A reader asked for the actual rates at which various global temperature data sets (featured in this post) are increasing, after one removes the estimated impact of ENSO (the el Niño southern oscillation), volcanic eruptions, and solar variations.
I welcome such requests, but caution strongly that I can’t fill them all, or even most of them. It’s too much work. But in this case, it’s a pretty simple request and I’ll go for it.
Here are the warming rates for the 7 global temperature data sets from the aforementioned post. The first 5 are surface temperature, the last two are satellite-based estimates of TLT (Temperature in the Lower Troposphere).
The numbers are the rates themselves, the vertical bars are those rates plus or minus two standard deviations. Note that all values are in °C/century. I usually give rates in °C/year, but I thought the 100-year rates would be more familiar and meaningful for some readers.
Rates for surface temperature data cover the time span from 1970 through 2017; those for satellite-based troposphere temperature the period 1979 through 2017 (satellite data don’t begin until 1979).
This blog is made possible by readers like you; join others by donating at My Wee Dragon.
Open Mind 
How do you come to choose a century? I can understand avoiding degrees per year.. but wouldn’t nearly the same data describe about a half century?
100 years, like 10 years and 5 years, is just a nice “round” number. 50 is a touch less “nice” in that respect, I would say.
Or, you could also think in terms of common long-term climate projections, which tend to be something like “by the end of the century” (of course, 2100 is not 100 years from now, but it’s close enough to wave a hand for).
Alternatively, depending on how much space one wants to allocate for a graph’s axis, rescaling the y-units to reduce the number of required significant digits can help with visuals. It’d be more difficult to fit 0.010 … 0.012 … 0.014 etc. onto the y-axis than 1.0 … 1.2 … 1.4. (The proportions of the graph would remain the same either way in this case, so long as the relative y-limits are kept.)
Richard – to get the warming per half century, just divide all the numbers by two.
How would that tell us when a linear trends is becoming exponential?
It doesn’t. It’s an average over a period (1970-2017, or 1979-2017). But that’s a different period from the units divisor (century/decade/year). It seems like you might be confusing those two things? We have to take shorter periods and compare them to detect accelerations. But the units used for that make no difference (the units remain per-century/per-decade/per-year, and they all work just the same, just change the units).
Many thanks once more Tamino, for this information in reply to a recent question. However this answer, though satisfying, raised in turn the further question of why the trend differences you computed differ by a lot from those computed by Santer & al. a few years ago:
(i.e. page 52 in the free access version of their article):
https://dspace.mit.edu/openaccess-disseminate/1721.1/89054
As noted in my earlier comment, the RSS3.3 time series chosen by Santer & al. gave at that time (end of 2012) a linear estimate of 1.24 °C / decade, what shows that their adjustment (0.86 °C / decade) was about 30 % less than the original trend.
Let us use Kevin Cowtan’s trend computer
http://www.ysbl.york.ac.uk/~cowtan/applets/trend/trend.html
to obtain the actual, unadjusted trends for those time series you performed the adjustements on (in °C / century):
– NASA: 1.80 ± 0.28
– HadCRUT: 1.75 ± 0.27
– NOAA: 1.75 ± 0.28
– C&W (HadCRUT with kriging): 1.88 ± 0.38
– BEST: 1.84 ± 0.36
– RSS (4.0 TLT): 1.91 ± 0.62
– UAH (6.0 TLT): 1.28 ± 0.60
We see that your adjustments result on average in a decrease of a best 5 % compared with KC’s original ones. They are therefore way smaller than those computed by Santer & al.
I would have pretty good understood such results if your adjustments had been performed on land-only records, as these show much higher trends. But you specified ‘global’ so I suppose you worked on land+ocean time series.
Thanks in advance for helping me in understanding the discrepancies.
[Response: Both RSS and UAH have undergone major revisions recently, which changed their estimated trends considerably. As for NASA, NOAA, C&W, Berkeley, when you remove the estimated effect of exogenous factors it reduces the trend estimate, in large part because the strong el Nino events of 1998 and 2016 are in the latter part of the record, major volcanic eruptions in the earlier part, and solar irradiance has declined (slightly) recently.]