In yesterday’s post, I observed that Nature’s recent news article on Met Office decadal forecasts failed to show the most recent Met Office decadal forecast and that its inclusion would not have permitted the Nature headline. I also showed the large change from the Met Office submission to IPCC AR5 and their current decadal forecast. Asked to comment by Anthony Watts, Richard Betts of the Met Office did not explain why the Met Office either signed off on or had not objected to the omission of their most recent forecast. Instead, Betts claimed that my plot was “wrong” because “HadGEM2 not an initialised forecast, so Steve is wrong to plot it from 2010 high point – exaggerates difference”… as though this were responsive:
However, I had directly plotted from data from the Met Office so there was no inaccuracy in my graphic despite Betts’ implication. Nor, needless to say, there is no scientific or statistical principle forbidding the illustration of initialized and uninitialized forecasts on the same graphic. Ironically, as shown below, the UK Met Office had themselves done so in the very article (Smith et al 2012 Clim Dyn) from which the Nature News article had been derived.
Here is the graphic that Betts criticized. The CMIP5 contribution,as Betts had observed, is “uninitialized”, while the two Met Office decadal forecasts (green and blue) are “initialized”. The Met Office IPCC contribution also included a hindcast, but I had shown the CMIP5 forecast from 2010 on to highlight the difference (taking care to note that I had shown only the forecast portion.) All data, as noted above, is Met Office data. I plotted the CMIP5 contribution from 2010 on, estimating , as stated in the post, that 2010 was the approximate start of the “forecast” given the timing of the CMIP5 contribution. In response to Betts’ objection, I added the hindcast portion into a revised graphic, observing that this was irrelevant to the conclusions of the post.
Figure 1. See yesterday’s post for explanation.
Now here’s something interesting.
Smith et al 2012, the academic article from which the Nature News article was primarily drawn, the lead author of which is a Met Office employee, also included a graphic contrasting the forecast portions of initialized and uninitialized runs, both from approximately 2012 on, as shown below. (No hindcast for the uninitialized run was shown.) Like my graphic, they showed forecasts in red, blue and green to the leeward of observations. However, they showed the “uninitialized” forecast in blue, while I had shown the uninitialized forecast in red. And while they had shown the initialized forecast in red, I had colored the two initialized runs as green and blue. (The green and lightblue lines in Smith et al 2012 are not relevant to the present discussion.) They showed a smoothed version of HadCRUT3, while I showed an unsmoothed version of HadCRUT4. I find it very hard to understand Betts’ precise objection to my graphic.
Figure 2. Smith et al 2012 (CLim Dyn) Figure 8a, showing Uninitialized (blue) and Initialized (red).
For comparison with my original graphic, here are the Met Office CMIP5 submission (magenta) and most recent decadal forecast (dodgerblue) plotted onto the Smith et al 2012, with irrelevant portions of the Smith 2012 graphic whited out. Betts had criticized my graphic which showed the uninitialized forecast (HadGEM2) from 2010, but the diagram in Smith et al 2012 shows the uninitialized forecast starting from a very similar point. Observations are slightly displaced downwards in this diagram, because Smith et al used HadCRUT3, while I used HadCRUT4, which has been amended and runs a bit warmer over the past 15 years.
Figure 3. Annotated and cleaned version of Smith et al 2012 Figure 8a. CMIP5 HadGEM2 (magenta). 2012 HadGEM3 decadal forecast (dodgerblue). Observations, following Smith et al 2012, are HadCRUT3, extended to current values.
Betts claimed that my graphic was “wrong”, but did not object to a similar graphic technique in the underlying academic article by Met Office scientist Smith. Not that there was anything “wrong” in either graphic. Betts’ claim was both ad hoc and unfounded, as too many claims arising from the climate science community.
Betts’ claimed that my graphic “exaggerated” the difference between the Met Office submission to IPCC and the most recent Met Office decadal forecast. Untrue. I plotted Met Office data. In my opinion, the exact contrary is the case: the Nature News graphic downplayed the differences between Met Office versions by not showing the most recent decadal forecast and by not showing the Met Office IPCC submission. If anything was “wrong” in this incident, it was that the Met Office either signed off on the Nature News’ omission of their most recent decadal forecast (if given a chance to comment) and that they failed to write in their own objection to the omission once the article was published. And that instead of conceding the point, Betts made untrue and unwarranted allegations that I had committed errors in my post reporting the problem.
