Why “GISTEMP LOTI global mean” is wrong and “UAHv6 tlt gl” is right

Ten days ago, Nick Stokes wrote a post on his “moyhu” blog where he – in his regular, guileful manner – tries his best to distract from the pretty obvious fact (pointed out in this recent post of mine) that GISS poleward of ~55 degrees of latitude, most notably in the Arctic, basically use land data only, effectively rendering their “GISTEMP LOTI global mean” product a bogus record of actual global surface temps.

Among other things, he says:

“The SST products OI V2 and ERSST, used by GISS then and now, adopted the somewhat annoying custom of entering the SST under sea ice as -1.8°C. They did this right up to the North Pole. But the N Pole does not have a climate at a steady -1.8°C. GISS treats this -1.8 as NA data and uses alternative, land-based measure. It’s true that the extrapolation required can be over long distances. But there is a basis for it – using -1.8 for climate has none, and is clearly wrong.

So is GISS “deleting data”? Of course not. No-one actually measured -1.8°C there. It is the standard freezing point of sea water. I guess that is data in a way, but it isn’t SST data measured for the Arctic Sea.”

The -1.8°C averaging bit is actually a fair and interesting point in itself, but this is what Stokes does; he finds a peripheral detail somehow related to the actual argument being made and proceeds to misrepresent its significance in an attempt to divert people’s attention from the real issue at hand. The real issue in this case of course being GISS’s (bad) habit of smearing anomaly values from a small collection of land data points all across the vast polar cap regions, over wide tracts of land (where for the main part we don’t have any data), over expansive stretches of ocean (where we do have SST data readily available) AND over complex regions affected by sea ice (where we indeed do have data (SSTs, once again) when and where there isn’t any sea ice cover, but none whatsoever when there is), all the way down to 55-60 degrees of latitude.

GISS 55N-55S

Figure 1. Note, base period here is 1981-2010, not 1951-1980. Data from KNMI Climate Explorer. 1200 km smoothing product.

Why does the thick golden curve in Fig.1 rise so much less from 1970 to 2015 than the thin black curve on top of it? The two after all cover the exact same latitudinal band, from 55 degrees north to 55 degrees south. Where 98-99% of the world’s human population happens to reside.

The reason is of course that the golden curve includes both the land AND the ocean surface, while the black one represents the land portion only. Land surfaces naturally warm (and cool) much faster than sea surfaces, both in the short term (as seen in the general interannual noise level) and in the longer term (as seen in multidecadal trends). The average of the two (land+ocean) thus always ends up somewhere in between either, depending on their respective area weighting. Within the 55N-55S band, the ocean/land ratio is 2.77 : 1, within the 55-90S band (the Antarctic), the ratio is 2.29 : 1 (sea ice excluded), while within the 90-55N band (the Arctic), it is 1 : 1.1 (again, sea ice excluded).

Let’s have a look, then, at GISTEMP’s 90-55N polar cap, the extended Arctic:

GISS 90-55N

Figure 2.

Yeah, so all of a sudden, there is a near-perfect match between the total curve (golden, meant to be land+ocean) and the land only curve (black).

Stokes says, in the quote above:

“But the N Pole does not have a climate at a steady -1.8°C. GISS treats this -1.8 as NA data and uses alternative, land-based measure. It’s true that the extrapolation required can be over long distances. But there is a basis for it – using -1.8 for climate has none, and is clearly wrong.”

(My emphasis.)

I wonder, what “basis” might he be referring to …?

What’s “clearly wrong” here is Stokes implying that all GISS do is simply leaving out ice covered areas, specifically whenever they are ice covered, because then we can no longer speak of the temperature of the actual sea surface in those particular areas (which is in itself true).

But that’s obviously not what they’re doing. They quite clearly ignore (remove, delete) existing SST data all the way down to 60-55 degrees of latitude, summer and winter, ice or no ice. And replace it fully with land data. In a region where the land data coverage is especially poor to begin with, at least beyond 60-65 degrees. In other words, they invent data, invent extra warming. Warming that isn’t really there. For as we all know, in a warming world, land data will rise at a much higher rate than ocean data, from differing thermodynamic properties alone. And when you then take out ocean data and put in land data in its stead, you increase the overall warming rate! It’s as simple as that.

So how can we tell if this is really what’s happening with the “GISTEMP LOTI global mean” product?

Once again, let’s bring “CERES EBAF ToA” in as the final arbiter.

First though, here’s how GISTEMP LOTI lines up with UAHv6 tlt within the 55N-55S band:

GISS vs. UAHv6, 55N-55S

Figure 3.

Not too bad, is it? The small divergence over the last couple of years is mostly a natural consequence of “The Blob” phenomenon in the North East Pacific Ocean, previously discussed here.

Watch what happens when we include the OLR data from CERES:

GISS vs. UAHv6 og CERES, 55N-55S

Figure 4.

This is just what we would expect. The surface heats the troposphere and the troposphere in turn emits (most of) Earth’s heat loss to space (as OLR at the ToA). This plot makes perfect physical sense.

But what if we were to move up to the 90-55N polar cap? Would we recognise this same tight relationship?

Well, between tropospheric temps (UAHv6) and the ToA OLR (CERES) we indeed still would (see Fig.6 below). But what about the surface, which we now know in this region is all land, according to GISS, even when it’s not …?

GISS vs. UAHv6, 90-55N

Figure 5.

Bear in mind, the troposphere is heated by the entire surface underneath, land AND ocean, not just the land. This goes for the 55N-55S band, and it likewise goes for the 90-55N polar cap. Naturally. Hence the obvious divergence here, where the surface all of a sudden appears to warm much faster than the troposphere above it, simply because the white curve is no longer the average of land and ocean; it is all land, no sea surfaces included … Watch how the amplitudes are considerably larger at the surface than in the troposphere. Now compare this inherently unnatural situation to the one in the 55-55 band (Fig.3 above), where things are how they should be.

We bring in the OLR once more:

GISS vs. UAHv6 og CERES, 90-55N

Figure 6.

Once again fits very well, if not perfectly, with the tropospheric temps. But no longer with GISS’s version of surface temps. And the reason is – still – very simple: They no longer include the oceanic portion, only the land part, weirdly replacing the ocean, thus effectively doubled in extent …

GISS quite evidently haven’t got this right.

You purport to compile a balanced (weighted) global surface temperature record, but include the ocean surface only between 55/60N and 55/60S, while beyond those latitudes you use land data only. Then what do you expect? Of course your 90N-90S warming rate will be higher than your 55/60N-55/60S warming rate. That goes without saying …

But does this peculiar way of doing things make your record correct? Of course not! It makes it completely detached from reality! It’s a bogus record!

The UAH team, on the other hand, seems to have nailed it quite nicely with their version 6. The tight agreement between tlt and OLR, not just globally, but also, as seen here, within separate zonal bands, gives great credence to their temperature product. Their 2000-2015 trend is almost certainly very close to spot on …!

Since the title of this post intentionally replicates the pattern of this one from a month ago, underlining the neat compatibility between the (adjusted) HadCRUt3 and the UAHv6 tlt datasets, this final plot should come as no surprise to anyone. It shows how the global surface temperature anomaly in reality – if you simply go by the right source – correlates wonderfully with that of the troposphere on top of it, with which it is tightly convectively coupled, and, as a consequence, also with Earth’s total radiant heat loss anomaly at the top of the atmosphere towards space. As it bloody well should!

This isn’t so hard. Tropospheric temps follow directly from surface temps, and OLR at the ToA in turn follows directly from tropospheric temps:

pause manifest

Figure 7.

15 comments on “Why “GISTEMP LOTI global mean” is wrong and “UAHv6 tlt gl” is right

  1. oz4caster says:

    Nice work. In Figure 3, why is there such a good match for about 1979 to 1987, but a gradual drift higher for the GISS afterwards?

    Also, I thought the satellite TLT estimates only went to somewhere around 55-65 N and S latitude, so that they have no coverage at the poles. I suspect the forecast system reanalyses have better polar data coverage than either of these sources.

    • okulaer says:

      To me, the slight GISS ‘drift’ seems to set in between ~1997 and 1999 rather, and this might very well have something to do with the introduction of the new ERSST.v4 dataset into the final LOTI product.

      “UAHv6 tlt” has all but full global coverage (99.6%), from 85N-85S.

      Thanks 🙂

      • oz4caster says:

        Thanks for the info on UAH v6. I thought the ERSSTv4 covered the entire span of the GISS estimates and not just the most recent years. Also the SST data should not affect the Arctic since they are extrapolating from land. I’m still puzzled over the gradually increasing discrepancy.

      • okulaer says:

        But you pointed to my “Figure 3” which is not about temp anomalies in the Arctic, but rather within the 55N-55S band, and so this is where a change between SST datasets would indeed affect the total.

        ERSST.v4 covers the entire span of the GISS estimates, yes, but since its trend post ~1996 is considerably steeper than either ERSST.v3b or Reynolds OI.v2 (the SST datasets previously used by GISS), even inside the 55N-55S band, then introducing it will have implications.

  2. markstoval says:

    Another great post. Thanks.

    You sure got Nick Stokes right!

    ~ Mark

  3. javier says:

    Very interesting, and quite thought provoking, thanks.

    I minor quibble. Some of the graphs only have data for about one third of the chosen Y axis. It would be better if they were expanded so we can get a better resolution and so the graph doesn’t appear flattened to make differences less evident.

    • okulaer says:

      Thanks, Javier.

      Actually, I made those plots the way I did for the specific reason that one should not focus on the ‘wrong’ things, that is, those minor differences in between. It matters not here that there is some divergence between GISTEMP and UAH even within the 55N-55S band. The two clearly track each other closely (even if not perfectly close), which is all I wanted to point out, because what that shows is that the GISTEMP curve obviously includes the oceanic part and not just the land portion. In the 90-55N band, on the other hand, it is equally obvious that the GISTEMP curve is only made up of land data, because it rises so much more/faster than the tropospheric (UAH) curve.

      And that’s what this post is about.

      • Do not focus on the “wrong” things. Here are those minor differences:

        I can see why you just wanted to expand the y-axis to hide the divergence after 1998.

        For the rest of this bizarre reasoning: You are saying that when there is sea ice, the temperature for the air above the ice cannot change. Must stay at -1.8 C. Always. All year.

        Believe that.

        UAHv6 is obviously in deep trouble. Like RSS. There is a drift after 1998 that just don’t make sense.

        Ps: UAH has global coverage 90S – 90N. They get that by interpolating over the poles. The unacceptable method in Gistemp. But it is obviously ok for UAH.

      • okulaer says:


        All your quibbles have already been answered in the post itself and in the comments.

        You say:
        “UAHv6 is obviously in deep trouble. Like RSS. There is a drift after 1998 that just don’t make sense.”

        No. As I’ve shown now over several posts, this one included, UAH and RSS are quite definitely correct:

        What doesn’t make physical sense is the post 1996 ‘drift’ of the current surface records. They are, to use your words, “obviously in deep trouble”.

      • okulaer: Your point was UAH v6 55S – 55N lines up with gistemp 55S – 55N.

        Not so. You had to change the y-axis to hide the divergence.

        OLR is not a measure for the temperature on earth. Non-increasing OLR demonstrates that the measured temperature increase cannot be attributed to increased solar insolation (less clouds/increased solar output).

        You can of course continue to claim that temperature over sea ice cannot change. Your choice.

  4. jerry l krause says:

    Hi Kristian and ehak,

    First, Kristian, is it okay that I correspond on your site. For previously it seemed that you stopped, for some time, posting and commenting on your own blogsite. While only you know the reason for this I consider the possibility that it was because you did not wish to continue our conversation. So if you would prefer I not comment on your site, I will not.

    Ehak, you only recently appeared on Roy Spencer’s blogsite and some regulars there immediately began to question who you were. And I believe they probably did identify you. And I researched who they suspected you were. So, I consider you are a person with significant accomplishments. And I like to have conversations with people with actual significant accomplishments.

    And even more recently, RichardLH began commenting on Spencer’s site. I have learned that he appears to be a person who has had significant accomplishments also. I consider the basis of all three of your efforts is computer analysis of averaged data and that you all consider what you do to be science (or scientific).

    I like to quote people of accomplishment. “Intuitive knowledge keeps pace with accurate definition.” was a common saying quoted by the publishers of Dialogues Concerning Two New Sciences.

    Kristian reviewed that Stokes had written: “The SST products OI V2 and ERSST, used by GISS then and now, adopted the somewhat annoying custom of entering the SST under sea ice as -1.8°C.” Of SST, OI V2 , ERSST, and GISS, the only one I know the definition of is SST (sea surface temperature). Hence, I must ask: What does the SST under sea ice have to do with climate? It would seem it is the surface temperature of the ice surface which has any relationship to climate. Now, it is a fact that as long as there is ice, its surface temperature cannot exceed 0 C. But relative to any analysis of climatic factor–temperature–it is not the surface temperature that is commonly considered. It is the temperature of the atmosphere a meter or two above the earth’s land surface that is commonly observed and considered in any ‘climatic’ analysis. Relative to actual SST of liquid seas and the temperature of the atmosphere above these liquid seas, I do not know which temperature is commonly considered.

    However, it is a fact for sea ice to form the temperature of the atmosphere in contact with the surface must be less than –1.8 C for that is the approximate freezing point of the salty sea water. And I know once that sea ice is formed it creates a solid insulating layer and the surface temperature must drop below this temperature for ice to continue to be formed at the base of the sea ice. The result of this insulating layer of sea ice is that sea ice does not continue to freeze at the base of the sea ice beyond a thickness of about 6 feet. This because the temperature of atmosphere in contact with the surface of sea ice does not drop beyond say –100 C (173K). Actually, I am reasonably sure that it never in the case of sea ice its minimum surface temperature is well above this temperature.

    Now to understand climate, one might try to answer the interesting question: Why cannot the sea ice surface temperature drop to such a minimum temperature?

    Have a good day, Jerry

    • jinghis says:

      Jerry, allow me to add a little more to your analysis.

      Obviously any combination of ice/water is going to be around 0˚C.

      The temperature question comes when the water is completely ice and snow covered, which is much of the time. If there is no solar insolation the surface temperature will plummet below zero. Where it is interesting is when the surface is frozen but exposed to sunlight, then the temperature just above the surface can soar from the reflected solar radiation. . .

      CO2 only absorbs in the IR band, reflected sunshine does not contribute to CO2 Forcing. . .

      CO2 forcing can only be applied to the actual surface temperature which is always Zero or below.

      • jerry l krause says:

        Jinghis, maybe I am confused because I considered what Kristian was referring to had nothing to do with any greenhouse effect. I considered it was all about the determination of the earth-atmosphere system’s average temperature.

        Correct me if I am wrong, Jerry

      • jinghis says:

        Jerry, the way the CO2 Greenhouse works is the shortwave radiation goes through the CO2 in the atmosphere and is absorbed by the water. Then the water radiates in the IR band which is absorbed by the CO2. That radiation absorbed by the CO2 is the Greenhouse effect.

        The issue here with the polar oceans (and all the other oceans too) is that the surface temperature doesn’t increase directly with increased shortwave solar radiation. The ocean cools almost 100% via evaporation not radiatively (the net radiation between the ocean surface and atmosphere is very close to zero). The ocean surface temperature stays relatively constant solely controlled by the rate of evaporation and solar radiation, period.

        If the oceans are warming (fewer clouds), CO2 has no effect on the process. . . And CO2 has no effect on the the rate of ocean cooling. The oceans are 84% of the earths energy budget.

  5. jerry l krause says:

    Hi Kristian and ehak,
    I now understand, at this point, my comments are way too detailed and therefore lengthy. But I cannot understand how you can make such an ado (fuss) about tenths of a degree when you know the data upon which these tenths of a degree is so limited; hence questionable. Why do you do what you do?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s