amplification is pretty high.

52. R (Comment #121265)
    November 16th, 2013 at 1:26 pm
    Is the station for Eureka the homogenized version from Environment Canada or the non-homogenized (raw)?
    ————–

    This is a world-class research station which started taking temps 4 times per day in 1953, gradually changing to hourly over time. The temps are quality controlled, and there is an occasional missing day but the raw and homogenized are the same. They even go out in the bay and hand-drill the sea ice to measure ice thickness every week and have done so since 1947.

    http://upload.wikimedia.org/wikipedia/commons/1/16/Eureka_Weather_Station_1997-08-04.jpg

    http://wattsupwiththat.files.wordpress.com/2010/04/labelled-eureka-arial-1.jpg

53. So you have checked that the raw and homogenized versions are the same? Afterall there was a TOB adjustment to the homogenized stations in the homogenized database.

54. looks to me like Antonys UHI

55. perhaps I should elaborate….the satellites have scanned the area for years and although their coverage is not perfect…even though they warn against using the data,.. surely they would have registered a trend…come on. Even bad data can show a trend.

56. Another thing that flummoxes me is how the guys at DMI obtained such wrong data…have they chimed in yet?

57. basicly everyone that has been measuring temperature at the poles has to re-write their data….just love it …fund me fund me.

58. “Kenneth Fritsch (Comment #121268)
    November 16th, 2013 at 2:17 pm
    I did a simple minded area weighted trend calculation ”

    Kenneth I did the same but comparing some of Nick’s zonal trends with model zonal trends. The only one’s I could get working on KNMI Climate Explorer ATM are CMIP3. I’m also lazier than Nick and did just 3 bands (80N-75N, 45N-40N, 25N-20N). So here are Nick Cow et al estimates (first) and CMIP3 (second)

    80N-75N, 1.452, 0.45
    45N-40N, 0.102, 0.25
    25N-20N, 0.091, 0.21

    (I’m sure somebody with more efficient data handling skills can fill in the gaps and do CMIP5)

    In the search for more heat they seem to have piled it into the poles were the measurements are lacking. In losing the pause and possible making the models more realistic again they seem to have lost a realistic rate of polar amplification. Or at least polar amplification that matches the models.

    I guess you can’t just pile a whole lot of heat into one region of the global and expect the wider story to remain unaffected.

59. HR (Comment #121279) November 16th, 2013 at 8:08 pm
    “In the search for more heat they seem to have piled it into the poles were the measurements are lacking.”

    But measurements aren’t lacking – they have used the UAH data. And UAH also has high trends near the poles. Not quite as high, and that may be a question. But much higher than CMIP5, which may be underestimating Polar amplification.

60. In fact, it isn’t just me thinking that CMIP5 may underestimate Arctic amplification. Article in Nature:
    “Recently, observations confirmed that the Arctic has warmed faster than the rest of the globe during the last few decades with a very pronounced seasonality as winter warming far exceeds summer warming in conjunction with sea ice retreat
    …
    Most CMIP5 climate models underestimate Arctic winter warming over the past two decades as compared to reanalyses data and observations, and overestimate summer warming”

61. My graph and findings were Dec/Jan/Feb for 7 years ending July 2013:

    When I graphed HADCRUT4 by latitude band (as of July) I got 1.73C/decade for 80-85 and .6 for 75-80.

    The next 7 bands (70-75 down to 40-45) were negative.

    I did not find enough data to graph above 85.

    C&W method found .825 at 67.5 while I found -2C per decade. (Different time frames)

    http://sunshinehours.files.wordpress.com/2013/10/highlighted-dec-jan-feb-hadcrut4-by-latitude-band-last-7-years-as-of-2013-7.png

    I’ll try and compare same periods later.

62. Shall we look at the last 100 years of HADCRUT4 data for the top 3 latitude bands with data.

    Its been warmer. A lot warmer. In the past.

    12C and 15C anomalies.

    http://sunshinehours.files.wordpress.com/2013/11/lat-hadcrut4-by-latitude-band-last-100-years-as-of-2013-9.png

    I stand be what I said a few days ago.

    A satellite based (or hybrid) only record would erase the past very conveniently.

63. Nick Stokes, “In fact, it isn’t just me thinking that CMIP5 may underestimate Arctic amplification. Article in Nature:”

    Arctic Winter Warming is probably a better term than “polar amplification” If the rate of heat transfer to the Arctic increases for whatever reason the temperature is amplified with or without CO2 involved. CO2 would change the amount of amplification, but you need a longer data period to compare degrees of AWW since it is part of the ~60 pseudo-cycle ala “Stadium Wave”.

64. Q1: Why should it matter whether kriging of surface temperatures is unrealistic?
    Polar [arctic only] amplification is due to decreased albedo from less ice causing more absorption of heat and to hot air currents going northwards from the tropics [Wikipedia paraphrase]. The latter happens all the time so would cause the Arctic and Antarctic to be slightly hotter than one would expect from straight radiation but it would not cause an increased trend [EVER]. Less ice less albedo might.
    A local area having an increased heating trend is primarily current and sea temperature driven, not an excuse to bump the global trend we have all relied on upwards

    The antarctic with an extra 1 million square kilometers of ice stretching out to latitudes much further out than the arctic ice reflects consequentially a lot more heat. In fact if you Krig or modify surely one should be expecting polar dis amplification at the South Pole only.
    The Cowman approach has a built in lightning rod of error.
    It purports to show complete correlation when testing all known areas by removing a known area and reinfilling. Obviously there should be some area most of the time with 50/50 positive and negative results.
    But what do we find when using it on unknown areas.
    100% positive trending warming.
    Not because it “exists” but because the program has a warming bias built into it using scanty information on suspected arctic warming trends.

65. Nick I’m not against the idea that polar amplification might be higher than already reported but there has to be some point at which it’s so high that it challenges the basic theory.

    https://courses.eas.ualberta.ca/eas570/arctic_amplification.pdf

    So back in 2006 Serreze was pegging it at 2X-4X i.e. the poles will warm 2X-4X the rate of the rest of the hemisphere, this included extra warming in winter and is in-line with the mid-20th century arctic warming. I haven’t done the maths but your numbers look like the arctic is warming 10X or more than the rest of the hemisphere. That seems quite high and deserves an explanation.

66. Angtech

    The change in albedo due to less ice is insignificant due to the low angle of incidence. There is more variation in albedo due to cloud cover than to changes in ice coverage.

    Change to evaporation rates as a result of less ice cover are a more plausible mechanism.

67. HR (Comment #121288)-“mid-20th century arctic warming.”

    Pardon me but I presume you mean *late* 20th century. There was no Arctic warming in the *middle* of the century.

68. Nick, thanks for the plot of HadCRU4, UAH and CW trends by latitudinal zones. I think your graph shows very definitively that CW has significantly increased the polar amplification of warming over the past 16 years – and even over UAH. Again this might be just weather noise for that 16 year period for both UAH and CW. A plot of the same zonal trends for the period 1979-2012 would be instructive.

    The point of contention from Nic Lewis and others familiar with the limitations of the MSU is in the region where CW finds most of the increased warming as applied to global warming. That might limit the analyses to those regions in determining the validity of CW’s approach and methods.

    I recall during the Steig and O’Donnell debates that I attempted, for my own purposes, to compare the spatial correlations from AVHRR (which both Steig et al and O’Donnell et al used in their methods) to the UAH MSU data in the Antarctica where there was overlap and as I recall the correlations were not good. Maybe someone could repeat that comparison. Recall that O’Donnell et al used the AVHRR for spatial correlations with weather stations. They presented evidence (not necessarily in the published paper though) that the AVHRR measurements were not stable over time.

69. Lucia, on December 22nd at the equator the temperature used is climate science is the Tmin, which occurs before day break, added to Tmax, which occurs in the early afternoon, and the total is divided by 2.
    What is the ‘temperature’ used at the North and South poles on December 22nd?
    Is it reasonable to compare (Tmain+Tmax)/2 with some other temperature metic?

70. “As a check, my global av was .1187; they gave .119.”

    My simple minded approach for weighting the trends appears to be reasonable as it agrees with Nick’s result above – unless he calculated the same way I did.

    I get 0.1187.

71. Andrew_FL: “There was no Arctic warming in the *middle* of the century.”

    HADCRUT4 (when it has data) disagrees. The warming from 1918 into the 1950s was pretty spectacular.

    http://sunshinehours.files.wordpress.com/2013/11/lat-hadcrut4-by-latitude-band-last-100-years-as-of-2013-9.png

72. Apologies if this has been previously mentioned, but the UAH documentation file contains the following:

    ALSO BE CAUTIOUS USING LT AND MT OVER HIGH TERRAIN ( >1500 M)

    The areas of poor anomaly values are : Tibetian Plateau, Antarctica, Greenland and the narrow spine of the Andes.

    Was this discussed in the paper?

73. “Was this discussed in the paper?”
    Yes. They say:
    “As has been noted the MSU lower troposphere data is potentially subject to biases over ice and at high altitudes. The use of anomalies for both surface and satellite data eliminates the effect of a constant offset, however differences in temperature variation over ice or high altitude surfaces could bias the results. The fact that the hybrid method gives better cross validation results than kriging for the Antarctic and high Arctic land stations suggests that this is not a major issue. Furthermore there is no obvious evidence of worse performance at high altitude in Figure (3).”

74. Thanks Nick. I’m at a bit of a disadvantage without the text of the paper itself.

75. Nick, I found the CW data for the hybrid and it agrees with yours for 1997-2012 and I now also have the latitudinal trends for 1979-2012. I note that you must have the UAH gridded data for the graph you linked to. I see that data on line, I think, but it is in separate files by year. The form in which it is usually reported is that which HaroldW evidently used and is broken down for polar only by 60-85 North and South. Do you have a link to UAH data that would be in a form similar to what CW used for the hybrid data?

76. John Vetterling thanks but albedo is given as one of the two causes of polar amplification
    The albedo effect at the South Pole is much greater than at the north as the angle of incidence is much greater (though still low) and the amount of heat reflected is much, much greater due to the ice area being further out.
    A trend of 10 times greater for the arctic amplification for 15 years compared to the hemisphere (HR) is physically impossible, Nick Stokes, it is so way out there unbelievable that only a 4 year old child could say it with a straight face and expect to be believed.
    Do your job as a scientist, not an apologist and insist on some credibility, please.

77. Nick, never mind. I found the UAH gridded data in the form I need at the CW website.

78. By the way if anyone wants these data the link is:

    http://www-users.york.ac.uk/~kdc3/papers/coverage2013/index.html

79. Andrew_FL (Comment #121290)
    November 17th, 2013 at 9:51 am
    HR (Comment #121288)-”mid-20th century arctic warming.”
    Pardon me but I presume you mean *late* 20th century. There was no Arctic warming in the *middle* of the century.

    No pardon me I actually meant the early 20th century warming (1920’s-1940’s)

    http://journals.ametsoc.org/doi/abs/10.1175/1520-0442(2004)017%3C4045%3ATETWIT%3E2.0.CO%3B2

    http://www.mpimet.mpg.de/fileadmin/publikationen/Reports/max_scirep_345.pdf

    http://www.the-cryosphere.net/6/1231/2012/tc-6-1231-2012.pdf

    http://acsys.npolar.no/meetings/final/abstracts/posters/Session_3/poster_s3_096.pdf

    etc etc etc

80. KF:
    “Do you have a link to UAH data that would be in a form similar to what CW used for the hybrid data?”

    I used the file in their “coverage” dir. This is in HAD 4 format. I’ve posted my code here with some comments on where the files come from.

81. HR (Comment #121302)- apologies, I thought it more likely to say mid century when meaning late century, than when meaning early. Of course early or late works in the Arctic. I am familiar with the earlier warming, too.

82. I have linked two graphs with the one in the first link showing the latitudinal zone trends for the CW Hybrid and UAH for the time periods 1979-2012 and 1997-2012 and the one in the second link showing the latitudinal trends for CW Hybrid and UAH for the period 1979-1996 and again showing for reference for the time period 1997-2012.

    These graphs show that the polar amplification for the periods 1997-2012 are present in both the CW Hybrid and UAH series with CW Hybrid showing a considerably greater amplification than UAH in both polar regions. The 1979-2012 period which may be sufficiently long that some of the “weather” noise in the 1997-2012 is subdued somewhat, but that longer period shows essentially no amplification for south polar region for both CW and UAH while the north polar regions shows polar amplification for both CW and UAH with CW being greater.

    The above comparisons lead naturally to a comparison of the 1979-1996 and 1997-2012 periods which is shown in the graph in the second link. Here we see no polar amplification for either CW or UAH in the south polar region and a smaller amount for CW in the north polar region and none for UAH.

    It appears from these graphs that the warming changes from 1979-1996 and 1997-2012 are relegated to the polar regions and in the extreme for CW Hybrid. It would interesting to see a direct comparison with some climate models and perhaps determine better whether these large changes in the polar regions can be considered “weather” noise.

    http://imageshack.com/a/img51/3035/fp9c.png

    http://imageshack.com/a/img36/7329/3m7k.png

83. There’s a review article of polar amplification by Serreze and Barry here. For Arctic, he points to longer ice-free, both because of albedo and just longer exposure of relatively warm water. For South, he mentions the ozone hole. The Arctic explanation suggests that there would have been a lot of warming recently, since we know the ice has receded.

84. https://courses.eas.ualberta.ca/eas570/arctic_amplification.pdf

    Kenneth Fritsch,

    The above Serreze papers show the polar amplification for a dozen or so models in Fig1

    On another point doesn’t Lucia live in Illinois? Any weather impacts?

85. Kenneth Fritsch your graphs are interesting. Suggests maybe we should be calling all this Polar Warming rather than Global Warming

86. I have put up a post which does a much simpler infill for missing cells for Hadcrut 4. It just estimates an average for each latitude band, and uses that to infill empty cells, rather than using the global average, either explicitly, or implicitly by omitting them. It would be equivalent to forming the global average by averaging each band, omitting missing cells, and then averaging the band averages.

    For the 1997-2012 period, it nearly doubles the trend, from 0.046 C/dec to about 0.084. The more elaborate C&W hybrid goes to 0.119. I think mine lags because its estimate near the N Pole, though better than global, still relies on arctic fringe numbers to stabilize the estimate.

87. Nick

    Before you do the weighting, if you take latitude bands of raw data and find the average, this should be the same as when you infill data. This is because you’ll be interpolating and typically it’ll reduce to an average of some subset or the complete set average for that band.

    If you then weight and the infilled data set has a different average than non infilled, accounting for a slight deviation due to residuals in the model you choose, you’ve done something wrong. Doubling of the trend is one such case.

    The C&W paper differs because they use a separate data source supposedly calibrated to Tsurf. So variations in this can be related to a predicted Tsurf. I have reservations about the claimed accuracy of Tsurf but thats by the by. Your method appears to only use one data set and as such there shouldn’t be a significant difference in band averages for non filled and infilled.

88. HR (Comment #121307)
    November 17th, 2013 at 9:27 pm

    I want to see the models graphed in exactly the manner I used for UAH and CW Hybrid. I may do that with some of the CMIP5 models if they are available at KNMI in gridded form.

    I live in a city adjacent to Lucia and we received no damage, but south of us they had some severe wind damage.

89. What all this seems to be saying to me is that far from being global in nature a disproportionate amount of the warming is happening at the poles (most true during the pause). Getting the poles right seems to be crucial. And the greatest uncertainty (poorest data and greatest debate about how to handle it) seems to be at the poles.

90. Nick Stokes (Comment #121309)
    November 18th, 2013 at 6:04 am

    Your model to me at first glance would appear as reasonable as any if you go the infilling route and takes advantage I would suppose of trends by latitude. You are, however, fitting your model with a supposedly “correct” result in mind. How does it differ from GISS? And why do we not see NCDC trends in these comparisons?

91. You plot is very nice Nick, you have shown that using your method we have 1.5 degrees warming in the antarctic, which shows that this method of infilling is artifactual. The one thing we do know is that the stations of the interior of the Antarctic show no warming since the 60’s. The UHA data shows the same for post-80’s warming.
    Always nice when you can compare a prediction against reality and find that your model fails.

92. DocMartyn:

    The one thing we do know is that the stations of the interior of the Antarctic show no warming since the 60′s.

    We don’t really know that since we have such a limited number of stations. Still I suspect coastal stations in the Antarctic Peninsula are getting smeared into the interior, so the interpolations aren’t valid.

    One thing I’ve still not seen is a breakdown by season. I’d like to see in which season the amplification is maximum for the extreme polar regions.

    Brief instructions for hot to generate these trends in GISTEMP: Start with this link. Make sure to set map type to “trends”, select the month, season, or annual values using “mean period” and of course set the years of interest. Click “make map”, which will open a new window with graphical data, then scroll down to “Download the zonal means plot as PDF, Postscript or text file” and select the “text” link.

    Here is my plot of their data

    Data

93. Carrick, In the Arctic is is DJF with the most amplification, jja has the least.

    http://www.nature.com/ngeo/journal/v4/n11/full/ngeo1285.html

    Since UAH is more likely sensing clouds, the effect is amplified. This leads back to the abysmal Arctic energy budget estimates due to mixed phase clouds. A subject that Curry has quite a bit of experience with.

94. As a by the way gang, mixed phase clouds produce effectively a radiant barrier to the “normal” energy transported by the atmosphere to the arctic. While the surface is being warmed by its own energy release interacting with the mixed phase clouds the upper troposphere energy being transfer in interacts with the top of the mixed phase cloud radiant barrier. You get greater than “normal” Stratospheric warming.

    In the early K&T Earth Energy budgets this lead to an error of ~18Wm-2 in their “atmospheric window energy” portion which was corrected with the Stephen’s et al. energy budget.

    Since the “best of the best” are working on this problem I find it amusing that they tend to be unable to update their knowledge base of basic “Weather” events and basic concepts of the “Earth Energy Budget”.

95. And for you radiant physics fans,

    http://upload.wikimedia.org/wikipedia/commons/9/97/Water_infrared_absorption_coefficient_large.gif

    Some consider mixed phase clouds, that would be all three lines together, “Effectively” saturated.

96. Kevin has done a great job putting together this FAQ portion of the website aimed at some of the more common questions we have been getting. There is a short discussion which can be toggled (more/less) to get a longer one for each.

    http://www-users.york.ac.uk/~kdc3/papers/coverage2013/faq.html

    There are a few things that we put on there to answer the points brought up by Dr. Curry. We will continue to update to answer additional questions that arise from the use of our data.

97. Its worth pointing out that the warming due to spatial interpolation (hybrid kriging in this case) is quite similar to that in GISS. I believe that the authors find that GISS + the new Hadley SST corrections leads to a trend just a tad smaller than in their new method.

    Specifically, they say:

    “If the Met Office sea surface temperature corrections are applied to the NASA data, the resulting 16 year trend (i.e. 1997-2012) is 0.103°C/decade. Using the Met Office data and a similar reconstruction method we obtain a similar trend of 0.108°C/decade. Our best reconstruction including the satellite data shows a trend of 0.119°C/decade.”

    http://www-users.york.ac.uk/~kdc3/papers/coverage2013/background.html

98. Zeke,
    They have only just begun to “adjust” the records. By the time they are finished “correcting” the record, it will turn out we wil have been fortunate the world did not catch on fire from this planetary fever.

99. Robert Way,

    Thanks for that link. Having read through some of the FAQ the main issue is your uncertainty in infilling. As someone who has used infilling methods for plasma distributions and for creating performance maps of ion thrusters, you and Kevin may be facing a losing battle. You have severely underestimated the uncertainty of measurements itself.

    For example, in order to get a “clean” surface signal you need every measurement to be taken in the same manner with the same uncertainty (including observer bias and local environment). Otherwise the uncertainties are realistically going to be around ± 2 °C per point. And it doesn’t matter how many times you take them: this is systematic error. In Judith Curry’s article she quotes John Kennedy who is saying the same thing. But speaking as an experimental physicist and engineer, if you don’t have a consistent set of dataq taken using similar methods then the process of measurement itself adds a large error to the data itself. And also to your correlations. Spatial correlation will not catch this.

    The satellite data on the other hand has less error as it uses the same instrument. The issue with the satellite data is different: you need proper calibration. On the Cryosat mission ESA had long campaigns of having test sites to calibrate the thickness measurements of the satellite. It was long and hard work and there are still uncertainties. In your update you state a 67% correlation of satellite data to surface trends for the Antarctic station of Vostok. That’s a 33% error on the real signal. It kind of dwarfs any statistical trend.

    I appreciate the work is interesting but there is just too much measurement noise to justify any trends that have been stated. If a statistical argument is to be used it has to be predicated on the Central Limit Theorem and I’m sorry for real systems and real measurements you can’t do that. CLT is time independent – eventually all data sets will show a normal distribution – but when? 1000 years time? Systematic errors dominate unless you have equipment with better accuracy and resolution.

    It’s a good paper to show how hard it is to produce meaningful results. But any resulting trends are academic.

100. http://climateaudit.org/2013/11/18/cotwan-and-way-2013/

     “Next, here is a simple plot showing the difference between the CW13 hybrid and HadCRUT 4. Up to the end of 2005, there was a zero trend between the two; the difference has arisen entirely since 2005.”

101. Zeke:

     Its worth pointing out that the warming due to spatial interpolation (hybrid kriging in this case) is quite similar to that in GISS.

     Agreed. But since I happen to think that GISTEMP is likely biased high for the Arctic (especially in the Eastern Hemisphere), that isn’t necessarily reassuring…

102. I have put up a post which does a much simpler infill for missing cells for Hadcrut 4. It just estimates an average for each latitude band, and uses that to infill empty cells, rather than using the global average, either explicitly, or implicitly by omitting them. It would be equivalent to forming the global average by averaging each band, omitting missing cells, and then averaging the band averages.
     For the 1997-2012 period, it nearly doubles the trend, from 0.046 C/dec to about 0.084. The more elaborate C&W hybrid goes to 0.119. I think mine lags because its estimate near the N Pole, though better than global, still relies on arctic fringe numbers to stabilize the estimate.

     ########################

     very nice nick.

103. Micky H Corbett (Comment #121310)
     “If you then weight and the infilled data set has a different average than non infilled, accounting for a slight deviation due to residuals in the model you choose, you’ve done something wrong. Doubling of the trend is one such case.”

     Well, I think the something wrong is the original non-treatment of missing data by just leaving it out. The data is non-homogeneous – we expect some regions to behave differently to others, and just omitting missing values means that the result depends on where those missings occurred.

     I like explicit infilling – you can see what you are doing. But an alternative is to identify reasonably homogeneous sub-regions; in this case latitude bands. Sub-average those, omitting missings. Here omission doesn’t hurt so much because of the homogeneity. Then average the averages. This would give identical results without missing data, but with missing data it’s different and better.

104. Micky H Corbett, as I point out above, at the equator, one gets a Tmin and Tmax every 24 hours, but at the poles you get a Tmin and Tmax every 365 days.
     I don’t much mind comparing Apples and Oranges, but this smacks me a bridge too far.
     Just what are they measuring as ‘AVERAGE TEMPERATURE ANOMALY’ has they make the transition from equator to pole?

105. DocMartyn (Comment #121314) November 18th, 2013 at 8:03 am
     “You plot is very nice Nick, you have shown that using your method we have 1.5 degrees warming in the antarctic, which shows that this method of infilling is artifactual.”

     Alas, please see the correction – I had the sign of latitude wrongly marked on the axis.

106. Kenneth Fritsch (Comment #121313) November 18th, 2013 at 7:52 am
     “Your model to me at first glance would appear as reasonable as any if you go the infilling route and takes advantage I would suppose of trends by latitude. You are, however, fitting your model with a supposedly “correct” result in mind. How does it differ from GISS? And why do we not see NCDC trends in these comparisons?”

     As I said in #121326, what I have in mind is homogeneity. You don’t have to see it as infilling; the same result comes if you start by averaging sub-regions chosen for homogeneity (omitting NA), and then average those.

     My result is very close to GISS (trend for 1997-2012). I haven’t included NCDC because C&W kindly provided the HAD, CW and UAH files in the same format (but not I think NCDC). NCDC has similar omissions, so a similar analysis would be interesting.

107. NickStokes, “I like explicit infilling – you can see what you are doing. But an alternative is to identify reasonably homogeneous sub-regions; in this case latitude bands.”

     Explicitly infilling what with what? This appears to be explicitly infilling SST data with land to mid-troposphere data. So you have a 5C change from -30C to -25C indicating a SST temperature change from -1C to 4C that more than doubles a trend known to be likely on the low side. If you correct hadsst with basically bogus infilling you have an unreliable product. You could name the new product something else, but the land product and SST product would have another step to include the ice/troposphere adjustment. Or do you think the SST correction based on winter mid-troposphere warming events is the right way to go?

     Now if the lower troposphere is the goal you could krige UAH and RSS to produce a better product possibly, but as it is, every time there is a winter warming event at either pole where each degree of anomaly has roughly 61% of the energy and an “average” anomaly you make global surface temperature an even more questionable metric to compare with OHC anomaly. Then to boot, these polar winter warming anomalies which happen in both poles are related to increase heat loss. You are basically making the argument that the globe is warming because it is cooling.

108. Nick Stokes (Comment #121329)
     November 18th, 2013 at 3:01 pm

     Nick, I thought you did some weighting with latitudinal distance that might be considered a parameter for fitting -perhaps I misunderstood.

109. Kenneth Fritsch (Comment #121331)
     Kenneth,
     The parameter just comes in when you try to get a latitude average when there is almost no data (eg N Pole). Then you have to accept data from adjacent latitudes. I used a smoothing system that had a parameter (which I varied to check).

110. dallas (Comment #121330)
     “Explicitly infilling what with what? This appears to be explicitly infilling SST data with land to mid-troposphere data.”

     With your best estimate of what the data might have been. You can’t get around the fact that the data is missing. The idea is to do least harm to the average.

     Suppose you have a month of 3 hourly readings and you want the monthly average temperature. That’s lots of info. Suppose a couple of 3am readings are missing. Still lots of info.

     Suppose you just average everything without the NA. That biases high – you know 3am is cold. You’ve effectively infilled with the monthly average.

     There are ways that you could do the average. All the numbers in one go. Or you could average each day, then average the days. Or you could average each hour over the days, and then average those eight sums.

     With no NA, it makes no difference. But with NA, it does. If you average over hours, you’re effectively treating the 3am NAs and being like any other 3am reading. That’s better than saying it’s like any other reading for the month. It removes the warm bias (of omission).

     I would replace the 3am NAs with a 3am average. It has the same arithmetic effect, but you can see what you are doing.

111. Nick

     My point was that in the absence of any other data then any interpolation must be consistent with the data you have. If it’s not then there is something wrong.

     Now if you then look at say terrain and notice that an interpolated point is on a mountain you could adjust using say the lapse rate or some other model. But that would add extra information to your data set.

112. Nick, “With your best estimate of what the data might have been. You can’t get around the fact that the data is missing.”

     Right, which for ice covered ocean would be ~the freezing point. That is an NA with a high degree of confidence in what the actual would be.

113. For those of us who do not have access to the paper, does it discuss why they take the approach of combining surface and satellite measurements instead of using only satellite measurements?

     Found this interesting article about satellite climate monitoring:
     “Satellites do not measure temperature. They measure radiances in various wavelength bands, which must then be mathematically inverted to obtain indirect inferences of temperature.[1][2] The resulting temperature profiles depend on details of the methods that are used to obtain temperatures from radiances. As a result, different groups that have analyzed the satellite data have obtained different temperature trends. Among these groups are Remote Sensing Systems (RSS) and the University of Alabama in Huntsville (UAH). Furthermore the satellite series is not fully homogeneous – it is constructed from a series of satellites with similar but not identical instrumentation. The sensors deteriorate over time, and corrections are necessary for satellite drift in orbit. Particularly large differences between reconstructed temperature series occur at the few times when there is little temporal overlap between successive satellites, making intercalibration difficult.” ”
     http://www.princeton.edu/~achaney/tmve/wiki100k/docs/Satellite_temperature_measurements.html

     Chris

114. As a humble member of the public who is expected to pay for most of this climate science research guff and who as a member of that low life uninformed public who are expected to lay down and be data raped by every passing wannabe climate scientist, it seems to me that the finangling [ I could use some quite unprintable language to describe this ] of the data where no data exists to get a result that ensures that what is seen to be happening in the climate, isn’t according to the non existent data.
     In this case attempting to dispel the idea that there is a “Pause” in the warming using some fancy and argued about statistical techniques applied to that non existent data taken from a region where there are almost no records to justify the claim there isn’t a Pause of over 16 years running in the climb in global temperatures.

     So why is it that this particular statistical lash-up couldn’t also be applied to those 20 years of supposed increasing temperatures from 1978 to 2007, a period which is only 4 years longer than the Pause and on which 20 years the entire global warming meme / ideology is based.
     Using Cowtan & Way’s statistical techniques and the same identical data base of temperatures it could probably be proven that there was no increase in global temperatures during that 20 years of supposed warming.

115. Nick, if you only have 6 hourly data, how the H**L do you get an average temperature, all year round, for Tmin and Tmax?

116. DocMartyn (Comment #121338)
     “Nick, if you only have 6 hourly data, how the H**L do you get an average temperature, all year round, for Tmin and Tmax?”
     I was referring to a notional calculation of linear average.

     (Tmin+Tmax)/2 has been used because a lot of our records are in that form, taken from 24hr readings of a specific device that records them. You work with the records you have. But the averaging principles are the same.

117. Sunshinehours has some more temps by latitude data. Looks to be an error at the end of the 85S-90S band for example. Otherwise, seasonality looks to be an issue.

     http://sunshinehours.files.wordpress.com/2013/11/lat-had4_hybrid-had4-by-latitude-band-last-15-years-as-of-2012-12.png

     Last 8 years zoom-in.

     http://sunshinehours.files.wordpress.com/2013/11/lat-had4_hybrid-had4-by-latitude-band-last-8-years-as-of-2012-12.png

118. “Nick, You work with the records you have. But the averaging principles are the same”

     Are they? Are the times the temperatures taken, every six hours, the same in 2013 as they were in 1960?
     Do you actually record Tmax and Tmin during the measurement?
     If your reading are taken at a time that is near Tmin, and especially Tmax, a shift of sampling time will make take you closer or further from the trough/peak.
     I am informed by Mosher that the TOB is a real problem with many records. Taking measurements every six hours, or worse, intermittently, is something of a can of worms.

119. ROM,
     The AGW movement is now regularly relying on sample sizes of zero to produce results to sustain the climate crisis and to prove just how crazy and conspiratorial skeptics are. For most people drawing conclusions with an absence of data is, to be kind, a creative approach. I can imagine some bright young actuary presenting this sort of product to an internal review board making decisions that actually risked a company’s own capital. The presentation would not last long and would be a career limiting move on the part of the young actuary.
     As you point out, if the creation of data from zip is good for the past 16 years, it should be wonderful over the past 34 years or so.
     It would be interesting to know what the authors would admit to as why they picked the dates they did.

120. I’ll start this by saying: I still like Cowtan and Way, and this post is not intended to suggest otherwise. Moreover, I will continue to like it as a contribution even if it turns out that some deficiencies are found in the future. The paper is solid.
     Lucia, After 119 entries, 87.6 of which are fairly convinced that the paper has been put out by the shonky brothers masquerading as scientists, do you still adhere to the papers solidness?
     As to your first point I still like Cowtan and Way as well. It rivals Lewindowsky in its breathtaking lack of insight as well as that Australian Girl who tired to show a Southern Hemisphere hockey stick.It is papers like these that will eventually destroy the AGW case.
     Made up data , with an obvious agenda in mind, with input that can only make the data go in one direction like the global warming models.
     As a mathematician, when the temperature can go in 2 directions [coin toss], how many models are now out that are heads [warming] to tails [cooler]
     At a guess over the last 23 years it would have to be at least 64 ie 2 to power 64? enough to call a 0.0000000000000000000001 chance of the models being realistic.

     I have left a few zero’s out.

121. In order to determine the potential “weather” noise we are looking at when viewing the latitudinal zone trends for the CW Hybrid temperature series, I downloaded a GISS CMIP5 RCP4.5 model with 5 runs and plotted the zonal trends for those series along with the CW trends for the periods 1979-2012, 1979-1996 and 1997-2012. The GISS model that I selected has an ARMA modeled red/white noise structure very similar to the observed series when compared using the p.value for a Kolmogorov-Smirnov test (in R as ks.test).

     The results are in the links below and show that the CW latitudinal trends are well within the noise levels for the GISS model for the period 1979-1996. The 1979-2012 period shows the CW trends probably within the noise level for the GISS model although one might argue that the Arctic polar amplification is higher for the CW trends. The 1997-2012 comparison shows that the CW trends has a noticeably higher polar amplification than the models and the models noise range.

     I am thinking that another analysis might be required to determine how sensitive the CW Hybrid results are to starting date for the series. It should be remembered that the polar regions at high latitudes are encompassing much smaller areas per incremental latitude zone than at lower latitudes and thus we should expect to see more “weather” noise in the polar regions – providing there are no other effects operating. Steve McIntyre had a plot at CA of the HadCRU4 and CW 13 series showing a divergence not occurring until 2005.

     1979-2012:
     http://imageshack.com/a/img59/3664/du6n.png

     1979-1996:
     http://imageshack.com/a/img801/439/qzys.png

     1997-2012:
     http://imageshack.com/a/img18/3757/gh73.png

122. Christopher J Shaker (Comment #121336)-I’m sorry but that article is more than a little misleading. These statements:

     “The resulting temperature profiles depend on details of the methods that are used to obtain temperatures from radiances. As a result, different groups that have analyzed the satellite data have obtained different temperature trends. Among these groups are Remote Sensing Systems (RSS) and the University of Alabama in Huntsville (UAH).”

     Are false. UAH and RSS results differ not because obtaining temperatures from radiances is an ambiguous problem. The methods are not really different there as far as I am aware. The primary differences come about as a result of methodological choices of how to deal with the issues of drift and intercalibration and others. Moreover there is the trumpeting of several “alternative” datasets which are frankly not to be taken seriously and full of known errors.

123. “DocMartyn (Comment #121341)
     November 19th, 2013 at 6:50 am
     “Nick, You work with the records you have. But the averaging principles are the same”
     Are they? Are the times the temperatures taken, every six hours, the same in 2013 as they were in 1960?
     Do you actually record Tmax and Tmin during the measurement?
     If your reading are taken at a time that is near Tmin, and especially Tmax, a shift of sampling time will make take you closer or further from the trough/peak.
     I am informed by Mosher that the TOB is a real problem with many records. Taking measurements every six hours, or worse, intermittently, is something of a can of worms.

     ##############################

     Doc it depends on your data source.

     Here are all the various types you will encounter.

     1. 1 minute readings
     2. 5 minute
     3, 1 hour
     4. 4 hour
     5. tmin, tmax
     6. tave

     So some historical records only have Tave ( tmin+tmax /2). This is the common denominator. That is, every other type ( 1-5) can be
     transformed into 6.

     From 1800 (or so ) on you start to get daily records of both tmin and tmax.

     Hourly data starts around 1900 or so but only for a subset of data..

     I cant recal the start date for synoptic data. Data every minute
     starts probably with the onset of automated stations for airports.

     So, if we want to go as far back in time as possible we use tave monthly. If we want to look at tmin tmax separately then we can get back to 1815 or so ( cant recall) and we can look at daily.

     If we want to use hourly or synoptic then your talking going back to 1900 for maybe only as far back as 1940.

     Its quite a mess.. At some point We might push out a daily global product or at least a daily US product.

     Hourly is possible but calculating the uncertainty might take a year.

     right now uncertainty takes two weeks to calculate.. thats 40000 stations 12 times a year by 250 years.

     Us hourly might be 20K stations by 24 hours by 365 days by say 60 years.. err.. my year estimate might be wrong but you get the idea

124. Steven Mosher (Comment #121346)-Synoptic refers to a spatial scale, what do you mean using it to refer to timing of observations?

     The first synoptic maps were drawn in the 19th century but only after the fact. Real time maps started to be drawn by the Smithsonian in the late 1840s and were possible because of the telegraph (allowing near simultaneous observations).

125. Steve, what I am having problems with is the sampling rate changing during the record.

     Have a look at this image

     http://ahualoa.net/eggsntea/wp-content/uploads/2008/10/templog_080907.png

     Now if I were to measure temp every minute I get spikes as my Tmin and Tmax, if I take hourly averages, then Wed 10th gives a low Tmax, and the Tmin of Tues 9th is overestimated.
     The greater my sampling rate, the greater the Tmax-Tmin, but the higher the Tav.

     Now, how are SST taken in the Arctic NOW, compared with 20 years ago. Are here more samples per day now, than in the past?

126. I haven’t read the paper yet but found the critique of kriging across boundaries odd. Certainly the discontinuity will be smoothed but it should be a reasonably trend-neutral smooth.

     Maybe I need to read the paper first but the critique doesn’t make sense on first read.

127. Jeff, see if this explanation makes sense:

     It’s because land and ocean are driven by different mechanics. At the least you’d want to look at the correlation as you cross from land to ocean.

     It’s easy to see with this graphic that land has a substantial polar amplification effect, but (at least where there is data), the ocean does not. If you did the kriging correctly (don’t assume axial symmetry for example), probably this wouldn’t matter, but the assumption of axial symmetry is used here.

     So in using coastal stations in the high arctic to infer Arctic Ocean surface temperatures, you are running a very high risk of smearing the land temperatures into the open sea.

     I think this is what happens with GISTEMP in summertime (of course they don’t use kriging, but the same issue of equating ocean dynamics with those of land are at play), especially over Russia and Siberia. Typically there’s a big huge red spot in that open water in the summer time for their 2-d anomaly maps, and very large trends for this time period.

     If these regions of the Arctic Ocean follow the same trend as the rest of the ocean (no amplification), then “crossing the boundary” is a huge problem. In that case, it will lead to artifactual warming.

128. Carrick (Comment #121350)- Do you happen to have seasonal breakdowns of that plot?

     dallas (Comment #121316)- Interesting comments about mixed phase clouds. Have you read Choi et al on this area?

129. Andrew Fl, “dallas (Comment #121316)- Interesting comments about mixed phase clouds. Have you read Choi et al on this area?’

     No, I forget the name but that was from a thesis and follow paper by a UWM student. If you take the GISS 70N-90N 250km resolved data you can compare the sept data to the full year where available and see the Arctic Winter Warming amplification is greatest when Arctic trends shift to neutral or cooling. It is like a pressure relief valve opening allowing more heat to sink to space.

     It is actually “proof” of the “pause” instead of disproof.

130. dallas (Comment #121352)- I’m actually referring to some work in a couple of papers on observations of cloud phase, I don’t know that it relates to the issue of trends per se but they did note some key differences from how models treat clouds in those conditions:

     http://link.springer.com/article/10.1007/s00376-010-9175-3#page-1

     http://www.pnas.org/content/107/25/11211.short

131. Carrick, in SKS forum comments, Cowtan observed that CRUTEM overweighted coastal stations as follows: “No, I never really sorted out the GISS stuff. The thing I did crack was (inconveniently) the CRUTEM3 plot, which overweights the costal stations because they treat all coastal cells as 100% land.”

     It would be ironic if their revised methodology ended up doing something similar or worse.

132. Andrew FL, Thanks or the links. This is the paper that started it all for me.

     http://www.ssec.wisc.edu/library/turnerdissertation.pdf

     Since then I have been looking more at the impact on simple energy balances instead of the exact cause since there are so many it can make your head swim. The aerosol impact on droplet size is something that I think Troy Masters has published on as well.

     Dynamically, it seems to boil down to the type of inversion. In the Arctic/Antarctic, poleward advection can slide over, under or into the mixed phase layer varying the amount of energy retained plus the shortwave impact varies with atmospheric lensing, inversion altitude and time of year. AQUA attempts to model the anisotropic absorption, but it is still hard to tell what dominates when and where.

     Since the Arctic has a lower average altitude, it has a stronger advection effect and the Antarctic appears to have a stronger SW anisotropic effect, but with Antarctic sea ice increasing, there would be a increasing advection effect.

     Modeling what happens is going to be a bit of a challenge.

133. Carrick,

     I get that part but it seems to me that we would get pretty balanced trend smearing in both directions with little net effect on larger regional trends.

134. There is still no reasonable explanation as to why anyone should take yet another AGW promotional study relying on conclusions derived from a population of zero seriously.
     One of the signs of bad science is when a study relies on marginal data barely discernible and requiring a great deal of massaging to deliver the obviously desired results. And those results are well within reasonable error bars and not actually significant.
     Please explain how this study does not meet that criteria perfectly.

135. Jeff:

     I get that part but it seems to me that we would get pretty balanced trend smearing in both directions with little net effect on larger regional trends.

     There’s little or no data on the ocean side in the Arctic Ocean (as with much of the interior of the Antarctic), so the trend can only move one way.

136. @Hunter, I’m seeing a lot of knee-jerk reactions to this study. From all I hear about it, it doesn’t seem to be an AGW promotional study at all. It’s an attempt to get a reasonable data set where there is missing data. HadCrut had big holes in the Arctic regions, and made a certain choice on how to fill them in. This paper is suggesting maybe a better choice, that’s all.

137. Andrew_FL:

     Do you happen to have seasonal breakdowns of that plot?

     see this link.

138. Carrick,

     There’s little or no data on the ocean side in the Arctic Ocean (as with much of the interior of the Antarctic), so the trend can only move one way.

     This is, of course, the major difficulty or challenge with infilling for earth’s temperatures. It happens that major regions to be infilled are in some way “exceptional”. If the regions lacked data for some non-climate related reason (e.g. imagine a weird dictator refuses to permit measurements or something); the lack data because they are located in regions relatively inaccessible to humans because they have extreme climates.

     My ‘issue’ if one can call it that is not that I think there can’t be any problems. Of course there could or might be some. I can even think of some “toy” problems that could take application of kriging to an extreme and show where it goes wrong– they would be ‘reductum ad absurdum’ applications and would be useful if anyone was claiming that kriging always works and can never have a problem. Or if they were even claiming that the validations done are 100% convincing that the method must work in this instance. (Doing the toy problem would be a lot of work to set up and do, and I think you, I, Jeff and a number of people who have a general understanding of what interpolation or extrapolation in general can or can’t do don’t need to see the application. We already know them!)

     I guess the way I see it more as: Is this method somehow obviously worse that other methods? Judy’s use of “physically unrealistic” sounds so strong. I realize there are discontinuities and changes. But C&W isn’t quite like taking measurements outside in dirt air outside my heated house and then “kriging” to estimates temperature inside my house.

     Though, of course, to some extent, the discontinuity of sea/land and extreme north shares the ‘problem’. It just seems to me that it’s an issue that needs quantification. The method can’t just be dismissed as obviously worse out of hand.

139. Carrick (Comment #121360)-Thanks.

     dallas (Comment #121355)-Interesting. I’ll give it a look, thanks.

140. MikeR,
     I think a more reasoned approach is to go measure the areas with missing data.
     Synthesizing data sets with sample sizes of zero, and then claiming conclusions out of those pseudo data sets is creative, and the work may involve scientific techniques, but the results are art, not science.

141. According to C&W figure 2, the potential coverage bias in HadCRUT4 relative to UAH was only -0.02 K in 2010. How then can the adjusted series be more like 0.05 K higher than HadCRUT4? [Per this figure at CA.]

     Again, apologies in advance if this is covered in the paper.

142. Hunter, I’m all in favor of better data: set up more thermometers in the Arctic. But we’ll still be left with spots in the Arctic that have data now and didn’t in 2005 and 2010. How do we figure out trends? We have no choice but to try to use the 2005 or 2010 data to get some kind of overall estimate, to compare it with what we have from today. Which estimate you get depends on how you “fill in” gaps. It’s not like you have a choice: HadCrut has to be using some method now.

143. I compared the seasonal Arctic polar amplification for the CW Hybrid gridded temperature anomalies by selecting the months Jan, Feb, Mar and Apr for maximum ice Arctic sea extent and the months Jul, Aug, Sep and Oct for minimum sea ice extent. I also included the Arctic polar amplification using all months. For the amplifications I used the latitudinal zones 60N-90N (AA)and 60S-60N (60SN). I also include below a link to a graph of the Arctic ice extent from 1979 to mid 2010.

     The link comparing all months for polar amplification shows the temperature anomalies in the latitudinal zone 60SN rising into the late 2000s and then leveling off to 2012. The AA zone shows a rate of warming faster that the 60SN zone as one might expect from a polar amplification from 1979 into the 2000s, but then as the 60SN zone plateaus in warming the AA zone warms at a faster rate from the early 200s to 2012 than it previously had from 1979 to the early 2000s. Observing these warming rates makes me wonder which zones are driving which in the Arctic polar amplification and whether we are observing somewhat independent phenomenon driving temperatures in these 2 global regions.

     I should immediately point out that on observing the graphs for the maximum and minimum Arctic sea ice extent from the CW hybrid temperature series that the y axis are to different scales. The amplification for months with the maximum sea ice extent has a greater polar amplification than that for the months with the minimum – based on the AA warming rates. I would have, from a simple minded view of this phenomena, expected that since the sea ice minimum extent is decreasing at a faster rate than the maximum that one might expect to see opposite seasonal effect. The rate of warming for the months with maximum sea ice appears to be constantly upward over the 1979-2012 period while that for months with the minimum sea ice extent the warming can be broken into two period by a breakpoint calculated to be around 1991 with the period after 1991 showing warming at a faster rate than before.

     I do not know whether these amplifications observed from the Cowtan and Way hybrid gridded temperatures can be used to test the validity of their series by comparing those observation to what is expected or whether, if we take the series as not only valid but perhaps unique in what we know about temperature trends in the Arctic, we now might change or modify our views of what is driving the Arctic polar amplification.

     http://en.wikipedia.org/wiki/File:ArcticSeaIceExtents.jpg

     Monthly All Months:
     http://imageshack.com/a/img843/443/pw98.png

     Minimum Ice Extent:
     http://imageshack.com/a/img199/2543/2b40.png

     Maximum Ice Extent:
     http://imageshack.com/a/img546/74/0gvm.png

144. MikeR (Comment #121366)-You can’t really set up stations in the middle of an ocean, even if there is ice over it.

     Kenneth Fritsch (Comment #121367)-Keep in mind that the ice minimum is a couple months after the temperature maximum. I’ve looked at the reanalysis data-which does have problems, to be sure-and found that there is little change in warmest days in the Arctic, especially compared to the coldest days. It would be interesting to see if that finding could be supported with other evidence.

145. Andrew_FL (Comment #121368)
     November 20th, 2013 at 11:59 am

     I was thinking in terms of maximum and minimum albedo which I have heard stated is a major factor in producing the Arctic polar amplification.

146. Thanks Carrick, I see what you mean about regions with no data. I also agree with Lucia’s point that the Judith’s comment seemed to be a bit stronger than my understanding of the problem.

147. Skeptical Science team members, “Kevin C” Cowtan and Robert Way shouldn’t be taken seriously after what we have witnessed from other team members.

     http://www.skepticalscience.com/team.php

148. Mike R,
     Again, this comes to activists using sciencey techniques to squeeze the data until they get something that they can claim does not look like a pause.
     Apples-to-apples, there is a pause. Apples compared with oranges, after mixed properly, and viola! no more pause.
     As I said in my first comment on this paper: sincere or not, it is post hoc excuse making. At best, this paper is demonstrating confirmation bias.

149. Kenneth Fritsch (Comment #121369)-Fair point. I think the evidence points more towards enhanced heat transport, personally, but that’s just me I guess.

150. Hunter,

     The best metric to use would be to actually judge the quality of the science. Yes, confirmation bias is potentially a problem, which is why it’s important to carefully and extensively validate new methods. But if the new method stands up to those tests, and it patches a hole in the data, then it really can’t be just called a result of confirmation bias. That’d be rejecting new science just because you don’t like the results.

     The alternative, of course, is to note how the new method affects the 1979-2013 trend very little, and you might reason that this should also eventually hold for the 1997-__ trend, so you can wait a while to see if the “pause” goes away by itself, even without the new adjustments.
     This new method is a patch for a short-term temp series problem (~15-20 years), over longer periods the heat will diffuse sufficiently that it’s irrelevant. Point being, if you want to wait another 10-20 years, the pause should go away anyway, if this method and the predictions are both correct.

151. HaroldW (Comment #121365)
     November 20th, 2013 at 10:43 am

     As far as I can tell in Figure 2 CW is referencing UAH to HadCRU4. CA is referencing CW13 to HadCRU4. Do we know what the UAH to CW13 coverage bias is? Is it 0.03? Or am I misinterpreting what coverage bias is?

152. Andrew, from Wikipedia I have excerpted two different mechanisms for the Arctic polar amplification that could work together but independently. The first is the albedo feedback which could be a “weather” noise factor operating over decades and specific to the Arctic due to changing weather patterns or it could require a warming globe to initiate albedo feedback. My analysis of the late half of the CW analysis would indicate that with the observed accelerated warming in the Arctic occurring in conjunction with a global warming that has “paused” we might think of the current Arctic warming as “weather” noise as I described it above.

     The second mechanism relies on tropical thunderstorms transporting heat to the Arctic and here you could relate warming in the tropics to more intense and/or frequent thunder storms there and thus global warming leading to
     faster warming in the Arctic. That might be what we would see from the observed rates of 60S-60N warming with faster warming of 60N-90N in the early half of the CW hybrid series.

     “Why are temperatures warming faster in the Arctic than the rest of the world? The loss of sea ice is one of the most cited reasons. When bright and reflective ice melts, it gives way to a darker ocean; this amplifies the warming trend because the ocean surface absorbs more heat from the Sun than the surface of snow and ice. In more technical terms, losing sea ice reduces Earth’s albedo: the lower the albedo, the more a surface absorbs heat from sunlight rather than reflecting it back to space.

     However, other factors contribute as well, explained Anthony Del Genio, a climatologist from NASA’s Goddard Institute for Space Studies. Thunderstorms, for instance, are much more likely to occur in the tropics than the higher latitudes. The storms transport heat from the surface to higher levels of the atmosphere, where global wind patterns sweep it toward higher latitudes. The abundance of thunderstorms creates a near-constant flow of heat away from the tropics, a process that dampens warming near the equator and contributes to Arctic amplification. From NASA, Earth Observatory, written by Adam Voiland.”

     http://en.wikipedia.org/wiki/Polar_amplification

153. MikeR,
     If you haven’t done a thorough read through of the comments on the latest CA post, maybe you’re missing a few data points that you haven’t heard. I certainly was.

154. Kenneth Fritsch (Comment #121376)-Yeah, the second is more or less what I have in mind (I would emphasize also the role of spatial heterogeneous radiative forcing in causing this). Mind you, albedo obviously plays some role. It’s just I feel the emphasis on albedo ignores important dynamics.

155. Kenneth,

     As pointed out earlier in this thread, during the winter months, the sun angle is so low, if it’s not actually below the horizon, that the difference in reflectivity between open water and sea ice is probably insignificant.

156. Dewitt Payne, “As pointed out earlier in this thread, during the winter months, the sun angle is so low, if it’s not actually below the horizon, that the difference in reflectivity between open water and sea ice is probably insignificant.”

     It can also impact SW absorption in the mixed phase clouds which is the reason I am not a fan of blending/smearing temperatures vertically. If you have a solid true surface temperature you can compare that with mid-tropo and stratosphere temperatures. Blend it all together and you have mush.

157. Look at the Arctic Stratosphere to Surface temperature profile over the past year. To the extent that UAH temps get into the higher troposphere and stratosphere levels, it really just becomes a big mess that has little to do with surface temperatures.

     Note there was a moderate Sudden Stratospheric Warming event in mid-January. About 2 out of 3 years have at least one of these. (While there have only been 2 total in Antarctica).

     Also note that the Surface and the lower Troposphere was mostly colder this year.

     http://www.cpc.ncep.noaa.gov/products/stratosphere/strat-trop/gif_files/time_pres_TEMP_ANOM_ALL_NH_2013.gif

     Back to 1979 here.

     http://www.cpc.ncep.noaa.gov/products/stratosphere/strat-trop/

158. Lots of interesting comments on possible physical mechanisms for doubting the C&W methodology. Can someone explain why these mechanisms dont show up as problems in the cross-validations that C&W use? Can someone show a cross-validation demonstrating an issue with the assumptions made?

159. Bill, given that C&W recalculate monthly, annual or seasonal variations should affect the method. Only that the deltaT is spatially smooth and relatively stable monthly. Cross-validation suggests these assumptions reasonably valid.

160. Phil Scadden:

     Can someone explain why these mechanisms dont show up as problems in the cross-validations that C&W use?

     It gets addressed here. For reference, I’m copying the comment from CA:

     Thinking Scientist:

     Robert – the cross-validation in the paper appears to only test for infilling bands between available data points, this is not the same as cross-validation where data coverage is one-sided relative to the latitude band ie extrpolation at the poles. Your xval will be optimistic.

     This is my worry too.

     Here you only have land stations and you’re using them to infill over open sea (think Arctic Ocean above Russia and Siberia). And because the land-based polar amplification is a maximum for these regions, any errors associated this “crossing the boundary” assumption will be maximized.

     Here is the coverage for October of this month, shown just for sake of illustration.

     Instead of no data in the Arctic cells, they have “virtually no data”.

     Can someone show a cross-validation demonstrating an issue with the assumptions made?

     Why do you assume that it has to fail before there is a problem?

     If you have very little data, the error bars are large, and just about anything you generate is going to pass cross-validation testing.

161. Lucia, I think we can break this down into two questions:

     1) Is the model physically realistic? The answer is certainly not. It falls far short of physically realistic. Then again neither is a harmonic oscillator, but see how often that gets used successfully

     2) Are there issues with the model that need to be addressed that could lead to biases in trends? I would say definitely.

     My attitude is that it is the responsibility of the authors to address these questions, dot the i’s, cross the t’s, but also recognize you don’t need a perfect contribution to have a useful or even a very good one.

162. Kenneth,
     What the thunderstorm study seems to imply is that so-called Arctic amplification is an intrinsic property of the weather mechanisms that dominate the Arctic.

163. Lucia:

     Though, of course, to some extent, the discontinuity of sea/land and extreme north shares the ‘problem’. It just seems to me that it’s an issue that needs quantification. The method can’t just be dismissed as obviously worse out of hand.

     Seems to me you have it backwards. The method *can* be dismissed as obviously worse out of hand. The quantification that’s needed is “how much worse?”

     It may be insignificantly worse or it may be so much worse that it’s meaningless, but I think the burden is on you — if you think it could be better — to show how a statistical technique with various assumptions about correlation and continuity is applicable across areas that are very different. (At the north pole, it’s not just across the land/water boundary but also includes land/water, land/ice, and land/water-over-ice boundaries which change on a seasonal basis.)

164. DaveJR, probably you didn’t mean the _very latest_ Climate Audit post. But I’ve never seen anything like it.

165. Re: hunter (Nov 21 07:29),

     Arctic amplification is an intrinsic property of the weather mechanisms that dominate the Arctic.

     Of course it is. Nobody who actually knows anything ever said it wasn’t. The clearest examples are the glacial/interglacial transitions. Temperatures at the poles change much more than temperatures at the equator.

     However, there is still the problem of how much of the recent warming in the Arctic is from quasi-cyclic behavior as exemplified by the AMO index. It got very warm at high northern latitudes in the early twentieth century when the AMO index was last in its increasing phase. At Svalbard, it was like a switch was thrown in about 1918.

166. Here are the official sea surface temperatures for the Arctic from the NOAA OiSST V2 dataset. [Not the anomaly, but the actual temperatures].

     Notice how they are physically explainable (as in it is consistent with sea ice conditions etc and yes there is some warming also consistent with sea ice conditions) and it is not really higher than Hadcrut4′s trends so there is no reason to up its trends by infilling some 1.5C increase.

     ie. it is already properly accounted for.

     And please show me where an extra 1.5C from Cowtan and Way 2013 can possibly show up without melting all the sea ice.

     http://s12.postimg.org/dcxwvcgh9/Oi_SST_V2_Arctic_SSTs_Oct2013.png

167. Phil Scadden

     Can someone explain why these mechanisms dont show up as problems in the cross-validations that C&W use? Can someone show a cross-validation demonstrating an issue with the assumptions made?

     I’m pretty sure I could come up with a hypothetical ‘toy’ model in which the cross-validations would appear to work but the kriging result would be wrong and possibly worse that omitting data. The hypotheticals to show problems that might arise tend to be extreme reductum ab adsurdum, but they show what can go wrong when assumptions are violated.

     I’m going up to Highland Park to deal with popsie wopsie today, so it might be a while though. (My sister moved from Highland Park to Chicago itself and we are moving Dad to a new home. So there are ‘move’ activities involved. )

168. DeWitt Payne (Comment #121379)
     November 20th, 2013 at 5:12 pm

     DeWitt, thanks for reminding me – no sun no albedo. I think that shifts the albedo effect entirely towards the season of all sun and maximum open seas and means maximum ice extent is not important in my considerations of a local Arctic amplification and albedo feedback.

     Minimum ice extent in the Arctic appears to have a visual break (I have not made a breakpoint calculation) around 1997. I might want look at different months than I have used for looking the maximum seasonal polar amplification due to albedo changes.

     I am currently doing what I did with CW Hybrid with the GISS gridded temperatures and looking for an increasing Arctic polar amplification during the global temperature pause.

169. Kenneth Fritsch –
     In considering the effect of Arctic albedo changes, one must consider, in addition to sea-ice cover, both insolation and cloud cover. See Hudson, Estimating the Global Radiative Impact of the Sea-Ice-Albedo Feedback in the Arctic.

170. “I’m going up to Highland Park to deal with popsie wopsie today, so it might be a while though. (My sister moved from Highland Park to Chicago itself and we are moving Dad to a new home. So there are ‘move’ activities involved. )”

     Good luck, Lucia and I hope your father is amenable to this move. I went through similar situations with my father and it truly tested my negotiating skills with all parties involved. It certainly helps to have siblings who are in agreement.

171. MikeR wrote: “DaveJR, probably you didn’t mean the _very latest_ Climate Audit post. But I’ve never seen anything like it.”
     .
     Yes, I didn’t know there was a new post at the time. I remembered reading one of Lucia’s comments suggesting that there was no apparent agenda behind the paper. I no longer believe that is a fair assumption. While this in no way affects its quality, IMO, I feel it requires a more cynical examination of the various decisions/assumptions made.

172. HaroldW (Comment #121393)
     November 21st, 2013 at 11:22 am

     Harold thanks for the link to that paper. It appears to put more uncertainty into the albedo feedback mechanism or at least clouds the issue.

     I am searching for what the greater Arctic polar amplification seen in Cowtan/Way (and given that it is valid and reveals an Arctic warming not previously used in these studies) means in light of the global warming pause. I think it might give further credence to the claims of the recent Arctic warming being more regional and even cyclical.

     I am a bit surprised that those in more or less acceptance of CWs findings are not pushing the “weather” noise issue as they might do for the warming pause or that I have not heard that CW pointed to weather noise in their 1997-2012 analysis and the potential ramifications for the mechanism for the recent Arctic warming. Maybe they did since I have yet to breakdown and purchase a copy of the paper.

     My strategy was if sufficient questions about the paper were posed and replies included excerpts from the paper I would have a comprehensive view of what it said – at least about the issues I see as important.

173. Carrick, if your statement is correct, then wouldnt it follow that if you replaced UAH gridded data with another relatively smooth set (say albedo of Mars), then cross-validation would work? Frankly I’m skeptical. Demonstrating a flaw in the cross-validation methodology would be definitely worth publishing though.

174. Kenneth

     Good luck, Lucia and I hope your father is amenable to this move. I went through similar situations with my father and it truly tested my negotiating skills with all parties involved. It certainly helps to have siblings who are in agreement.

     It’s going to go fine. Dad is pretty easy going and he trusts Mary Beth and me to be doing whatever is best. The most difficult period will be that the car drive is about 40 minutes and he will get restless. That should happen in about a month. Otherwise, there are just a bunch of “to dos”. Mary Beth is the one doing most of everything. But there are key periods where either someone substitutes or where we need more than one person. She might have been able to get Dad to the doctor by herself, but it was much easier to have one person getting the car in place while the other waited with Dad and so on. Dad’s now seen the MD, and the nurse injected him with the TB test stuff. The test needs to be read in two days, then a few more things need to happen. Dad will probably move in a month.

175. I have now compared the GISS Infilled series, that I obtained from the CW website, with the CW Hybrid series for the latitude zones of the Arctic 60N-90N (AA) and 60S-60N (60SN) for all months and for the months July, August, September and October. GISS is infilled but continues to have some grid areas with no data.

     The graphs are linked and show that the major difference in the graphs between CW Hybrid and GISS Infilled is that the GISS series shows a definite bending of the AA temperature anomalies with some lag with the global warming pause, while the CW Hybrid series is tending with perhaps an ever so slight bend unabatedly upward in the global warming pause time period. In other words the GISS series show an apparent reaction somewhat in kind to the global warming pause in the AA while the CW Hybrid series do not.

     All months:
     http://imageshack.com/a/img9/7859/zxyn.png
     

     July, August, September and October:
     http://imageshack.com/a/img707/6980/b0tk.png
     

176. hunter (Comment #121386)-“What the thunderstorm study seems to imply is that so-called Arctic amplification is an intrinsic property of the weather mechanisms that dominate the Arctic.”

     Actually, you can really see just how true this is, from the data. Richard Lindzen mentioned in alot of talks he has given that the temperatures in the Arctic during the summer months are mostly determined by local radiation balance-and hence they vary very little. But during the winter, storms from further south erratically penetrate the Artic and when this happens there are brief and sudden swings in temperature. You can see this effect here:

     http://ocean.dmi.dk/arctic/meant80n.uk.php

     Where there are large swings of temperature in the colder parts of the year. It’s actually a really cool illustration of the importance of these dynamic factors.

177. The rerference to Arctic amplification is interesting because the AGW narrative has implied that Arctic amplification is due to human caused CO2.

178. hunter (Comment #121401)-Yeah, it’s pretty misleading. The main mechanism they cite would act regardless of the source of the warming.

179. Huh. I’ve always heard about Arctic amplification as simply a result of temperature change, as the main feedback is due to melting (or growing) ice, which should act regardless of where the temp change comes from.

     It’s similar to the tropospheric hotspot in that regard: the source of the warming is unimportant.

180. Windchaser (Comment #121403)-There have been claims recklessly bandied about of Arctic warming as “fingerprint” of CO2 warming. I’m sure people who actually knew what they were talking about and not trying to mislead people were consistent in saying it was an indicator of warming regardless of the cause. Confusion may have come about because the warming itself is being attributed to man, those same people would indeed assert that it was caused by CO2, and people who don’t know what they are talking about would probably misunderstand that.

     Ah, the iron triangle/rice bowl.

181. “Kenneth Fritsch (Comment #121399)
     November 21st, 2013 at 2:35 pm ”

     Were you to compare only the grid points overlapping between the infilled GISS and original Hadley in the Arctic you would find that Hadley’s points are warming faster than GISS because of the 1200 kmsmoothing GISS applies.

     As an example since 1979 the Hadley cells overlapping with GISS (north of 60N) warmed at a rate of 0.504°C/decade whereas those same set of cells for GISS warmed at 0.438°C/decade. So then as you can imagine if you are using the observations to estimate other regions (with satellite guidance) you will have a higher trend using the hadley data as opposed to the 1200 km smoothed GISS data.

     There is the question as well of whether the 1200 km smoothing is appropriate during the summer in the Arctic which our results and those of Rigor et al (2000) seem to indicate is too large a radius during those seasons.

182. As an example for comparison with UAH the trend over that same period for overlapping cells is 0.450.

183. Smoothing makes for prettier maps though.

     Anyway I think there are good reasons to think the same radius should not apply everywhere-in fact it should be inversely proportional to the sine of the latitude. Let’s see if people can guess why.

     (well, I should also note it is kind of dubious in the boundary layer period, but that is kind of giving people another clue, isn’t it?)

184. Have commented before but the total sea ice area for year 2013 is going to be above average for a WHOLE YEAR very shortly. This will put a big crimp in any “Kriging” by Way,Cowtan et al.
     With the extent positive any Arctic amplification will be wiped out by Antarctic Deamplification and there will be a Hockey stick spike down in global cooling for 2013. Cannot wait.

     Comments on the arctic[Antarctic]
     The temperature should follow a seasonal pattern based on solar input [and lack thereof in winter].
     The anomalous warming of the Arctic[Antarctic] comes purely by changes in the temperature of the sea currents feeding into the Arctic and should fluctuate around a mean which means based on past history we will be due for anomalous cooling in the very near future.
     The extra heat is then dissipated by the melting of the ice which prevents any serious rise in arctic temperature.
     Virtually the whole Arctic ice cap would have to melt before any sustained rise in arctic temperature could occur.

185. Actually, that makes me wonder what would happen if you substituted the interpolation over Antarctica for O’Donnell et al’s efforts over the continent. HM….

186. For those of us who do not have access to the paper, does it discuss why they take the approach of combining surface and satellite measurements instead of using only satellite measurements?

     It is easy to be suspicious of methodologies which mix data from different measurement devices

     Thank you,
     Chris Shaker

187. @angech (Comment #121408),
     Since amplification is an intrinsic characteristic of the Arctic, the warming is not anomalous by definition.

188. hunter agree but the anomalies posited by Way have been ascribed universally to [anomalous] arctic amplification

189. If arctic sea ice loss is attributed to warming temperatures and given as a proof of arctic amplification, why is it then that almost all of the loss is consistently concentrated to the area north of Novaya Zemlya between Novaya Zemlya and Spitzbergen? If it’s GLOBAL warming should the ice loss then not be kind of equally distributed all around the arctic area? I’ve been following this trend for a few years now and this fall we are again headed into the same situation (look at the daily values)
     http://nsidc.org/data/seaice_index/
     I’ve asked it a couple of times at different fora but have not got any response.

190. Re: Sven (Nov 22 08:50),

     Of course it’s not all global warming. There was rapid warming and ice loss in the same region in the early twentieth century starting in 1918. That’s why it became practicable to mine coal for export on Spitsbergen. IMO, it’s largely a northward shift or increase in flow of the North Atlantic Current. Assuming that the AMO index has peaked and is now starting to head downward, we could see some recovery in Arctic ice volume over the next 20 to 30 years.

191. The area north of Norway, the Kola Peninsula (the Barents) and the Pechora Sea is bathed by warm waters which originate far to the South and are part of the Gulf Stream flow warming Europe. The other pathways where the Arctic Ocean waters are connected to the Atlantic and Pacific oceans are fairly narrow, choked with ice, and tend to have surface currents flowing south.

     I would love to have a climate model run when the Atlanic was smaller, say in the Paleocene. I bet sea currents were quite different.

192. Not part of the Arctic of course, but the other bizarre thing about the NSIDC map is that it shows the gulf of Finland (and even parts of the gulf of Livonia) already completely covered with ice when there is, of course, not yet the tiniest glimpse of ice yet. NSIDC is consistently wrong with gulf of Finland usually showing ice cover as late as June when people are already swimming in the sea. I think they really should check this oddity out in order not to lose credibility.

193. Sven,
     Nearly everything used by the AGW promotion industry lost credibility long ago.
     Only now is the AGW community learning that soot causes ice to melt, and are still not widely acknowledging that a great deal of ice melt occurs due to dirty smokestack emissions and dirty diesel transportation fuels.
     If people considered the opportunity costs of the CO2 obsession, and how a small fraction of that money could have done something useful, like reduce soot and other particulates, they would be very unhappy.
     As to Arctic anomalies, anomalous to what? Cowtan and Way set out to erase the pause like Mann set out to get rid of the MWP. The anomaly is that the science community puts up with this tripe.

194. If the GISS in-filled extrapolation into the Arctic underestimates the warming there and if indeed the CW hybrid is the more valid view of that warming, a comparison, using the CW hybrid, of the latitudinal zone of 60S-60N temperature anomaly changes and recent pause there to an ever increasing warming trend in the Arctic would point, in my view, to the “weather” noise in the Arctic being mostly independent of the “weather” noise globally – at least for the 1997-2012 period.

     Obviously the extra heat from AGW could be transported to the Arctic from the mid latitudes and in that case the lesser warming in the 60S60N and even perhaps the 60S90S zones could be offset globally by the greater warming in the 60N90N zone. Given the validity of CW Hybrid series and global temperature trends generated from it, we are still looking at a substantial global warming pause during the 1997-2012 period which would mean that not much of the extra 90S-60N heating is transported to 60N-90N.

     From the discussion I see on this thread concerning the Arctic polar amplification it is not entirely clear to me whether it can be mainly a locally and reoccurring phenomena unrelated to AGW or whether it requires mainly or partly the extra heat from global warming in the mid latitudes transported to the polar regions. It sound likes the former mechanism is favored in the discussion on this thread.

     Since the favorable comparison in Cowtan and Way of the satellite/kriging to kriging alone methods of estimating missing temperature data would indicate that kriging alone methods would allow that methodology to be applied going further back in time than 1979 and thus an examination of 60S60N warming/cooling and perhaps 90S-60S to that for 60N-90N over long time periods in attempts to find any warming/cooling cycles that might occur.

     I would suppose the Antarctica would be a problem prior to 1956 since we have no temperature records that I am aware of prior to that time. It would be interesting though to compare the Antarctica to the Arctic and look at what I believe DeWitt has referenced as an anti opposed cyclical warming/cooling phenomena between the poles. I know there are also ice core proxies for temperature that Steig reported on in the Antarctica that indicated a warming period there, as I recall, in the 1930s.

195. Christopher J Shaker (Comment #121410)
     November 21st, 2013 at 9:29 pm

     Because satellite measurements are not surface measurements. The trending of satellite and surface temperatures are not supposed to be the same necessarily and can vary with the region of the globe. CW use the satellite data only for spatial reference and reference to surface grid cells and not for temporal references much like AVHRR was used by O’Donnell et al in the Antarctica.

196. Kenneth Fritsch (Comment #121419) -Well anything that is adding heat to the tropics that is counteracted by moving that heat toward the poles, not just greenhouse forcing.

     *Given* that the heat is ultimately from GHG forcing, it follows that the amplification is “caused” indirectly by that. *If* something else has caused heat to be added to the tropics (say, a circulation induced change in cloud cover) and that heat is mostly being transported to the poles, that “something else” is responsible for the Arctic amplification.

     Both are basically regional scale feedback mechanisms. In the case of the transport mechanism, it acts effectively like a negative feedback in the tropics and positive at the poles.

197. Re: Andrew_FL (Nov 22 13:45),

     In the case of the transport mechanism, it acts effectively like a negative feedback in the tropics and positive at the poles.

     Precisely. It already does that. If you look at this plot of radiative flux emission and absorption of solar radiation vs. sine(latitude) to correct for area, you see that between ~40S to ~40N (-0.6 <sin(θ)<0.6) there is excess absorption compared to emission and vice versa at high latitudes. If you increase temperature, the emission curve should get flatter for two reasons. One, the albedo at high latitude decreases leading to higher absorption and two, heat transfer from the equator to higher latitudes increases. Because there is a continent surrounded by ocean at the South Pole and an ocean surrounded by continents at the North Pole, the effect will be much larger at northern high latitudes than southern. That, in a nutshell, is what polar amplification is all about.

198. I have felt for a while that one needs to understand the mechanisms whereby heat is transported from the equator to the poles, and how those mechanisms can be altered, to explain Budyko-Izreal curve, not the other way around. There seems to have been a great strengthening of that heat flow during the Eocene, for explain, even compared to what one would expect from other cases.

199. Andrew FL,
     I think a reasonable question is how important is energy transportation to the poles in the first place?
     Polar transport is mostly convection. The real issue is how much energy is radiated in and out. Convection is mostly rearranging deck chairs.

200. DeWitt and Andrew, I have a general understanding of the mechanism of the heat transport to the poles but not in sufficient detail to understand what the Cowtan and Way temperatures series means in terms of that transport with the two time periods from 1979-1996 and 1997-2012 differing in the warming rates between the mid latitudes and the Arctic polar region.

     In the early period I see a general global warming including the mid latitudes and a corresponding, but faster warming in Arctic polar region. In the later period the mid latitude warming, and to a somewhat lesser extent the global warming, has reached a plateau while the Arctic polar region continues to warm at an even a faster rate. The CW Hybrid shows a much exaggerated view of this relative warming for these two periods than other temperature series do.

     I have some ideas on how one might attempt to explain these warming trends – given that the CW hybrid is the providing the more correct polar Arctic temperatures- but I would be interested in what the both of you might propose.

     It is difficult for me to conceive a situation where the transport of heat to the Arctic polar region would over a short period of time go from transporting some of the extra mid latitude heat to the Arctic polar region and then go to transporting all the extra there.

     Could it be that the pause in global warming, as indicated by the mid latitudes near complete warming pause, is effected by some other mechanism independent of the transport mechanism and that given the previous heating and warming in the Arctic polar region (from heat transported from a previous warming mid latitudes) that region continues to warm, and even at an faster rate, due to albedo feedback?

201. hunter (Comment #121456)-Can’t quite agree. For one thing, since the infrared opacity will be reduced in drier air, and humidity drops off significantly as you move poleward, a movement of heat around can also alter the flow of heat in and out. Think like taking a chair from below deck, now there is a much greater probability that chair can get tossed overboard.

     I also have a chart somewhere of what the temperature distribution would be without transport. It’s about double the present range from equator to pole.

202. Kenneth Fritsch (Comment #121459)-I’ll let DeWitt answer for himself, but I am inclined to agree an increase in the efficiency of transport suddenly occurring seems a little improbable. It would be nice to explicitly examine local radiation “balance” (or rather imbalance) and implied transport free temperatures for comparison to an absolute temperature field. Which us the other thing, we’d need absolutes not anomalies.

203. re:hunter (Comment #121456)
     November 23rd, 2013 at 10:55 am
     Hi Hunter,
     How meridonial transport changes may turn out to be critical to long-term (multicentennial) temperature gain. The problem is that we really don’t know. I have a working 2-D meridonial model which I play with when my wife isn’t watching. All of the AOGCMs predict an increase in meridonial heat transport towards the poles. The strength of this effect appears to be the critical element which controls in the model the magnitude of polar amplification and how much the relationship between net flux and average temperature moves away from simple linear for an assumed constant forcing. This in turn determines how effective climate sensitivity changes with time, or if you prefer, it determines the magnitude of the ratio between the Equilibrium Climate Sensitivity and the Transient Climate Response for that model. In a nutshell, the issue may not seem germane to prediction of temperatures by 2050, say, but it may be highly pertinent to predicting temperatures by 2200. It also tells us that we should never, never, never take an ECS from an AOGCM and use it as though it was derived under an assumption of a constant linear feedback.
     This is a model phenomenon. Is it a realworld phenomenon? I still don’t know.

204. Re: Kenneth Fritsch (Nov 23 11:33),

     In the later period the mid latitude warming, and to a somewhat lesser extent the global warming, has reached a plateau while the Arctic polar region continues to warm at an even a faster rate.

     That’s not quite true. Currently, it’s not continuing. Looking at UAH lower troposphere NoPol anomalies, there was little warming at all from 1979-1994. Things pick up in 1995, but then in 2003 things slowed down again. In fact, the EWMA filtered NoPol anomaly is currently at 2003 levels. I believe the geographic coverage of UAH NoPol is 60-82.5N latitude.

     This, by the way, is exactly what you would expect if most of the increase in the NoPol anomaly is proportional to the AMO index, which should be getting ready to head down.

205. Re:Andrew_FL (Comment #121461)
     November 23rd, 2013 at 11:45 am

     It would be nice to explicitly examine local radiation “balance” (or rather imbalance) and implied transport free temperatures for comparison to an absolute temperature field. Which us the other thing, we’d need absolutes not anomalies.

     I’m not quite sure what you are asking for here. There is already a substantial meridonial heat flux, the net component of which is a large and critical element in any local heat balance. It can’t be “disappeared” even if you want to assume that the perturbation to that transport system is small. Absolute transport-free temperatures would be a long way from anywhere.

     The paper below gives some idea of how the GCMs balance the system.

     http://www.atmos.washington.edu/~dennis/Zelinka&Hartmann_2012.pdf

206. Thanks to Paul_K and Andrew_FL
     At least I qualified my observation with “mostly”…. It makes sense that the heat transported to the poles is going to radiate out more easily, between albedo, atmospheric optics and low heating and long cooling cycles.

207. Paul_K (Comment #121469)-I wasn’t clear, I meant one should calculated the change in temperature from one state to another assuming no change in transport. The difference between this delta and the observed delta is the change due to transport.

     Or one could ask how the heat content of the boundary layer in the area changed and how much the vertical heat flow changed. The difference is the change in horizontal heat flow.

208. DeWitt Payne (Comment #121465)
     November 23rd, 2013 at 12:33 pm

     DeWitt, what you describe for UAH for the warming of the polar Arctic region over the period 1979-2012 would appear to put the GISS series somewhere in between UAH and CW Hybrid. I’ll need to compare all three of these series together. Do you think the lower troposphere temperature is well correlated with that of the surface temporally in the Arctic polar region?

209. Re: Kenneth Fritsch (Nov 23 17:18),

     Do you think the lower troposphere temperature is well correlated with that of the surface temporally in the Arctic polar region?

     I think that should be the working assumption until falsified.

210. My limited experience that I can recall from examining USHCN versus UAH was that atmospheric and surface temperatures over limited regions could be very poorly correlated over monthly timescales (I think, it was a long time ago) but the correlation was much better on annual timescales. Of course, that was over land. I have no clue about over ice, although I kind of suspect over open water the correlation on monthly timescales would be much better (than land).

     I got into trouble with that analysis because it suggested USHCN was *underestimating* warming in the US. I almost got my skeptic cared revoked. Pretty crazy!

211. I will the post the spline smoothed anomalies for the CW Hybrid, GISS and UAH temperature series for the latitudinal zones from 90N to the equator in 5 degree increments for the 1979-2012 time period. I’ll link to those graphs in three separate posts so as not to over burden each post. I do this in an effort to look for differences between these series with reference to the Arctic polar amplification and determine whether the newer series, CW Hybrid, makes better sense or shows a problem in that regard.

     All three series show progressively flatter anomaly trends and a more definite flattening in the latter half of the time series when going to lower latitudinal zones. That bending and flattening starts to occur at higher latitudinal zones with GISS and less with UAH and lesser still with CW. The zones from 70N to 90N are definitely different between the three series with CW showing the greatest trend, GISS a lesser trend and UAH a substantially lesser trend than either of the other two series and with a trend that starts later in the time period. The GISS series Arctic polar region of 60N-90N appears more in phase with less lag, but a lag nonetheless, than the either the CW or UAH series do with the warming and the leveling of that warming in the mid and lower latitudes. That could, of course, be the result of the GISS series failing to detect the extreme warming in the upper most latitudinal zones. It is the UAH series at the upper most zones that is the most out of phase with the lower zones. In the early half of the time series the UAH series shows no warming in those uppermost zones while the lower zones are warming and then when the lower zones level, the UAH upper zones are warming at a fast rate.

     As a general observation, these data I show here would indicate that the Arctic polar amplification lags the warming/cooling in the lower latitudes and would imply some kind of feedback occurring in the Arctic polar region – at least from my simple minded viewpoint. Interesting also that the CW Hybrid series shows a warming trend of nearly 3 degrees C over the 34 year period from 1979-2012, or nearly 10 degrees C per century, and at actually a faster rate than that for the 1997-2012 period for the latitudinal zone 70N-90N. Even though that zone includes only about 3 percent of the earth’s surface area that large of a trend when averaged over the globe has not a small effect on the global mean trend.

     If there is a lagged polar amplification that adds/subtracts out of phase to the global mean perhaps it is time to look at global trends by Taminoating the Arctic polar amplification. (I do not do smiley faces).

     CW 60N-90N:
     http://imageshack.com/a/img35/981/ga3v.png

     CW 30N-60N:
     http://imageshack.com/a/img543/488/n761.png

     CW 00N-30N:
     http://imageshack.com/a/img824/5100/8733.png

212. Here are the links to the UAH series:

     UAH 60N-90N:
     http://imageshack.com/a/img28/6673/0pfy.png

     UAH 30N-60N:
     http://imageshack.com/a/img14/934/i11q.png

     UAH 00N-30N:
     http://imageshack.com/a/img36/2792/dxph.png

213. Here are the links to the GISS In-Filled series.

     GISS 60N-90N:
     http://imageshack.com/a/img203/9467/y1ce.png

     GISS 30N-60N:
     http://imageshack.com/a/img59/5053/9jw2.png

     GISS 00N-30N:
     http://imageshack.com/a/img716/3950/8ki6.png

214. Kenneth,

     Having the three series on separate graphs makes them difficult to compare. Personally, I don’t see how CW get such a high trend for 80-85 and 85-90. Speaking of which, I didn’t think that UAH covered 85-90N. I know that RSS cuts off at 82N and that satellite sea ice data has a hole at the pole too. Also, doesn’t 85-90 have ice coverage year round, not including the occasional polynya?

215. DeWitt Payne (#121358)
     “I didn’t think that UAH covered 85-90N.”
     I was surprised too, because the product which I usually look at, supplies zonal values for tropics (20S-20N), north extra-tropics (20N-85N), and north polar (60N-85N) — so I assumed that there was a coverage hole above 85N. And there’s a daily zonal file which contains, if I recall correctly, 67 2.5-degree latitude zones.

     However, their gridded product includes all latitudes.

216. At KNMI, they claim the UAH gridded goes to 88.75N. My understanding is the reported global average is 85N-85S

217. Re: Andrew_FL (Nov 25 12:09),

     I suspect they don’t include it because it’s too noisy because the surface area is so small. The NoPol anomaly is already far noisier than the tropics, probably for that reason. Yet another reason to discount CW trends at high latitudes.

     Which reminds me, do CW put any sort of error estimate on their trends? IMO, they should be wall-to-wall.

218. Andrew_FL (#121543)
     “At KNMI, they claim the UAH gridded goes to 88.75N.”

     KNMI has this:
     “UAH MSU v5.5 Tlt anomaly […] Spencer and Christy (U. of Alabama Huntsville) MSU lower troposphere temperature v5.5
     X axis: whole world in 144 2.50° steps, first point at 178.75° W, last point at 178.75° E
     Y axis: regular grid with 72 2.50° steps, first point at 88.75° S, last point at 88.75° N ”

     So the last Y-row contains 2.5°cells centered at 88.75°N; in other words from 87.5°N to 90°N

     I agree that one would expect the error in the UAH polar cells to be larger than elsewhere due to coverage…but I don’t see any quantification of this. The saving grace is presumably that the region beyond 85 degrees is < 0.4% of the global surface area.
     .
     Edit: Yes, UAH's reported global average covers 85S-85N. I suspect this is due to unreliability of polar measurements. The conspiracist ideator will, of course, assume that the global average excludes latitudes beyond +/-85 in order to reduce trends.

219. Kenneth,

     I’m betting that if you plotted CW temperatures vs sine latitude so it reflects actual distance on the surface, you get a high temperature gradient. I don’t believe that you can physically justify a high temperature gradient in the polar region.

220. Here is a quick calculation of the trends in degrees C per year for the Arctic polar region in 5 degree increments from the CW Hybrid gridded series from 1979-2012. The 1st column after the latitudinal zone ID is the trend, the 2nd is SE for the trend and the 3rd is the t.value. None of these calculations are adjusted for auto correlation, but I suspect/assume that that correlation is similar for all zones and the corrected values would remain relatively the same.

     85N-90N: 0.07923136 0.008085646 9.799015
     80N-85N: 0.08845048 0.007482762 11.820566
     75N-80N: 0.08031298 0.005675114 14.151785
     70N-75N: 0.06748949 0.004927798 13.695671
     65N-70N: 0.05256289 0.005034898 10.439712
     60N-65N: 0.03826198 0.004449203 8.599738

     All the gridded series data for UAH and GISS Infilled were taken from links from the Cowtan and Way paper.

221. Now here is a similar calculation of the trends in degrees C per year for the Arctic polar region in 5 degree increments from the CW Hybrid gridded series from 1997-2012. And again the 1st column after the latitudinal zone ID is the trend, the 2nd is SE for the trend and the 3rd is the t.value with none of these calculations adjusted for auto correlation.

     85N-90N: 0.15085296 0.02390942 6.309353
     80N-85N: 0.16916092 0.02237897 7.558922
     75N-80N: 0.14515695 0.01605377 9.041923
     70N-75N: 0.11960372 0.01355527 8.823413
     65N-70N: 0.08248035 0.01350313 6.108239
     60N-65N: 0.03534776 0.01296714 2.725949

     Obviously the lesser degrees of freedom using a shorter time period is in effect here on the t.value.

222. HaroldW (Comment #121547) -“So the last Y-row contains 2.5°cells centered at 88.75°N; in other words from 87.5°N to 90°N”

     Thanks, I didn’t realize they were centered, I am not always clear what KNMI is doing.

     “The conspiracist ideator will, of course, assume that the global average excludes latitudes beyond +/-85 in order to reduce trends.”

     Then one can assure them by showing that including it makes basically *zero* difference, with the data from KNMI. You can try it yourself-I did something like this. Partly because it is such a small area, it’s contribution to the global mean is negligible, despite large trends.

     DeWitt Payne (Comment #121546)-Yes, I think it’s in the early publications somewhere that the noise is far too great. Of course, it makes very little difference because it is such a small area.

     DeWitt Payne (Comment #121548)-Interesting thought, I’m not sure whether I understand your reasoning though. Why (and I sincerely don’t know why and would like to) is a high anomaly gradient unlikely?

223. Re: Andrew_FL (Nov 25 20:24),

     Why (and I sincerely don’t know why and would like to) is a high anomaly gradient unlikely?

     There are no geographic features that would prevent mixing, like mountain ranges. The surface is effectively smooth. Any significant increase in the latitudinal temperature gradient is going to increase pressure gradient flow, not to mention the more or less continuous circulation in the polar cell. The correlation length should be quite large. 1200km gets you from 70N 180E to 70N 0E, if I did my sums correctly. Other than Greenland, there’s pretty much nothing but the Arctic Ocean inside that circle.

224. DeWitt Payne (Comment #121566)-Thanks, that’s a pretty clear explanation, mostly.

     I don’t have much expectations for the appropriate smoothing radius at the surface-other than that I would expect a larger one above the boundary layer. I do expect that *above* the boundary layer, the smoothing radius is probably inversely proportional to the sign of the latitude.

     I’ve been wanting for a while to analyze anomaly correlations in LT data to see if this holds up (it should, the physical reason is pretty basic) but keep finding other projects to do. I wonder if anyone else has done this…probably.

225. DeWitt Payne (Comment #121548)
     “I’m betting that if you plotted CW temperatures vs sine latitude so it reflects actual distance on the surface…”
     Kenneth Fritsch (Comment #121561)
     “Now here is a similar calculation of the trends in degrees C per year for the Arctic polar region in 5 degree increments from the CW Hybrid gridded series from 1997-2012.”

     I calculated the same trends as Kenneth (numbers agree), and plotted them against sine(lat) here. The CW time trends actually drop a little in the top band 85-90.

226. Andrew FL. “Other than Greenland, there’s pretty much nothing but the Arctic Ocean inside that circle.”

     Right, so the impact is more East than West and Arctic instead of sine lat. The UAH trend for 75-85N 0-180W from 1997 is half that of 0-180E. It is like the Arctic version of the Santa Anna winds. Then when you consider that as the ENSO region shifts westerly you have more of the higher Tibetan Plateau region in poleward heat transfer.

     So is heat of compression due to internal variation a “sign” of CO2 and the end of the “pause”?

227. Nick Stokes that interactive plot you show puts a lot of information in one graph. Do you have code for doing that that you might share?

228. Similar to what dallas noted for the UAH case above, for the CW Hybrid gridded series in the latitudinal zone above 60N trends vary significantly by longitude. I divided the 60N-90N zone into 6 equal area parts by longitude and plotted the spline smooth for each area. The graphs are in the link below. The resulting series for each longitude was derived by weighting by the area of the 6 latitudinal zones from 60N-90N from which the data originated.

     http://imageshack.com/a/img713/9867/azg5.png

229. Kenneth Fritsch (Comment #121576)
     “Nick Stokes that interactive plot you show puts a lot of information in one graph. Do you have code for doing that that you might share?”

     It’s Javascript, so it’s in the html. If you save that, there are just two sections. Search for, say, DJF. It’s between “input” tags, which is the radio button, and all between table tags that lay out the image and buttons. That’s all HTML, but the link to JS is ‘onclick=”DxClick(0,1)”‘.

     So search for DxClick. It’s a function in between two script tags. That script and the table section are all you need, and you could move that to a separate .html file; it should work stand-alone.

     The web address is in the script – you can substitute your own, where you would put your own plots.

     The place to learn about all this is here. Notepad++ is good for highlighting the JS functionality. JS looks like C.

     I posted code here for the spaghetti code viewer, which is somewhat related. There I got the R program to print out the Javascript.

230. Thanks, Nick, I’ll see what I can do.

231. I used 5 runs of the CMIP5 model,GISS-E2-H_p1_rcp45, for the period 1979-2012 and looked at the spline smoothed anomalies for the 6 most northern latitudinal zones in increments of 6 degrees as a comparison with the CW Hybrid, UAH and GISS In_Filled series.

     The 5 runs give a better idea of the effect of “weather” noise on the results than a single run would. The results for the 5 runs are shown in the graphs in the links below. The progressively steeper trends seen in the observed series on proceeding northward are also seen with the model runs.

     Runs 1 and 2:
     http://imageshack.com/a/img199/158/6rp7.png

     Runs 3 and 4:
     http://imageshack.com/i/nsn2uup

     Run 5:
     http://imageshack.com/a/img33/4385/gbrs.png

232. Kenneth Fritsch (Comment #121590),
     For goodness sakes man, write a guest post. The dribs and drabs are like torture.. and confusing to boot.

233. Here is a better view of Runs 3 and 4.

     Runs 3 and 4:

     http://imageshack.com/a/img856/443/n2uu.png

234. I would also appreciate seeing Kenneth write up a post. I often enjoy the contributions here from the “community” we’ve got. For my own part, I’ve rarely felt my work up to a level where I would ask someone else to host it (although I did that one time at Jeff’s, mostly because I was not saying much that was controversial).

235. This site has some animation of the NH SSW events that I think helps visualize what is happening during Arctic Winter Warming.

     http://p-martineau.com/ssw-animations/

     There is actually a trend starting about 1988 to 1998, but the coverage of the area is so limited hard to determine the magnitude of the events. Martineau gives it a good shot though.

236. DocMartyn says:

     “Now if I were to measure temp every minute I get spikes as my Tmin and Tmax, if I take hourly averages, then Wed 10th gives a low Tmax, and the Tmin of Tues 9th is overestimated.”

     Sorry if I’m being obtuse, but I still don’t understand. A maximum is a maximum, even if only lasts half a minute, isn’t it? The good old min/max thermometer goes up to max (and down to min) rather quickly and keeps these values, by construction of the apparatus, until the operator doing the reading resets it. There is no question of “taking averages”, or there shouldn’t be; nor should there be any possibility that a max or min happens unnoticed between readings, *because the max/min thermometer by construction sticks at the extreme values even after the temp changes again*. At least that’s the principle on which the cheap, pretty old specimen hanging outside my garden shed works. Only if more than one full day/night cycle has passed between readings will an ambiguity occur, because it will be impossible to reconstruct which of, say, three nights that have passed between two readings 72 hours apart was the coldest, or what the min temps of the other two nights were. In such a case it would be wrong to assume “This is the min temp of each of these three nights” (or one would indeed overestimate two of them), but still quite correct to say “This is the min temp of the last 72 hours”. If one would leave a min/max thermometer undisturbed for a full year, it should display the absolute max and min temps during the whole past year, though of course the contraption couldn’t tell you when in time these extremes happened and how long they lasted. There is *nothing* in its simple construction capable of producing “averaged” or “biased” readings, provided that the thermometer itself is well-calibrated!

     OTOH, if we are talking about hourly (or even more widely spaced) samples of current/momentary temps at the time of reading, I immediately understand the TOB issue caused by undersampling, but I am under the – naive? – impression that even a hundred years ago, people were perfectly aware of this and eliminated this confounding effect by means of the “mechanical” min/max thermometer which always gives *the actual min/max values since the last reading*. All care that’s needed is to read it at approximately 24-hour intervals, preferably at hours not too close to the time when mins and maxes actually happen to avoid accidentally having two daily mins maxes between readings – a trivial task since the average start and end of a work day provide suitable points in time for a human researcher (extreme temps won’t usually happen just after the sun has risen, nor shortly after sundown). What’s the fuss all about? Talking of TOB with daily min/max readings looks suspiciously like an attempt to discredit – or arbitrarily “adjust” – historical records, in the hope that nobody in the 21st century is familiar with 19th century technology any more. Our grandfathers weren’t all stupid!

237. Moderator: Sorry I don’t seem to be able to delete the draft version of my text (#121600) manually, maybe because it’s still in moderation. Only #121601 is valid, please delete my earlier post. Thank you!

238. SteveF (Comment #121591)
     November 26th, 2013 at 7:47 pm

     It is a work in progress. Once finished it could be summarized in 2 or 3 paragraphs relating back to analyses already shown. Actually my postings here have been to learn from others inputs – as my understanding of the Arctic amplification in detail is weak.

239. I will summarize my analysis and inputs – as I interpret them from others on this thread – with reference to the question that I originally posed. That question was essentially: Given the not insignificant change in the global temperature trend from 1997-2012 that results from the Cowan and Way Hybrid gridded series combined with the fact that much of it comes from the Arctic polar region and further that the Arctic trend rate is increasing while the lower latitudes are decreasing, what does that mean in terms of what we know about the Arctic polar amplification and further should we expect that if the pause in warming continues for some years will the Arctic warming subside and produce an even smaller global trend in temperature.

     The observed gridded series, CW Hybrid, GISS and UAH and the gridded modeled series, GISS-E2-H_p1_rcp45 (see link to Run4 spline smoothed anomalies by latitude below) show a lag in the slowdown of the Arctic warming compared to the slowdown in lower latitudes. That relative slowdown varies with series and with the CW hybrid showing the least amount of Arctic slowdown. The gradient of anomaly trends to warmer going northward from 60N is large and evident in all the series mentioned above. That this area of the globe is a flat open area and amenable to efficient mixing of air would appear counter to what is observed and modeled.

     If the Arctic polar amplification is dependent on transport of heat from the lower latitudes as the term implies than I would guess that the lag in the Arctic warming rate must be the result of some feedback effect in the polar region that has momentum. If that is the case then I would expect a continuing pause in the global warming will eventual reduce the Arctic warming. On the other hand if the Arctic polar amplification or warming rate is more or less independent of the lower latitudes one would not necessarily expect to see a reduced Arctic warming. From the inputs to this thread I got the idea that the Arctic warming depends at some point in time on the heat generated at the lower latitudes.

     http://imageshack.com/a/img89/4498/pcqb.png

240. Kenneth Fritsch says, November 27th, 2013 at 1:02 pm:

     “If the Arctic polar amplification is dependent on transport of heat from the lower latitudes (…)”

     Of course the Arctic polar amplification is dependent on transport of heat (energy, that is) from the lower latitudes. That is what the Arctic polar amplification is all about. Energy from the Sun absorbed by the Earth system in the tropical/subtropical oceans is transported poleward. An equal amount of energy can raise the absolute temperature of the cold, dry Arctic atmosphere much more than it can the hot and humid tropical atmosphere.

     So where is tropically absorbed shortwave energy from the Sun by far most efficiently transported to a polar region? The North Atlantic. This is where the Arctic polar amplification has its origin. The Pacific transport from the tropics to the Arctic is much, much less (because it is primarily zonal rather than meridional), and so is pretty much all transport of energy (‘heat’) from the tropics to the southern polar region, Antarctica.

241. Kenneth Fritsch says, November 27th, 2013 at 1:02 pm:

     “If the Arctic polar amplification is dependent on transport of heat from the lower latitudes (…)”

     Of course the Arctic polar amplification is dependent on transport of heat (energy, that is) from the lower latitudes. That is what the Arctic polar amplification is all about. Energy from the Sun absorbed by the Earth system in the tropical/subtropical oceans is transported poleward. An equal amount of energy can raise the absolute temperature of the cold, dry Arctic atmosphere much more than it can the hot and humid tropical atmosphere.

     So where is tropically absorbed shortwave energy from the Sun by far most efficiently transported to a polar region? The North Atlantic. This is where the Arctic polar amplification has its origin. The Pacific transport from the tropics to the Arctic is much, much less (because it is primarily zonal rather than meridional), and so is pretty much all transport of energy (‘heat’) from the tropics to the southern polar region, Antarctica.

242. Kenneth Fritsch says, November 27th, 2013 at 1:02 pm:

     “If the Arctic polar amplification is dependent on transport of heat from the lower latitudes (…)”

     Of course the Arctic polar amplification is dependent on transport of heat (energy, that is) from the lower latitudes. That is what the Arctic polar amplification is all about. Energy from the Sun absorbed by the Earth system in the tropical/subtropical oceans is transported poleward. An equal amount of energy can raise the absolute temperature of the cold, dry Arctic atmosphere much more than it can the hot and humid tropical atmosphere.

     So where is tropically absorbed shortwave energy from the Sun by far most efficiently transported to a polar region? The North Atlantic. This is where the Arctic polar amplification has its origin. The Pacific transport from the tropics to the Arctic is much, much less (because it is primarily zonal rather than meridional), and so is pretty much all transport of energy (‘heat’) from the tropics to the southern polar region, Antarctica.

243. Kenneth Fritsch says, November 27th, 2013 at 1:02 pm:

     “If the Arctic polar amplification is dependent on transport of heat from the lower latitudes (…)”

     Of course the Arctic polar amplification is dependent on transport of heat (energy, that is) from the lower latitudes. That is what the Arctic polar amplification is all about. Energy from the Sun absorbed by the Earth system in the tropical/subtropical oceans is transported poleward. An equal amount of energy can raise the absolute temperature of the cold, dry Arctic atmosphere much more than it can the hot and humid tropical atmosphere.

     So where is tropically absorbed shortwave energy from the Sun by far most efficiently transported to a polar region? The North Atlantic. This is where the Arctic polar amplification has its origin. The Pacific transport from the tropics to the Arctic is much, much less (because it is primarily zonal rather than meridional), and so is pretty much all transport of energy (‘heat’) from the tropics to the southern polar region, Antarctica.

244. Kenneth Fritsch says, November 27th, 2013 at 1:02 pm:

     “If the Arctic polar amplification is dependent on transport of heat from the lower latitudes (…)”

     Of course the Arctic polar amplification is dependent on transport of heat (energy, that is) from the lower latitudes. That is what the Arctic polar amplification is all about. Energy from the Sun absorbed by the Earth system in the tropical/subtropical oceans is transported poleward. An equal amount of energy can raise the absolute temperature of the cold, dry Arctic atmosphere much more than it can the hot and humid tropical atmosphere.

     So where is tropically absorbed shortwave energy from the Sun by far most efficiently transported to a polar region? The North Atlantic. This is where the Arctic polar amplification has its origin. The Pacific transport from the tropics to the Arctic is much, much less (because it is primarily zonal rather than meridional), and so is pretty much all transport of energy (‘heat’) from the tropics to the southern polar region, Antarctica.

245. Not sure what happened there. Could you please delete all but the first of my (identical) comments above, lucia? I tried to, but seemingly couldn’t. Thanks.

246. Wow. Déjà déjà déjà déjà vu.

247. Kristian (Comment #121636)
     November 28th, 2013 at 7:53 am

     No problem, Kristian, my wife likes to communicate with me in this manner sometimes with perhaps a longer lapse between comments.

248. Kristian, my recent readings about the Arctic polar amplification put forth a number of mechanisms for the source and extent of the amplification. Most of these papers admit to an unsettled debate at this point on the matter.

     My immediate interest is in putting forth a rationale, or conjecture if you will, that might explain what I see in these observed and model gridded temperature series.

249. Kenneth it’s one of the points that interest me.

     I’m thinking of it as one of those variety entertainment shows where a man is trying to keep spinning plates on poles. You spend too much time on one plate and the rest start to look like thet are going to fall.

     Cowtan and Way have discovered some of the missing heat and placed most of it in the Arctic. But by doing that we now have too much heat in one part of the system and too little in others, And we have polar amplification that starts to look far higher than earlier periods (like the early 20th century).

     Having said that I agree with you that nobody (apart from Kristian) has yet suggested a definitive explanation for polar amplification. All the mechanisms you mention are in play.

     What Kristian said is also true it would seem that Atlantic water play an important role. It’s possible that those water only peaked around the mid/late-2000’s (http://prj.noc.ac.uk/ExtendedEllettLine//research-and-impact). To be consistent one would have too assume that the rate of warming in the Arctic is going to have to slow in order to come in line with the slowing (halted??) sub-polar NH warming.

250. I have found what I judge to be the most comprehensive review of the Arctic polar amplification and relating to heat transport to the Arctic polar region and feedback effects. The link to this article is given below.
     This review shows me that the positive feedbacks from sea ice albedo and perhaps secondarily from the clouds, vegetation growth and black carbon aerosols are probably what leads to the amplification of warming from the mid latitudes. The transport mechanisms of heat from the mid latitudes to the Arctic are discussed in the article include conveyors in the atmosphere and sea with various sub routes.

     The article notes the historical evidence for Arctic amplification in Arctic warming and cooling modes for Milankovitch cycles and those associated with other historic periods. The article points to requirement for the positive feedbacks to obtain the high estimated warming ratios of the high northern to mid latitudes caused by the Milankovitch cycles.

     The article further explains the expected and observed greater Arctic polar warming in the winter and autumn months for the positive sea ice albedo as the result of more open waters absorbing solar energy in the summer and not adding significantly to the surface air temperature and then releasing the heat from the sea into the surface air in the autumn and winter.

     The latitudinally observed and modeled gridded temperature series I have shown in this thread and the seasonal temperature warming from Nick Stokes interactive graphic linked above seem to me to agree to an amplification from mainly reduced sea ice albedo and ancillary from increased vegetative growth. It appears that the feedback effect lags the warming and cooling of the lower latitudes and that observation points to the important differences seen beween the CW Hybrid gridded series warming by latitude to that for GISS and UAH. CW shows little or no evidence of an inflexion in the accelerated warming trend in the Arctic polar region since 1997 while the GISS Infilled series shows a definite inflexion and bending of the warming there while the UAH warming rate at those latitudes shows a hint of a bending of the warming trends. The GISS model with 5 runs shows rather mixed results for the what we might expect from a pause in the lower latitude warming.

     Of interest to me here is at what rate might we expect the Arctic warming to subside if the warming pause at lower latitudes continues for several more years. From what I read it might well be related to how fast the sea ice reduction can physically turn around although none of the literature that I have found relates the polar amplification to the warming pause. I find the validity of Cowtan and Way temperature estimates in the upper Arctic polar regions critical to this issue.

     http://lightning.sbs.ohio-state.edu/geog8901/paper/polar_Serreze2011.pdf

Comments are closed.