A CA reader emailed me, observing that there may be relevant differences in insulated and uninsulated buckets in the post-World War 2 period, which could easily affect adjustment schedules. This makes a lot of sense to me and might reconcile a few puzzles and opening others.
Let’s say that the delta between engine inlet temperatures and uninsulated buckets is ~0.3 deg C (and here we’re just momentarily adopting one of the canonical Folland numbers as this particular number surely deserves to be cross-examined). Insulated buckets would presumably be intermediate. Kent and Kaplan 2006 suggest a number of 0.12-0.18 deg C. So for a first rough approximation to check our bearings on this – let’s suppose that it’s halfway in between. Maybe it’s closer to engine inlets, maybe it’s closer to uninsulated buckets. We’re not trying to express viewpoints on such conundrums here – we’re merely examining what assumptions are latent in the temperature estimates.
We know that 90% of all measurements in 1970 with (supposedly) known provenance were done by buckets (Kent et al 2007), while there was a turnover in proportion to about 90% engine inlet and hull sensor by the 2000s. In my first cut at estimating the effect of unwinding some of the erroneous adjustment assumptions, I posited that the above information implied that the 0.3 deg C adjustment between buckets and engine inlets didn’t disappear merely because of reversion to “business as usual” after WW2. On this information, the only time that the delta could be introduced was between 1970 and 2000. This in turn poses new conundrums, as you’re getting into periods with satellite measurements. So there are issues with pushing the delta entirely into the post-1970 period.
However, let’s suppose that there was a transition from predominantly uninsulated buckets immediately post-WW2 to predominantly insulated buckets as at 1970 or so. Then the 0.3 deg C total adjustment would be spread proportionally between the two periods – with the delta between uninsulated buckets and insulated buckets being allocated to the 1945-1970 period or so (together with other relevant instrumental changes) while the delta between insulated buckets and engine inlets would be allocated to the 1970-2005 period (again together with any other relevant instrumental drifts e.g. changing proportion of hull sensors, buoys, whatever.)
I’m headed away for the weekend, but I’ll redo my rough guess based on these variations in a day or two.
A couple of observations, which readers should bear in mind.
In doing an “audit”, if an auditor identified a situation where the assumptions did not warrant the conclusions (e.g. the IPCC hypothesis of a sudden and permanent changeover to engine inlets at Pearl Harbour), he would not substitute his own assumptions; he would reject the assumptions and throw the problem back at the authors. I’ve been pretty consistent about this in the proxy areas, avoiding the temptation to posit what really happened. I don’t want to go beyond this policy here either.
Now I did a graphic showing the impact of abandoning the Pearl Harbour hypothesis under certain other assumptions. Please construe any such calculation as equivalent to a sensitivity calculation to show the effect of (say) excluding bristlecones, to illustrate the impact of certain assumptions, but not advocacy of a specific alternative. My language at the time may not have been explicit on this, but it’s the way that I do things. There are any number of intricacies in the interpretation of SST buckets. What are the effects of hull sensors? Drifting buoys? Etc. etc. I have no information on such matters at present. Could hull sensors and drifting buoys offset changeover from insulated buckets to engine inlets? Could be. I’m not opining on this. I’d like to see proper expositions of these populations and biases and one of the good outcomes of Thompson et al is that it will almost certainly achieve this goal.
As to my original observations on this: I was in a position to observe that the Pearl Harbour assumption did not hold up and that the 0.3 deg C adjustment did not occur in one bite in 1941, but needed to be spread much later. Right now, as noted above, it looks like some proportion gets spread to the period before 1970 and some proportion after 1970.
I might also make a point here about reviewing. Posts at this blog are not “peer reviewed”, but they are read by a lot of people, who have an opportunity to comment on them. In this particular case, a reader drew my attention to a distinction between insulated and uninsulated buckets that I hadn’t thought about and which makes sense in the context. One could make a case that blogs are actually a very good way of carrying out review of articles. In economics, journals like articles to have been around for a while and to have been reviewed in a variety of venues before being published; the whole idea of embargo-to-publication that has been promulgated by Nature and Science seems like something that needs to be talked about.
The audience is also astonishingly active. Demetris Koutsoyannis wrote to thank me a favorable recent comment; he said that he got 2700 downloads of the article in one day after being recently mentioned and linked at Climate Audit and 5700 since then, with the vast majority coming directly from CA. It’s no accident that we rank so high on any number of Google searches.
Provenance Bias versus Sudden Changeover
A second point that I’d like to quickly mention before leaving for the week-end.
My original criticisms were based on the arm-wavingness of assuming that there was a “sudden but undocumented” change in observing procedures coinciding with Pearl Harbour. This seemed far too much like Briffa’s Cargo Cult explanation – and originated once again from essentially the same crowd. Here is the precise quote from Folland and Parker 1995 (originally cited here in 2005):
Barnett (1984) gave strong evidence that historical marine data are heterogeneous. He found a sudden jump around 1941 in the difference between SST and all-hours air temperatures reported largely by the same ships. Folland et al. (1984) explained this as being mainly a result of a sudden but undocumented change in the methods used to collect sea water to make measurements of SST. The methods were thought to have changed from the predominant use of canvas and other uninsulated buckets to the use of engine intakes. Anecdotal evidence from sea captains in the marine section of the Meteorological Office supported this idea. .
The abrupt change in SST in December 1941 coincides with the entry of the USA into World War II and is likely to have resulted from a realization of the dangers of hauling sea buckets onto deck in wartime conditions when a light would have been needed for both hauling and reading the thermometer at night. The change was made possible by the widespread availability of engine inlet thermometers in 1941 (section 4)
Now there is a real discontinuity in World War 2 and it’s related to the above. Thompson et al 2008 provide a far more plausible and convincing explanation of this discontinuity than the one provided by Folland and Parker 1995. It’s partly a difference in nuance, but the difference is important. Here is a graphic from Thompson et al 2008 showing a change in US contribution to the total population of SST measurements. Clearly the discontinuities in the increased proportion of US measurements (engine inlet) to the total population in WW2 is directly connected to discontinuities in the SST history. It’s not that thousands of ships suddenly and overnight changed how they did things; it’s that the data sources changed. In a way, it’s exactly like the Hansen Y2K problem.
This explanation makes sense, while the Folland and Parker 1995 one doesn’t. If you then look at their graphic with this idea in mind, one wonders what is happening in the 1960s, when US contribution inexplicably goes to nearly zero for a while, which seemingly coincides with a downdip in world SST, with a later sharp increase in US contribution coinciding with an increase.
Left undiscussed by Thompson et al 2008 is the thorny issue of wooden buckets prior to 1940. One of the distinctive innovations of Folland in the 1980s and Jones et al 1991 was the hypothesis that heavy wooden buckets (which behaved like insulated buckets) were used in the 19th century, with a more or less linear changeover to uninsulated canvas buckets.
In the last few days, I’ve been reading some of the older literature on buckets. Brooks (Weather Review 1924) is an early and interesting consideration, reporting on experimental comparisons of canvas bucket and engine inlet measurements in a variety of situations. There’s not a whiff of mention of wooden buckets. One of the discussants of the paper wonders why they use canvas buckets at all, when engine inlet appears superior; and another says that it’s just a holdover from sailing vessels. Another discussant mentions the use of engine inlet measurements by the Canadian Meteorological Service in the NW Pacific, a point that may tie in to an issue in Rayner et al 2006 [note to self].
Folland and Parker 1995 carried out elaborate measurements on heat dissipation from an oak bucket, noting, in passing, that there was no evidence that this type of oak bucket had ever been used to measure SST.
A nineteenth-century oak ships’ bucket covered in iron bands has been studied though there is no indication that it was used for taking sea temperatures
This did not stop the development of an elaborate adjustment methodology based on a transition from oak buckets to canvas buckets, a transition which had the effect of lowering adjusted 19th century temperatures relative to what would have been calculated if canvas buckets had been used.
Surely at some point, someone has to show that these wooden buckets actually were used in the 19th century. If I had to take a stab at it right now, I think that it’s far more likely that canvas buckets were used until engine inlet changeover and that it would be well worthwhile doing a variation on this assumption. Also the assumption of 0.3 deg C isn’t necessarily written in stone. Brooks 1924 observed many values well in excess of this. Saur 1963 (J Appl Met online) is an interesting direct comparison of bucket and engine inlet measurements on the same voyage – the sort of evidence one would like to see more of – and the differences were higher than 0.3 deg C.
Whatever the end result, there are a lot of models being developed on this sort of information and careful cross-examination of the wooden buckets is long overdue.