My calculation uses a 3D integration of GHCNV4 and HadSST3 temperature data. If I use GHCN V4C (corrected data) I get 0.93C for September, whereas if instead I use the uncorrected V4U I get 0.70C. In the first case 2020 is the warmest September ever recorded, while the uncorrected data shows 2016 to be just a bit warmer.
Comparison of the corrected and uncorrected temperatures.
Here are the annual temperature comparisons, where 2020 is simply the average of the first 9 months.
Annual temperatures where 2020 covers the first 9 months
2020 is on track to be the warmest year although only marginally so for the uncorrected data. The values though have reduced slightly since August.
Finally here are the spatial distributions for the Northern and Southern Hemispheres
Spatial temperatures for October 2020
Thanks again, Clive. I do have one question, and apologise if it’s been dealt with before: is there any methodologically-valid estimation of error or confidence that could be presented with the graphs? I ask because I wondered about the significance, within the precision of your estimates, of the “slight” differences between Sep temperatures for 2020 and 2019.
Cheers.
I think all you can do is compare the different methods to compute the global average.
Clive
“If I use GHCN V4C (corrected data) I get 0.93C for September, whereas if instead I use the uncorrected V4U I get 0.70C.”
That is a surprisingly large difference. Using the same anomaly base of 1961-90, and the same GHCN V4 sets, but with ERSST, I get 0.843°C using V4U (my usual posted source) and 0.87°C using V4C (adjusted). The choice of SST does not affect the difference, since it is common to both.
I think perhaps your method (LOESS) naturally smoothes out the uncorrected data. This may be why you get almost the same result for both. I take the recorded temperatures at each triangle vertex and then calculate the area average.
Clive,
I was quoting numbers from the mesh average method, which is the same as yours. I show LOESS results too, which are much the same as mesh. I don’t think spatial smoothness is a differential issue, but even if it were, it would have no effect on a global average, which is the ultimate in smoothing.
Nick,
You can see there are some cold spots in the raw data. Especially one large patch in Africa
a better image for Africa
Clive,
There aren’t many stations in that region of Africa, so I’ve looked through on the GISS station explorer:
https://data.giss.nasa.gov/gistemp/station_data_v4_globe/
Can’t see any which should produce that signal in the unadjusted data. Can you pinpoint which station(s) is giving this result for you there?