el Niño and Satellite Data

Having compensated surface temperature data for el Niño, volcanic aerosols, and solar fluctuations, it’s appropriate I should do the same for satellite temperature data. After all, upper atmosphere temperature (what the satellites estimate) responds to these factors much more strongly than the surface temperature, so it can be argued that it’s more important to compensate satellite data than surface data. Not doing so can cause some very misleading conclusions about temperature trends.


I’ll use the lower-troposphere data (TLT) from RSS (Remote Sensing Systems) and UAH (Univ. of Alabama at Huntsville, v5.6). Here’s the data from RSS, together with the model incorporating all three exogenous factors:

rss

Just as with surface temperature data, the match is impressive, and demonstrates that the extreme high temperature during 1998 is just about entirely due to the el Niño. The same is true for UAH data:

uah

Impressive indeed.

After removing the estimated influence of el Niño, volcanic activity, and solar fluctuations, we have a much better picture of what the satellite data are saying about the part of temperature change due to human influence. Here’s the corrected data (annual averages, UAH in blue, RSS in red) compared to the surface temperature data from NASA (in black):

adj_1yr

It’s surprising how well they all agree, until after 2011 when RSS diverges and after 2013 when UAH begins to diverge. A more direct comparison is available by subtracting the surface temperature data (from NASA) from the satellite data. Here’s UAH data minus the NASA data:

uah_giss

Here’s RSS data minus the NASA data:

rss_giss

The big difference between RSS and NASA is the declining RSS temperature after 2011, but there’s also distinct sign of a consistent decline from 2000 onward.

To compare the two primary satellite data sets, here’s RSS data minus UAH data:

rss_uah

Again there’s a distinct sign of decline in RSS (relative to UAH) after 2000, with RSS warming more rapidly than UAH pre-2000 and less rapidly post-2000.

Readers are invited to draw their own conclusions from the foregoing. But one conclusion is blatantly obvious from the corrected data: that claims of “no warming since 1997” or “no warming since 1998” based on satellite data are wrong. The false impression is a direct consequence of the impact of el Niño; once that confounding influence is removed, it’s easy to see how silly those ideas are. Unless, of course, you’re a republican running for president like Ted Cruz.


If you like what you see, feel free to donate at Peaseblossom’s Closet.

Advertisements

31 responses to “el Niño and Satellite Data

  1. dikranmarsupial

    I’m surprised just how well the model fits are in the first two figures, very interesting demonstration.

  2. After removing the estimated influence of el Niño, volcanic activity, and solar fluctuations

    That’s considered poor form in discerning denialati circles.

    As I’ve posted before, the satellite 50-60 GHz microwave data seems quite interesting, and I suspect it could be reprocessed and mined for further information. But with RSS able to devote only limited resources to the problem and Spencer and Christy at UAH providing deeply suspect analysis, such a task may require new bodies.

    The third figure shows why it’s generally inadvisable to draw conclusions by eyeballing a noisy trend.

  3. UAH v5.6, I take it?

  4. Be fun [¹] to do the same sort of fits on the outputs of model runs to see if they result in similar parameters.

    [¹] for a rather specialized definition of “fun”.

    • Timothy (likes zebras)

      It would certainly be interesting since the models have different levels of success when it comes to modelling ENSO.

  5. yeah difference, but how?

  6. No question you’ve shown the “conclusion is blatantly obvious … that claims of ‘no warming since 1997’ or ‘no warming since 1998’ based on satellite data are wrong.” Tamino, you’ve knocked another one out of the park with your usual aplomb.

    But what was the point of comparing the RSS and UAH data without the influences of el Niño, volcanic activity, and solar fluctuations with NASA GISS (presumbly still with those influences present)? I Yes the resulting lines appear similar, but how is this not more than a distraction from your principal argumentative grand slam? Maybe I’m missing something.

    • “Skeptics” like to argue that the surface records and the satellite records are appreciably different (in particular, that the satellite records are better, which is likely not true). If the corrected series match though that’s just another domino knocked over. (That may not be the correct analogy.)

    • You may also be missing the idea that it is *interesting* when you get disagreements like this. One team or both have made a mistake, and finding it and correcting it is part of the scientific process. Tamino has given a good case that there is something worth investigating here.
      As you say, in the big picture looking at this is perhaps a little luxury, but I think it is worth following up.

      • Thanks Alex and John. Good insights both. I read and commented on this article before noticing Tamino’s previous post “Correcting for more than just el Niño“, in which he actually plotted NASA GISS without el Niño, volcanic activity, and solar fluctuations. It seems to me that would also be a useful thing to be comparing the correspondlingly modified RSS and UAH datasets against. (That one does look a little different than the NASA GISS plot used here.)

    • That’s not the uncorrected GISTEMP record; it’s the corrected one Tamino described in the last post.

      • I think you may be right, but after stretching the one to make the scales the same, it still looks to me like there are some subtle differences. It would be nice if Tamino could verify that.

  7. Lawrence McLean

    Hi Tamino, I hope you do not mind me to ask a favour, before I reply to a comment on you-tube the favour is to verify my proposed reply.

    The comment is: “Show me a graph of Mauna Loa CO2 to any variable of climate in any data set on earth, and show me a least square regression fit, then an r2 correlation over 0.7 to show a significant correlation.”

    My proposed reply is:
    “It is not unexpected that as you call it a least squares regression fit with your given criteria, will not confirm climate change, at least at this stage. That technique is not appropriate for climate data. The climate data is affected by very significant periodic, but not totally random forcings, in particular the El-Nino and La Nina Pacific Ocean events, there are also others such as the Indian Ocean Dipole.

    The climate data is analysed and statistical trends are regularly published. If you are certain that they are mistaken you should highlight their mistakes and let them know.

    It is possible that even seasonal data, that is, correlating the perfect geometrical rate at which the ratio of night time to day time length changes against daily temperature data during a 4 week period in a mid season cycle of a given hemisphere, may also fail the least squares regression fit for the criteria that you specify. It if were to fail, would that invalidate the Theory of Seasons?

    Although I am rusty on statistics myself, the last I did any was at University many years ago, nevertheless, I am somewhat sceptical that you actually know what you are talking about.”

    What do you think?

    • I think maybe your reply fails the politeness test. I suggest the following:

      “You are a denier, plain and simple, and almost certainly an idiot to boot. Any explanation would surely be wasted on you. I encourage you to keep on denying, as you will find it considerably less taxing than thinking.”

      There, I think that passes the politeness test.

    • Actually, maybe I was a bit harsh there. They may not realise that the dependence of temperature on CO2 levels is logarithmic. If they just try and do a linear fit of temperature to CO2 level, of course it won’t work.
      Then of course you’d have to throw in that CO2 is not the only factor. And then you end up with an analysis like Tamino’s.
      But when all is said and done, my bet is they are just an out and out denialist.

    • M. MacLean,

      The correlation between CO2 and Hadley CRUTEM4 temperature anomalies for the years 1850-2014 is about r = 0.91, and r^2 = 0.82. I believe this meets your criterion.

    • The comment is: “Show me a graph of Mauna Loa CO2 to any variable of climate in any data set on earth, and show me a least square regression fit, then an r2 correlation over 0.7 to show a significant correlation.”

      Uhh…how about temperature? For GISTEMP regressed on the Mauna Loa record, I get r^2 of .89 for annual data, .78 for monthly.

  8. Lawrence McLean

    There are a number of issues. I used to work in Industry, for a while I had to manage instruments. Certain instruments required recalibration very day, obviously Satellite designers would not launch something that required recalibration every day, nevertheless, is it possible the instrument has degraded in some way, maybe micro asteroids or cosmic rays have done some low level damage. Is the data being processed correctly. I cannot see how this is “the best we got”

  9. Barton. Great web site.

    My own contribution from 1979 till present for 12 month averaged HadCrut 4 data is http://tinyurl.com/j6fmxad and likewise (in this case the C02 data is also averaged over a year) at http://tinyurl.com/j47wg25 . For the satellite data you get a similar picture but with lower correlation coefficients due to the greater effects of El-Nino etc..

  10. Tamino, have you looked at the satellite data for the stratosphere?

  11. Why does the upper atmosphere respond to ENSO much more strongly than surface temps?

  12. Hi,

    Just saw this post from esa
    http://www.esa.int/Our_Activities/Observing_the_Earth/SMOS/SMOS_gets_help_from_Tibet

    One of the systems to map satellite data to earth data :)

  13. Pete Dunkelberg

    I think it would be helpful to see this analysis for the radiosonde data, if you have the time for it.