How hot is the sea?

I am currently reading Richard Heinberg’s new book “Blackout – Coal, Climate and the Last Energy Crisis,” and will, hopefully before long, give you my review. Now it does contain a fair bit on climate change, but that is not why I would have bought it, if TOD had not arranged for me to get a review copy. The reason that I am reading it, and that I will review it relates to the discussion that it continues on the availability of coal in the different nations around the world.

But even if I were to review the climate change bit (and it will be hard not to make some comments, given that it is interwoven in the discussion throughout the text) I will promise that I will not subject the book to the sort of scrutiny and, in part nit-pickiness that Gorge Monbiot is currently subjecting Ian Plimer to over his book, Heaven and Earth: Global Warming – The Missing Science. (I did get a copy of that book, and suspect that I will now read it with the several piles of criticism from a number of the sources quoted by the Guardian printed out beside me on the table. – though I note that in the week since I ordered and got my copy it has gone out of print).

However it did get me curious about the genesis and history of a couple of plots of global temperature issued by the Hadley Climate Research Unit. Some time ago I had used the plots that used to sit on the top left of the Hadley site as a source of information and comment, and the one I used was this one.


Temperature changes since 1850 (Hadley)

However just recently when I went back to look, the easy graphs to find are now these:

Temperature changes since 1850 (Met Office)

Which should not be confused with the summary plot of annual temperatures which Hadley posts:

Temperature changes since 1850 (Hadley )

The difference that has me interested in the bit between 1940 and 1970. It used to be, (top graphs) that that was, overall rather flat – agreeing also with the bottom. But if you look at the new way of combining the data, the global temperature is now shown to have been steadily increasing since 1950.

This is due to the way in which the data is now combined, since it is being done through an integration of land and sea temperatures, rather than the integration of Northern and Southern Hemispheres, or so it would appear.

Now getting the average sea surface temperature (SST) is not an easy task. To illustrate this consider this thermal picture of the Gulf Stream, and you can see the great range of temperatures as that heads out from the American Coast over to Western Europe.

Satellite thermal Image of the Gulf Stream

Thus even when you get the satellite image it is subject to interpretation. It also only reads the temperature at the skin of the surface (i.e. top 0.1 mm) rather than at any defined depth. In contrast there are also buoys out there which can give a real set of measurements at defined depths (but which still have to be weighted and integrated to get an average). And I took the following plot from a talk by David Llewellyn-Jones which shows the correlations that can be achieved.

Comparison between buoy measured (AVHRR) and satellite (ATSR) readings of SST.

One can compare this data with that from NOAA

Global temperature changes from 1880 NOAA.

This shows a slightly different path between 1940 and 1970, but again not the steady increase since 1950. The problem with the data that is used to derive the averages historically is that it has been subject to the odd correction. There were corrections made because of the use of different types of buckets to collect the water, and whether they were insulated, rather than taking the temperatures from the water in the engine inlet ports of the measuring vessel.

Different (metal canvas and wood) buckets used to measure sea surface temperature

The “first guess” estimates also tried to take into account the speed of the ships at the time of sampling (since it affected the evaporation rates and thus the amount of cooling of the sample). It was as a result of the estimates of changes in the errors caused by these different methods of reading that the actual values measured were changed.

Annual anomalies of corrected SST (solid), uncorrected SST (dotted) and corrected NMAT (dashed) for (a) northern hemisphere, (b) southern hemisphere, and (c) globe. Values have been smoothed with a 21-term low-pass binomial filter.

With the original readings being the dotted lines, you can see how the pre-1940 values have been adjusted. The original work showed that the variations between buckets and non-buckets were location sensitive., and it has been shown that it is also seasonal.

With these uncertainties it appears that there are still questions as to which data is the most accurate. The problem is also that the different techniques measure temperatures at different points in the sea, and when one changes from one method to another then corrections must be estimated and applied to bring everything to a common level. As a result the confidence that the historic data, as reported by the different Climate groups, is accurate seems less secure than it should be, given the importance that is being placed upon it.

Sadly it is more likely that the curves will be chosen which best reflect the bias of the selector and used without recognition of the potential errors and assumptions that have been made in generating it.