By Roger Andrews
HadCRUT4, a joint production of the UK Met Office Hadley Centre and the Climatic Research Unit of the University of East Anglia, is the world’s “official” global surface temperature time series. It’s the time series that tells us how much the Earth has warmed. It’s the time series the IPCC uses to “verify” its climate models and to support its claim that the warming was dominantly anthropogenic. It’s the time series that underpins the world’s efforts to cut carbon emissions and transition to a “sustainable” energy future. It’s difficult to overstate its importance.
Figure 1: HadCRUT4 “global surface temperature” time series
One hopes and trusts it’s reliable.
Unfortunately, however, it isn’t.
HadCRUT4 exists because it’s believed that we need a single global surface temperature time series to properly evaluate global warming. We don’t have a consistent set of temperature measurements covering all of the earth’s surface, but we do have surface air temperature measurements in land areas and sea surface temperature measurements in ocean areas, and so we construct what is believed to be a representative global surface temperature time series by combining the two.
We therefore construct HadCRUT4 by taking an area-weighted average of the CRUTEM4 land surface air temperatures, which contribute ~30% of the final HadCRUT4 value, and the HadSST3 ocean sea surface temperatures, which contribute the remaining ~70%. The three series are plotted in Figure 2:
Figure 2: CRUTEM4 land, HadSST3 ocean and HadCRUT4 land+ocean time series
And immediately we see that CRUTEM4 shows roughly twice as much warming as HadSST3 since 1890 (and also since 1970), raising the question of whether HadCRUT4 isn’t averaging apples and oranges. The response is to present comparisons like Figure 3, which shows a good match between HadSST3 and HadMAT surface air temperatures (measured on board ship) over the oceans. It’s claimed that such comparisons confirm that the differences between CRUTEM4 and HadSST3 occur simply because the land warms faster than the ocean and not because of any fundamental difference between SSTs and surface air temperatures:
Figure 3: HadMAT marine air temperature series vs. HadSST3
These results indeed look fairly convincing if we take the Figure 3 data at face value. The problem, however, is that they have been “corrected”. Figure 4 shows what the comparison looked like before the “corrections” were applied.
Figure 4: Figure 3 data before “correction” (mean both data sets = 0)
If corrections this large are needed to support the claim that there are no fundamental differences between surface air temperatures and SSTs then clearly the claim clearly isn’t a very robust one. The fact that the specific goal of the corrections was to match the SSTs to the air temperatures also doesn’t help. As a result we can’t dismiss the possibility that HadCRUT4 really is averaging apples and oranges, in which case it won’t be representative of anything, except possibly a lemon.
The IPCC also uses a “HadCRUT4 equivalent” variable rather than direct climate model output to project 21st century surface warming. The AR5 projections come from the CMIP5 suite of climate models, which directly output only one global temperature variable – surface air temperature (tas). They can also output tas over land (equivalent to CRUTEM4) and tos over the oceans (equivalent to HadSST3), but these data sets aren’t global. Figure 5 plots the CMIP5 model means through 2100 (RCP85 scenario) for these three variables. They give quite different warming projections:
Figure 5: CMIP5 multi-model means, RCP85 case
Which one should be the “official” global surface temperature projection? Global surface air temperature is at least global, but it ignores the oceans, which show considerably less surface warming and which contain over 99% of the heat in the atmosphere and the oceans. Alternatively we could use the HadSST3 projection, but this ignores the atmosphere, which is where we live. Obviously we can’t define” global surface warming” with a single number. But the IPCC insists on doing it this way, so it makes up a “HadCRUT4 equivalent” model variable by averaging tas over land and tos over the oceans in the same way as it averages CRUTEM4 and HadSST3, and then it adopts this number as its official “global surface temperature” projection:
Figure 6: CMIP5 multi-model means plus “HadCRUT4 equivalent”, RCP85 case
But what does this number mean? Nothing. It doesn’t refer to the air, it doesn’t refer to the ocean. It refers to a medium consisting of 70% sea water and 30% air, which doesn’t exist as a physically-definable entity. It’s pure apples and oranges, and the result in this case is unquestionably a lemon. (A good way of looking at it is to assume that the HadCRUT4 projection is accurate and that the warming is evenly spread across the globe. Where on the Earth’s surface will you experience a temperature rise of 4.2C relative to 1860 in 2100? There isn’t anywhere.)
And now it gets worse.
CRUTEM4 and HadSST3 have been “corrected”.
It’s not possible to say exactly what corrections have been applied to CRUTEM4 because no one publishes a land surface air temperature time series that uses only raw records. However, a few years ago I put together a surface air temperature time series from scratch using ~1,000 selected and unadjusted Global Historic Climate Network (GHCN v2) records, and Figure 7 compares it with CRUTEM4:
Figure 7: CRUTEM4 vs. author’s global surface air temperature series
The short-term fluctuations match closely, but CRUTEM4 shows about 0.3C more overall warming since 1890. It’s likely that this warming is manufactured by “corrections”, in which case CRUTEM3 will show about 0.3C more surface air temperature warming over land areas during the 20th century than has actually occurred, but I can’t be sure about this so I can’t call strike three here.
The “corrections” applied to HadSST3, however, can be quantified simply by plotting the difference between HadSST3 and the ICOADS SST series, the raw data set from which HadSST3 is derived:
Figure 8: HadSST3 vs. unadjusted ICOADS SST series
I reviewed these corrections in detail a few years ago (a pdf documenting the results of the work is here) and found that they were based almost entirely on ad-hoc assumptions and backed up by effectively no hard data (as was the case with the Figure 4 data, the goal of the corrections was again to match the SSTs to surface air temperatures). They are certainly wrong to some degree and quite possibly seriously wrong. At the very least a case can be made that a data set that needs corrections this large to make it “correct” was too heavily distorted to have been used in the first place.
Yet HadSST3 contributes ~70% of the final value of HadCRUT4.
And HadCRUT4 is the data set that underpins the world’s efforts to cut carbon emissions and convert to a “sustainable” energy future.
Strike three, and you’re out.
CRUTEM4, HadSST3, HadCRUT4 from CRU
ICOADS SST & marine air temperatures, HadMAT from KNMI Climate Explorer
The author’s surface air temperature time series data are available on request.