Gavin Schmidt and James Hansen say that errors in the U.S. “don’t matter” because it is only 2% of the earth’s surface (about 6% of the land surface). This implies that the accuracy of measurements in other parts of the world can be relied on. In the U.S. the 1930s have a similar level to recent levels, while the ROW has a striking difference. The surface area of Africa (30,300,000 km” ) is about one-fifth of the Earth’s land surface. 148,939,100 km”), about 4 times the size of the contiguous 48. I guess that this would be a good place to look for the high-quality stations that Schmidt and Hansen are counting on.
From the collated GISS data set, I selected those stations meeting the following criteria:
1) country code less than 200 (yielding Africa plus, as it turns out, a couple of islands;
2) designated R for “rural” by GHCN/GISS (taking no position on whether this is accurate)
3) beginning by 1931 (to cover the 1930s) and ending after 1990.
This yielded a grand total of 10 stations (of which 2 are oceanic islands), which are shown below, with the ones in the first graphic being more northerly and the ones on the left, more southerly.
A few observations about the Africa Ten. My search required that there be values post-1990 and yet some of the graphics appear to end much earlier. For Sao Tome, for example, there are a couple of oddball singleton readings in the 1980s and one in the 1990s, which triggered its inclusion here, although the record is obviously not a continuous record into the 1990s. An inspection of the Ain Sefra suggests at least the possibility of inhomogeneity in the decades-long interval between measurements.
So if you’re looking here for information from rural sites in Africa that show the relative levels of 1934 and 1998, there’s not a lot to go on. There are relatively long records from St Helena and the Seychelles; both of these data sets are interesting and probably quite valuable, but they don’t per se give a lot of information on Africa. The only candidate station, Mongu, shows elevated 1930s, with 1934 appearing to be warmer in Mongu than 1998, as in the United States.
To move past this data set, Hansen has to blend in urban sites, which, even Hansen says should be avoided in climate studies. It doesn’t seem like one can draw very forceful conclusions about relative land temperature levels in Africa between the 1930s and the present from these rural sites.