CMIP6 temperature cycles

Nick Stokes has an interesting post comparing a blended version of CMIP6 tos/tas  to compare models to data. The blending is intended to correct for the fact that  Ocean temperatures are measured at the surface (tos) while land temperatures are measured 2m above the surface (tas). He made available the CMIP6 model data in an easy to use csv file. So I thought it would be easy to compare CMIP6 model results to my global temperature monthly data. However I discovered another effect.

CMIP6 monthly temperatures (Univ. Melbourne). Note the regular monthly oscillation in all models . For example the yellow post 2100 model signal (MRI ESM2)

Each model produces a monthly cycle of changing temperatures like a sine wave. The models work in absolute temperatures and you typically find a variation of 4 or 5 degrees during each year. The reason for this is probably 2 fold. Firstly there is an asymmetry between the land/ocean areas in the  Northern Hemisphere to that in the Southern Hemisphere. Secondly there is a slightly elliptical orbit of the earth (perihelion/aphelion). This causes the Southern Hemisphere summer to be about 4.1 million miles closer to the sun than the Northern Hemisphere summer. So why don’t we see this in the measured data?

The  fact is that none of the global temperature data show any sign of such an oscillation, although it is observed in Meteorological ‘reanalysis’ data. The main reason for this is that everyone always work in temperature ‘anomalies’ so they just calculate deviations from a ‘normal’ monthly average. I always take a 30 year normalisation period for weather stations of 1961-1991 which is the same as HadSST3/4. Let’s compare all 102 CMIP6 models with the data from Nick’s csv file where I calculate model “anomalies” on the same baseline.

Global temperature anomaly data compared to 102 CMIP6 model runs with tos/tas blending. The purple arrow shows the normalisation period for both.

The data are up to March 2012. The agreement appears reasonable good partly because they have been normalised to the same baseline. However more recent temperature data favours those models running slightly cooler.

What I find interesting though is that global temperatures actually change by ~5 degrees every year.  This is then superimposed onto an overall global “warming” of just 1 degree over the last 180 years. It is only by using temperature anomalies that this small effect can even be measured.  The above monthly model  temperature data are derived form monthly temperature data that actually looks like this !


About Clive Best

PhD High Energy Physics Worked at CERN, Rutherford Lab, JET, JRC, OSVision
This entry was posted in AGW, climate science. Bookmark the permalink.

7 Responses to CMIP6 temperature cycles

  1. entropicman says:

    You see a similar seasonal cycle in the GISTEMP monthly data.

    • Clive Best says:

      it seems to be MERIS data onto which they subtract/add the global temperature anomaly. It is not measured directly.

      Update: Sorry this is reanalysis data from meteorological models using data assimilation to match satellite and weather sensors over a fixed period

      The GISTEMP temperature series can only produce monthly averaged t
      temperature anomalies. These are then simply added or subtracted to the MERRA2 profile for the given year and month.
      The temperature profile always remains fixed.

  2. entropicman says:

    That would fit. The GISS seasonal changes graph gives monthly temperatures as anomalies but to detect the annual cycle it would need to be based in an absolute temperature dataset.

    It is probably a good sign that CMIP6 has picked up the annual cycle. It shows that the model resolves regional differences well enough to detect the differences due to the land imbalance between hemispheres and the orbital insolation variations.

  3. Chic Bowdrie says:

    Clive, I don’t understand “superimposed.” Is that the same as saying despite the 5 degree swing back and forth due to the hemispherical differences, the swinging disappears when averaged?

    Also, the last sentence is a bit confusing. Is it exactly correct or are the words “absolute” and “model” missing so that it should read, “The above monthly model anomaly temperature data are derived from monthly absolute temperature data that actually looks like this.

    Is the cyclical nature of CO2 emissions a possible third reason for the NH/SH swings?

    • Clive Best says:

      Sorry – yes the data is not superimposed. Basically each month of GIStemp anomaly data is added or subtracted to the single MERRA2 profile.
      It gives the impression that the long term GIStemp data measured profiles back to 1880, but of course it didn’t.

      For the models I actually do the inverse. I calculate 12 monthly normal (average) monthly temperatures over a 30 year period. I then subtract these for each month from each model to end up with model temperature anomalies which I then compare to the classic temperature series (GHCN/V4 or GISTEMP or whatever). This is a better way to compare models to measurements.

      CO2 oscillates with seasonal vegetation growth and decay in the northern hemisphere. There is a small feedback effect but not large enough to explain these large temperature variations. I think the sun is responsible.

  4. frankclimate says:

    Hi, if the orbit has this strong impact on models it would be striking: The 2nd law of Kepler demands a higher speed of the earth on perihelion which limits the impact of a bigger TSI at this part of the orbit. And indeed the annual cycle of the CMIP5 is overestimated ir relation to the absolute SST-data of ERSSTv5 for 1940-1970 (to avoid stronger ERF:

  5. The annual variation in temperature on the equatorial band has a fundamental influence on ENSO behavior. Every year a maximally sensitive thermocline density gradient occurs which is then forced by the transient tidal factors occurring at that time,

Leave a Reply