Pierrehumbert recently made the following statement about the truncation of data:
Whatever the source of the purported … data, there is no legitimate reason in a paper published in 2007 for truncating the … record … as they did. There is, however, a very good illegitimate reason, in that truncating the curve in this way helps to conceal the strength of the trend from the reader, and shortens the period in which the most glaring mismatch … occurs.
I totally agree with Pierrehumbert’s condemnation of graphics that are truncated to “conceal” mismatches from a reader. This is a matter that I’ve discussed previously in connection with IPCC and which I would like to review today in light of Pierrehumbert joining with climateaudit in the condemnation of this practice. Prior discussions of the topic at CA include here here and here .
IPCC TAR
Let me refresh the discussion by showing how IPCC TAR concealed the mismatch between the post-1960 decline of the Briffa et al 2001 reconstruction and temperatures, by simply deleting the post-1960 values of the Briffa reconstruction. First here is a graphic from the original Briffa article showing the “divergence problem” – values of the Briffa reconstruction declined sharply in the 2nd half of the 20th century, such that closing values were similar to those in the early 19th century, long before modern warming.
The Briffa et al 2001 reconstruction was one of only three reconstructions in the IPCC TAR spaghetti graph. However, as shown below in the left panel, the graph does not show any mismatch between the Briffa et al reconstruction and the other reconstructions or with temperature, even though late 20th century values of the Briffa recon would be at early 19th century levels – well below the supposed “confidence intervals”. The detail shows why: the Briffa MXD series has been truncated. The Briffa series is in green and ends in 1960, but the truncation is virtually impossible to spot in the spaghetti graph as the green series seems to merge with another dark-colored series. Without the truncation (as you can estimate by examining Figure 1), the late 20th century values of the Briffa series would go to values about equal to early 19th century values, yielding a glaring mismatch.
Figure 2. IPCC TAR Figure 2-21 with blowup.
This unscrupulous truncation was previously reported at CA.
IPCC AR4
This was bad enough in TAR – what about IPCC AR4. While their spaghetti graph included more series, their truncation of the Briffa et al 2001 series was the same as in IPCC TAR as shown below – the Briffa et al 2001 recon is in light blue – see the 1960 truncation in the detail at right.
Figure 3. IPCC AR4 Box 6.4 with blowup.
Review Comments
In my capacity as an IPCC AR4 reviewer, I noticed that the Briffa et al 2001 reconstruction had once again been truncated in 1960, which had the effect, in Pierrehumbert’s words, of “concealing” the “mismatch” from the reader. I observed in language not dissimilar to Pierrehumbert:
Show the Briffa et al reconstruction through to its end; don’t stop in 1960. Then comment and deal with the “divergence problem” if you need to. Don’t cover up the divergence by truncating this graphic. This was done in IPCC TAR; this was misleading. (Reviewer comment ID #: 309-18)]
The IPCC review comments (unavailable at the time of my original post on this but now online here – Go to chapter 6 – Final review comments and then to comment 6-1122) stated:
Rejected: though note ‘divergence’ issue will be discussed, still considered inappropriate to show recent section of Briffa et al. series
So let’s return to Pierrehumbert’s statement:
Whatever the source of the purported … data, there is no legitimate reason in a paper published in 2007 for truncating the … record … as they did. There is, however, a very good illegitimate reason, in that truncating the curve in this way helps to conceal the strength of the trend from the reader, and shortens the period in which the most glaring mismatch … occurs.
While Pierrehumbert made this statement in respect to Courtillot et al 2007, the same principles obviously apply to IPCC AR4. Indeed, the IPCC circumstances are far worse than Courtillot circumstances. First, Courtillot’s interest was in the earlier period, as he recognized the post-1990 divergence. Second, Courtillot used an obsolete data set resulting in a shortened series, but did not actively truncate the data. Neither of these justifies the use of obsolete data, which, like Pierrehumbert, I have criticized.
In the IPCC case, there was an active truncation of “inconvenient” data which had the effect of concealing a mismatch from the reader. Worse, the matter was clearly and explicitly brought to IPCC’s attention and they refused to address the concealing.
In Pierrehumbert’s words, there was no “legitimate reason” for what IPCC did, but a “very good illegitimate reason”. It’s gratifying that Pierrehumbert and realclimate are lending their authority to the condemnation of such practices.
