Is GISTemp4 running a little hot?

September 2020 was the warmest globally since their record began in 1980 with a monthly anomaly 0f exactly 1°C above the 1951-1980 LTA. This surprised me a little, as I was expecting the recent surge in global anomalies to level off or even fall back. I track the monthly temperature of seven of the world leading estimates of global temperature. To do this I like to keep a record of each 12 month moving average, and as you can see although they report differing anomalies they all keep on the ‘same track’ so to speak. The only wildcard in the pack is the UAH series, they use satellite data to construct their series and this appear to be a lot more sensitive and reactive to changes in the lower atmosphere.
The GISTemp and Berkeley temperature series have always produced very similar results as you can see for the moving average for the last 20 years, that is till around 2011, when they have slowly parted company. We are not talking by massive amounts here, because remember changes in global temperature are measured to two decimal places or more per decade, but the gap between the two has widened to just over 0.05°C over the last nine years. Luckily we can rule out the LTA they use because by chance they both use the same 1951 to 1980 period. So why is it that the GISTemp series has been warming more quickly than the Berkeley series, or conversely why has the Berkeley series not warmed as much as the GISTemp series since 2010? Obviously something looks to have changed in the methodology one of them uses in the algorithm they use to estimate a global monthly anomaly with. As always answers on a postcard please to the usual address on that one.

Because a variety of LTA’s are used in calculating global temperature anomalies, it’s difficult to make direct comparisons between any of them. The best way of doing this that I can think of is to zero, or equalise, all seven of the 12 month average anomalies as I’ve done for September 2020 in the chart below. Now these results will have a statistical bias because all series don’t use the same LTA as I said previously. I am no climatician if there is such a word but this may be causing the divergence between the temperature series which as you can see is not just a difference of 0.05°C but almost 0.5°C over the last twenty years.

Out in the lead at the moment, and by quite a margin in the global temperature sweepstakes, are the new kids on the block, in the shape of ERA5 series from Copernicus and the ECMWF, closely pursued by the possibly overheating GISTemp4 series. In last place with the lowest increase are the much more conservative HadCRUT4 and the JMA series, which have a 12 month average increase since 1990 which is 0.15°C lower (dare I say colder) than the American GISTemp4 series. These results are far from scientific I know, and am sure will be readily dismissed by the powers as statistical bias or something such, but it does make you wonder why they can’t all agree to use the 1981-2010 LTA for starts.

In Conclusion

I have a feeling at the start of 2021 that the Americans, and hence the rest of the world, will announce that 2020 overtook 2016 as being the warmest year on record on the strength of the results from the GISTemp4 series. It maybe that ERA5 will come to the same conclusion, but I have a strong suspicion that the NOAA, JMA or the HadCrut4 series will not, which in the scheme of things will make very little difference to all those headlines.

Scroll to Top