New Kid in Town

El Niño (“the little boy”) is the warm phase of an ocean/atmosphere oscillation; it helps heat go from the ocean to the atmosphere and warm up our weather. During its counterpart, la Niña (“the little girl”), it does the opposite, moving heat from the atmosphere into the oceans. The whole phenomenon is called the El Niño southern oscillation, or ENSO, but it often happens that we just say El Niño for the whole thing.

It’s one of the things that affects Earth’s global temperature — temporarily — and there are lots of ways to quantify it, i.e. to “put a number on it.” One of the best is MEI, the Multivariate El Niño Index. It’s the way I describe El Niño when I adjust global temperature for temporary factors (volcanic eruptions, solar variations, and yes, El Niño).

There’s a new kid in town, or at least, a new way to quantify El Niño. The scientists who constructed MEI have come up with a new, improved version (version 2), so of course I’ve re-computed the adjustments based on the new version of MEI data. Since the new version covers the year 1979 to the present, that’s the time span for which I’ve computed adjusted data.


Let’s start with the global temperature data before we try to remove the influence of temporary factors. Here it is (monthly averages, the whole globe, since 1979, from NASA):

My estimate is that the whole el Niño thing had this much effect on global temperature during this time:

We can see the warm spikes from the big el Niño events of 1998 and 2016. But there’s an even bigger one in 1983, even though that didn’t show up as a super-hot year like 1998 and 2016 did.

That’s because there are other temporary factors at work, and as chance would have it, one of them — the eruption of the el Chicon volcano — canceled out most of the 1983 el Niño warming. Here’s my estimate of the effect of all three fluctuation factors (el Niño, volcanic eruptions, and solar variations) on global temperature:

The biggest upward peaks, the times when the world was warmest because of fluctuations (not trend), were the big el Niño peaks of 1998 and 2016.

When we subtract this estimate from the original data, we get an estimate of how temperature has changed in addition to those temporary factors. We can call it the “adjusted” data, and here it is:

The trend is still there, unchanged, but the fluctuations are much smaller, and the steadiness of the uninterrupted warming trend is easy to see.


If you’re read this far, you probably have more than average interest in global temperature. You might even know already that more than one organization estimates global average temperature. Here’s the yearly average, for the whole globe, according to two of the best-known data sets, one from NASA (we’ve just seen their monthly data), the other from HadCRU (the Hadley Centre/Climate Research Unit in the U.K.):

There’s an obvious trend, and just as obviously it fluctuates around that trend. In fact, as obvious as the trend is, the fluctuations are plenty big enough to be quite noticeable.

We’ve already computed adjusted data for NASA; we can do so for the HadCRU data too. It gives this:

The trend is still the same, but now the fluctuations are a lot smaller. Of course we can never eliminate them, but by at least getting rid of the fluctuations we can explain, we make our view of the trend that much clearer.

And the world’s warming rate? NASA data put it at 0.0178 °C/year, the HadCRU data estimate 0.0177 °C/year. That’s plus or minus 0.0018 °C/year. Basically, they’re telling the same story: 95% chance it’s somewhere between 0.0160 and 0.0196 °C/year.


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


33 responses to “New Kid in Town

  1. Thanks for the update.

    If I may return to a previous topic (now closed), I’d like to note once again Jay Inslee’s focus on climate change. I think it’s important that we do what we can to reward that priority by at least talking about his candidacy!

    https://slate.com/news-and-politics/2019/04/jay-inslee-campaign-climate-change.html

    [Response: I agree.]

    • Inslee has been asking supporters for donations so that he can qualify to get into the upcoming Democratic Party Presidential candidate debates. According to the DNC: “Candidates may qualify for the debate by demonstrating that the campaign has received donations from at least (1) 65,000 unique donors; and (2) a minimum of 200 unique donors per state in at least 20 U.S. states”

      A donation of any size qualifies. Put your two bits worth where your mouth is people…

  2. So the Multivariate El Niño Index is calculated using the recorded temperatures of the El Niño.
    Hardly surprising it smooths the graph.

    [Response: You should learn how the MEI is computed *before* you show how ignorant you are.]

    • Tamino, I at least am interested in how the MEI is calculated. The first thought I had when you said there was a new, better, one, was ‘How does that get generated’. If you could do a blog on that I’m sure a lot of people would find it interesting. And you could educate Jeff too :-)

      • “MEI is determined as the first principal component of six different parameters: sea level pressure, zonal and meridional components of the surface wind, sea surface temperature, surface air temperature and cloudiness.”
        https://en.wikipedia.org/wiki/Multivariate_ENSO_index#Overview

      • Jeff,
        The Wikithing page misses out some explanation which should be pretty obvious to someone not wanting to “show how ignorant (they) are.” To quote from ESRL

        “The bi-monthly Multivariate El Niño/Southern Oscillation (ENSO) index (MEI.v2) is the time series of the leading combined Empirical Orthogonal Function (EOF) of five different variables ( … ) over the tropical Pacific basin (30°S-30°N and 100°E-70°W).” [My bold]

      • Al,
        The above post says
        “We can see the warm spikes from the big el Niño events of 1998 and 2016. But there’s an even bigger one in 1983, ”
        So the fair assumption is the el Niño events affect the average global temperature.
        So I still think that using el Niño surface temperature to calculate the index will inevitably smooth the global temperature graph when the index is applied.

      • “….the fair assumption is the el Niño events affect the average global temperature.”

        And just what would be the point of applying an ENSO factor to the temperature record if the former *didn’t* affect GMST?

      • “The Wikithing page misses out some explanation”

        Well, the wiki page is describing version 1 of the MEI index. That version was a combination of six components, as Jeff quotes, and the new version (2) uses five components.

      • Jeff,
        The El Niño of 1983 conicided with the eruption of El Chich’on, the pair of them cancelling each other out. That is why the adjustments of global temperature adjust for Solar & Volcanic as well as ENSO.
        The adjustment you appear to be describing suggests that ENSO raises Pacific temperatures and at the same time raises global temperatures. This is not the case. Firstly, the indices that measure Pacific temperature (NINO3.4 or ONI) are only part of the MEI index. They do not fully capture the strength of ENSO. Secondly, the impact of ENSO on global temperature is not immediate but occurs after a time lag of (usually) some four months.
        The similarity of ENSO wobbles & global temperature wobbles is marked (see this graph plotting MEIv1 against HadCRUT4 – usually 2 clicks to ‘download your attachment’) and so the assumption of ENSO impacting gloabl average temperature is not a difficult one to make.

    • Timothy (likes zebras)

      The very simplest way to define El-Nino is with sea surface temperatures in various areas of the Pacific Ocean, but I have no idea how the Multivariate El-Nino Index (MEI) is calculated. So I used Google to look it up.

      Turns out it is calculated using five different variables, which are:
      1. Sea level pressure.
      2. Wind in an east-west direction.
      3. Wind in a north-south direction.
      4. Outgoing longwave radiation – in the tropics this is mostly an indirect measure of how much deep convection is happening, therefore how high/cold the tops of the clouds are.
      5. And, yes, sea surface temperature, though because of the statistical way that the index is calculated it’s not as simple as saying that high temperatures lead to a higher value of the index. There is a pattern to match which involves colder temperatures in some areas too.

      See: https://www.esrl.noaa.gov/psd/enso/mei/

  3. @tamino:
    After adjusting, is there a annual cycle in the data which can also be adjusted?
    Can you confirm a significant 2 year cycle in the monthly data I’ve found?

    • Rather OT, but here’s an analogy of a potential El Niño scenerio:

      You feel feverish, and on average your temperature has been higher than
      100 F for 24 hours. NOAA, however, won’t declare an “official” fever until 5 consecutive similar days have passed.

      For the next 4 days your average T remains above 100 F, but near the end of that period the fever starts to break.

      Finally, after 5 days of suffering, you’re back to your old self…..98.6 F.

      This is when NOAA makes it official. You have a fever!

      ***
      (I could have added a 3-day running mean to the story, but it wouldn’t have changed the basic idea).

    • David B. Benson

      Search for information on the quasi-biannual temperature oscillation.

  4. Is the new kid able to explain more variability than the old geezer?

    • I checked this just now on the overlapping years 1979-2016. No difference, both show a standard error in the residual to a linear fit of 0.051ºC. Not such a big surprise, although there are some differences in individual years. The old measure has 1998 peaking at +0.38ºC and 2016 at +0.30ºC. The new one has them both peaking at +0.32ºC. But after taking annual averages they both look the same.

  5. I’ll likely regret asking this, because I probably won’t be able to understand the response; but could you explain, in *really* simple terms, how you went about computing the adjustments for MEI, volcanic and solar?

    Thanks

    [Response: Take the global temperature data, and find its “best fit” (mathematically) to the data for ENSO, volcanic, and solar variation. Allow for the fact that there can be a lag between some event (like a volcanic eruption) and when it gets around to affecting global temperature, so find which lags also give the “best fit” (mathematically). This best fit represents an estimate of how these fluctuation factors have affected temperature, so subtract that away and you’re left with the adjusted data.

    Yes, it’s more complicated than that. If you want the details, you can find them here.]

  6. Tamino

    Did you also see that MEI V2 rates the El Niño 1982/83 higher than the previous version? Previously, 1997/98 had a higher value in the MEI time series than 1982/83.

  7. Thanks, Bin! I’m surprised by how different some of the events are when using MEI as the metric versus ONI. For example, the MEI table shows the 1982/83 El Niño to be much stronger than the 2015/16. Just the opposite in the ONI table.

    • Snape

      Yes indeed, and ONI shows 2015/16 even stronger than 1997/98. But this is due to ONI actually using ERSST V5; please have a look at ONI’s earlier revision based on ERSST V4 (link above the data table).

      Btw: an interesting corner to look at for ENSO comparisons is JMA’s Global Temperature map:
      https://ds.data.jma.go.jp/tcc/tcc/products/gwp/temp/map/temp_map.html

      • The current version of MEI uses JMA reanalysis data for the wind component of their analysis. I believe ship records from COADS was the previous source for that data.

        I wonder why JMA prefer NINO3 region for their ENSO monitoring rather than NINO3.4.

    • Snape,
      Plotting out the MEI(v2)-MEI(v1), there is a distinct slope. OLS puts MEI(v2) declining at 0.07 per decade relative to MEI(v1). Thus the earliest El Niño (1982/83) will appear relatively stronger than later ones (eg 1997/98) under MEI(v2).

  8. “But this is due to ONI actually using ERSST V5; …”

    I think you mentioned that once before and I forgot. 😏

  9. Just so I understand correctly:
    ERSSTv4 was designed to allow comparisons of one month to another. ERSSTv5 is simply an improvement to that end. So the v5 ONI table should give a better picture (versus v4) of how the 2015/16 El Niño compared to the 82/83 El Niño. Is that right?

  10. @tamino: Do you find that the influence of the solar radiation variation is significant? Or there are problems due to the correlation with the volcanic forcing?
    The timing of the major volcanic eruptions in the second half of 20th century is really bad for the detection of solar temperature variations.

    [Response: The solar influence is statistically significant, but very small.]

  11. Can I extrapolate that to +1.8°C by the year 2119?

    • Martin,
      That would be a very bold extrapolation. But if you did it would be +1.8°C above the 2019 temperature which looks to be about +0.4°C above 1995 temperatures which in turn were perhaps something like +0.6°C above late-19th century temperatures. So your 2119 projection would be appraching +3°C, not a healthy place to be.
      The linear nature of the last 40-years of global temperatures is remarkable (see a HadCRUT long-term trend graph here – usually 2 clicks to ‘download yhour attachment’) but is it going to last 140 years?

      • “The linear nature of the last 40-years of global temperatures is remarkable but is it going to last 140 years?”

        Hell of a good point. Sure hope that it does, since if it doesn’t, the curve is apt to start bending in completely the wrong direction.