It’s been a while since Foster & Rahmstorf (2011) took global temperature time series and removed our best estimate of the changes due to known fluctuating factors, the el Niño Southern Oscillation (ENSO), aerosols from volcanic eruptions, and variations in the output of the sun. After removing fluctuations of known origin, what was left over was a very steady rise in global temperature.
I’ve improved (I believe) the method by allowing for more detail in the response to ENSO. I now allow both a “prompt” (not necessarily immediate) and “more delayed” response, as well as a seasonal pattern to the ENSO response. I’ll probably expound on those details at some point, but not now. Now, let’s look at the results.
I studied temeperature since 1950 (when we have MEI, the Multivariate el Niño Index), for all the datasets studied in Foster & Rahmstorf: five surface temperature data sets and two satellite sets for lower-troposphere temperature. They are:
NASA: from NASA’s Goddard Institute for Space Studies
NOAA: National Oceanic and Atmospheric Administration
HadCRUT4: Hadley Centre/Climate Research Unit in the U.K.
Cowtan & Way: New method for HadCRUT4 base data
Berkeley: Completely new method, all-inclusive data
RSS TLT4: Lower-troposphere temperature from RSS
UAH TLT6: Lower-troposphere temperature from UAH
Let’s begin by looking at what these data sets say before “adjustment” to remove known fluctuations. Here are annual averages for the five surface temperature data sets:
Clearly the different data sets are largely in agreement. They show temperature just fluctuating, not really going anywhere, until about 1975 when it starts to increase. Of course it keeps fluctuating too; the most notable, to my eye at least, are the peak in 1998 (due to the strong el Niño of that year) and the peak in 2016 (due to the strong el Niño of that year too).
Removing my best estimate of the influence of known fluctuating factors, the annual averages show less fluctuation so the trend becomes clearer:
To each surface temperature data set I fit a model consisting of two straight lines, one for the early standstill, the other for the rise since about 1975. I actually found the best-fit “turning point” time for each series. Then I recorded the recent rate and its uncertainty. For the satellite data sets, I just fit a straight line. Here are the rates I found, with error bars (2σ):
Most of the data sets agree on the recent rate with the notable exception being the UAH TLT data, clearly an outlier. The NOAA data give a slightly lower rate, while NASA, Cowtan & Way, and Berkeley give a higher rate, and the RSS TLT data slightly higher still.
We can use the five surface data sets to define an average, then look at how each individual series deviates from that average. Without further ado, here they are:
For the most part the deviations stay between -0.1 and +0.1 °C. Relative to the others, HadCRUT4 and Cowtan & Way started high while NASA and NOAA started low. Most recently (2018), NASA is high and HadCRUT4 is low.
When we compare the lower-troposphere satellite data to our five-set surface average (from 1979 when the satellite data begin) the deviations are quite a bit larger, reaching about 0.3 °C.
The deviations are bigger for UAH than for RSS, and UAH seems to show a sizeable downward trend (relative to the surface average) while RSS might show a slight upward trend.
If we study deviations from the five-set surface average, but using adjusted rather than raw data (for both the satellites and the surface data), the deviations are meaningfully reduced:
We conclude that even after removing the influence of known factors, the lower-troposphere temperature series show more fluctuation than the surface temperature data sets, and while the RSS TLT data show a slight upward trend relative to the surface, the UAH TLT data show a big downward trend with a more complex pattern. In terms of modern warming rates, UAH TLT is an outlier.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.