Why Are Surface and Satellite Temperatures Different?

Many are puzzled by the difference between surface temperature, measured in Stevenson screens, and atmospheric temperature, as measured by satellites. Some sceptics suspect surface temperatures cannot be trusted; some global warming enthusiasts claim satellite data are not accurate. The truth is both are accurate enough to be useful for their own purposes. But why the difference?

I have used data from the Bureau’s Climate Change Time Series site for monthly rainfall and surface temperatures for Australia, and from University of Alabama-Huntsville (UAH) for Temperature of the Lower Troposphere (TLT) anomalies for Australia, from December 1978 to October 2015. I converted rainfall and surface temperatures to anomalies from monthly means 1981 – 2010, the same as UAH. Throughout I use 12 month running means.

Firstly, surface temperatures are supposed to be different from atmospheric temperatures. Both are useful, both have limitations. The TLT is a metric of the temperature of the bulk of the atmosphere from the surface to several kilometres above the whole continent, in the realm of the greenhouse gases- useful for analysing any greenhouse signals and regional and global climate change. Surface temperature is a metric of temperature 1.5 metres above the ground at 104 ACORN-SAT locations around Australia, area averaged across the continent- useful for describing and predicting weather conditions as they relate to such things as human comfort, crop and stock needs, and bushfire behaviour.

The context:

This map shows the location of the Acorn surface temperature observing sites.

Fig.1: ACORN-SAT sites

Acorn network

Note the scale at bottom left, and that they are concentrated in the wetter, more closely settled areas. As with rainfall observing sites:

Fig.2: Rainfall observation sites

Rainfall network awap

Fig.3: mean annual rainfall:

Avg ann rain map

The scale is in millimetres: divide by about 25 to get inches. Consequently a very large area of Australia is desert, and another large area is grassland with few or scattered trees. Very little of Australia is green for more than a few months of the year. More on this later.

The data:

Here are 12 month running means of the Bureau’s Acorn maxima and minima since December 1978.

Fig. 4: 12 month running means of monthly maxima and minima anomalies (from 1981-2010 means) for Australia

Max v min

Note that minima frequently lags several months behind maxima- which is why mean temperature doesn’t give us very much useful information.

Now compare surface temperature with the lower troposphere:

Fig. 5: Minima vs TLT anomalies

min v uah

Fig. 6:  Maxima vs TLT anomalies

max v uah

TLT approximately tracks surface temperature, but with smaller variation. So what causes the difference between surface and atmospheric temperatures?

The culprit is that wicked greenhouse gas, H2O.

In the following graphs 12 month mean rainfall is scaled down by a factor of 25, and inverted: dry is at the top and wet is at the bottom of these plots.

Fig. 7: Maxima vs Inverted Rain

max v rain

It is plainly obvious that very wet periods mostly coincide with low maxima, and dry periods with high maxima.

Fig. 8: Minima vs Inverted Rain

min v rain

Again, minima has no immediate relation with rainfall (although cloudy nights are warmer), lagging many months behind.

Next I calculate the difference in anomalies- surface temperature minus TLT- to analyse the difference between surface and satellite data. As minima lags many months behind rainfall a close relationship is not expected.

Fig.9: Acorn minima anomalies minus TLT anomalies compared with rainfall anomalies

min diff v rain

However, Acorn maxima minus UAH matches rainfall remarkably well.

Fig.10: Acorn maxima anomalies minus TLT anomalies compared with rainfall anomalies

max diff v rain

It is not an exact match of course. The next graph plots the surface maxima- TLT anomaly difference against 12 month mean rainfall anomaly sorted from smallest to largest, with the horizontal axis showing monthly percentile rank by rainfall:

Fig.11: Comparison of maxima-TLT anomaly difference with ranked rainfall anomalies

Max-UAH v rain%

Note that the surface- atmosphere difference tracks rainfall quite closely (+/- about 0.5C), with the largest positive and negative differences at the rainfall extremes, and also that the 12 month period where the rainfall anomaly crosses from negative to positive is at the 59th percentile: there are more dry months than wet months.

Another way of showing the relationship is with a scatterplot:

Fig.12: Surface maxima- TLT difference compared with rainfall

max diff v rain scatterplot

Note the R squared value: 0.76! At least three quarters of the difference can be explained by rainfall variation alone- not bad across a whole continent with a northern wet summer / dry winter and a southern wet winter / dry summer pattern.

An over simplified explanation of a complex process:

In wetter than normal weather, more and thicker clouds reflect sunlight and shade the surface, keeping it cooler than normal. Moisture from the surface (and vegetation) is evaporated, also cooling the surface. Deep convective overturning occurs during the day and evaporated moisture ascends in the atmosphere, where it condenses, releasing heat. The troposphere anomaly is thus relatively warmer than the surface anomaly in moist conditions such as during wet weather.

In a drought, fewer clouds allows more sunlight to heat the surface. The ground is dry; surface water is scarce; vegetation is thinner, drier, and shades less of the ground. Therefore the surface is hotter than normal. Less evaporated moisture means less condensation releasing heat in the troposphere, and therefore the troposphere anomaly will be relatively cooler than the surface anomaly.  As well, as the Bureau explains, ” the rate at which temperatures cool with increasing altitude (known as the lapse rate) is greater in dry air than it is in moist air.”  Thus in dry weather, ignoring convection, the atmosphere will be cooler than normal.

Yes, but…

So how does this explain why the October 2015 surface maximum anomaly was a record +3.08C above the 1981-2010 mean, while the UAH anomaly was a mere +0.71C, and the rainfall anomaly was only -12.75mm, nowhere near the lowest?

This map shows the Normalised Difference Vegetation Index for October. The Bureau explains the index as a measure of “the fractional cover of the ground by vegetation, the vegetation density and the vegetation greenness”.

Fig.13: Normalised Difference Vegetation Index (NDVI) October 2015

Vegetation Oct 2015

What do the dark brown areas look like on the ground? Here’s a photo I took recently around about the area circled red:

Fig.14: Droughted country, Western Queensland, September 2015

Bare, dry dirt with scattered tussocks of dead grass- scattered prickly acacia in the distance.

A large area of Australia is relatively bare and bone dry, therefore hotter. Over wide areas, much less moisture is convected into the atmosphere, which will thus be relatively cooler than surface anomalies. North winds blowing from the interior towards the south will bring hot dry air even to green areas, causing much hotter surface temperatures there as well. Much of the moisture evaporated from these wetter areas is blown out to sea (outside the UAH Australian grids) so the TLT over even these green areas is relatively cooler than expected.


Atmospheric temperature anomalies are necessarily different from surface anomalies. Usually, atmospheric anomalies are less than surface maxima in hot periods and higher than surface anomalies in cool periods.

There is no conspiracy: over three quarters of the difference between surface and atmospheric temperature anomalies is due to rainfall variation alone.

Tags: , , ,

18 Responses to “Why Are Surface and Satellite Temperatures Different?”

  1. Geoff Sherrington Says:

    Clearly explained, thank you, Ken.
    I have often wondered what this association of Tmax with water does when sensitivities to GHG are calculated.
    If there is a cooling mechanism (rainfall) that is essentially independent of GHG, at least in a first order look, should sensitivities not be calculated using temperatures corrected for the effect of rain?
    Maybe they are, but the path is not so clear if they are. It is really a massive effect and it cannot be ignored.

  2. MikeR Says:

    Hi Ken,

    Figure 12 is amazing!. That is a great pickup! Explains a lot of things about the difference between UAH and land based measurements and the influence of water vapour. Are Spencer and Christie aware of your data? This should be widely disseminated.

    It would also be good to see if the USA data shows the same trends but, as you say, it may be also related to soil and land usage which presumably is different for the US versus Australia.

    Sorry about some of our previous exchanges which have on occasions been a bit heated but I recognise good work when I see it.

    • MikeR Says:

      Finding I have a little bit more time on my hands I just have had another look at your data.

      I think your material above may be relevant to the discussion at http://www.drroyspencer.com/2014/10/do-satellite-temperature-trends-have-a-spurious-cooling-from-clouds/ and http://link.springer.com/article/10.1007%2Fs00382-013-1958-7 , despite these discussions being about clouds and precipitation over oceans rather than land.

      Interestingly the Australian annual data for Tmax-UAHv6 and pan evaporation also show a similar correlation of Rsquared of 0.74 (positive slope) compared to the correlation for rain of 0.78 (negative slope). Cloud cover also correlates with value for Rsquared of 0.56.(negative slope).

      It appears that the discrepancy between surface and lower tropospheric satellite temperatures are, as you point out related to rain and/or water vapour.

      • kenskingdom Says:

        I haven’t looked at evaporation or cloud in this context so those results are interesting. I’m not surprised- cloud can act both ways and pan evaporation very much depends on humidity, therefore similar correlation to rain.

        • MikeR Says:

          The major difference is that the correlation is in the opposite direction for evaporation as compared to rain.

          I have also had a look at the USA48 data and it also shows on an annual basis a significant correlation (Rsq=0.41) between Tmax-UAH versus Rain. The slope was negative as for Australia.

          Finally using the BOM data for global temperature and rainfall I found that T mean- UAH versus Rainfall was correlated for the land data with Rsq=0.23 (negative slope). Despite the lower Rsq this is significant as the linear regression statistics had a P-value of 0.0033 much lower than the 95% confidence value of p= 0.05. Interestingly doing the same for the UAH ocean data the correlation was much smaller (Rsq =0.12) but with a positive slope. This result is not quite at the 95% confidence level (p=0.079) but getting close.

          I guess these results are not that unexpected as the impact of rain upon T mean for land is going to be much greater and different than its impact upon the temperature of the surface of the sea. Differences in evaporation are also likely to be very different as a consequence.

  3. kenskingdom Says:

    Hi Mick
    Your apology is accepted. Yes I briefly discussed my initial findings with John Christy two years ago, and he put me onto the convective overturning- but I promptly got too busy with other issues to explore it properly. Another factor with USA might be snow.
    The important thing to remember about any dataset is not whether individual months or year are records but what the broader data tell us. Which is why I prefer satellite data as a measure of what the bulk atmosphere is doing.

  4. Neville Says:

    Ken I understand that UAH is measuring the lower troposphere and BOM is using much sparser ground measuring system. But shouldn’t the UAH data still be showing faster warming than ground temps? IOW when we consider GHG theory shouldn’t temps measured at the surface be warming at a lower rate than UAH TLT? But this doesn’t seem to be the case, in fact ground temps are warming faster. Doesn’t this disprove their theory?

    • kenskingdom Says:

      Unfortunately for that argument, the TLT trend is much greater than the Australian surface trends over the satellite era, which tends to warm the cockles of global warming enthusiasts.
      However in GHG theory Troposphere anomalies should be increasing as higher surface temperatures-> greater evaporation-> higher tropospheric humidity-> higher TLT anomalies plus greenhouse amplification due to greater H2O downwelling radiation. These results suggest this is false (at least above Australia) for two reasons. 1. It should have a stronger effect at night, on surface minima, whereas the correlation is higher with maxima, in fact rainfall has no apparent correlation with the minima- TLT difference (R2= 0.004!) and the Acorn minima trend is about half that of maxima (0.8 vs 1.7 /100 yrs). 2. Higher surface maxima are demonstrably associated with lower TLT, and also with lower rainfall. Higher temperatures are not the cause of either more droughts or more rain, but rainfall variation affects both surface and TLT anomalies.

  5. Neville Says:

    In a 2014 study Ross McKitrick looked at Balloon data from the tropical troposphere from 1958 to 2012. He found that nearly all the warming over that period came from one shift change ( PDO ?) in Dec 1977 and warming before and after that shift was not statistically significant. Don’t the balloon and satellite records show similar trends since 1978? Here’s his summary————

    Bottom Line

    “Over the 55-years from 1958 to 2012, climate models not only significantly over-predict observed warming in the tropical troposphere, but they represent it in a fundamentally different way than is observed. Models represent the interval as a smooth upward trend with no step-change. The observations, however, assign all the warming to a single step-change in the late 1970s coinciding with a known event (the Pacific Climate Shift), and identify no significant trend before or after. In my opinion the simplest and most likely interpretation of these results is that climate models, on average, fail to replicate whatever process yielded the step-change in the late 1970s and they significantly overstate the overall atmospheric response to rising CO2 levels.”

    And here’s a discussion of that study at Climate Audit at that time. http://climateaudit.org/2014/07/24/new-paper-by-mckitrick-and-vogelsang-comparing-models-and-observations-in-the-tropical-troposphere/

  6. kenskingdom Says:

    The UAH satellite record December 1978 to October 2015: Global- +1.14C/ 100 yrs; Australia- +2.34C/ 100 yrs, both +/- 0.1C. Both show flat trends to the early 1990s, a sharp step up, then flat or declining trends since about 2001. For the Tropics, the trend since 1978 is +1.02C/ 100 yrs. I understood McKitrick found the Pause (no significant warming) goes back to the late 1980s.

  7. Neville Says:

    Thanks for that Ken, I think RSS global trend is 1.2 c/ 100years. But the near similar trend for UAH tropics is very strange if you believe that co2 is the major driver of CAGW.
    In fact your UAH data above shows that the tropics actually show a 0.1 c lower temp per century than the globe. Shouldn’t the tropical troposphere be the area that warms at a much faster rate according to AGW theory? Do you have any idea why OZ should be showing about twice the warming trend compared to the rest of the planet ? If it’s co2 it must have very snazzy and magical powers to double just the OZ trend surely? I wonder how NZ trend has fared since 1978?

  8. Neville Says:

    Ken here is NOAA radiosonde balloon data with measurements from surface (100 mb) to 300mb to 800mb ( 300mb is nearly 10 klms above the surface)


    According to Tony Heller at Real Science that radiosonde surface data shows no warming from 1978 to 2010. Here is his post —————–

    • reichforthesky Says:


      Last time I looked 100 mb is at about 16,000 metres above the surface i.e. in the stratosphere.

      The steven goddard site you have linked to shows the radiosonde data for the surface to 100 mbar. Consequently it includes a significant part of the stratosphere. All global warming models , I gather, predict cooling in the stratosphere.

      So it is little wonder that this data, because it includes the cooling stratosphere as well as the warming low troposphere, shows no significant change in temperature.

      The other point of interest is that the original data shows very strong negative trends for pressure ranges from 300 mbar to 30 mbar. For 300-100mbar the, trend is -0.22 degrees C/decade, for 100-50 mbar the trend -0.6 C/decade, while for 100-30 mbar the trend is -0.44 C/decade. This is all consistent with climate change modelling.

      Finally the UAH TLT data corresponds to a height of about 3000m or 700 mbars, see https://en.wikipedia.org/wiki/Satellite_temperature_measurements).
      This is another reason that the above radiosonde data has no real relevance to discussions about UAH data.

  9. Neville Says:

    Ken here is NOAA radiosonde balloon data with measurements from surface (100 mb) to 300mb to 800mb ( 300mb is nearly 10 klms above the earth)


    According to Tony Heller at Real Science that radiosonde surface data shows no warming from 1978 to 2010. Here is his post —————–

  10. Neville Says:

    Sorry about that second comment above, I don’t know why that occurred.

  11. Neville Says:

    A 2014 Santer study that removes both ENSO and Volcano effects seems to show that the pause started in 1993 . Over 20 years ago.


  12. UAH, ACORN and Rainfall: Something’s Wrong | kenskingdom Says:

    […] as I showed in my post “Why are surface and satellite temperatures different?”  in 2015, most of the difference between UAH and BOM maxima can be explained by rainfall variation […]

Comments are closed.

%d bloggers like this: