I’ve noticed that even Hansen’s rounding seems to give a little edge to modern values relative to older values. It’s not a big effect, but Hansen doesn’t seem to like to leave any scraps on the table. Cara, Siberia has 3 versions – all of which appear to be the same station; all 3 versions cover more or less the same period (Data here ) and, in this particular case, the values only go to 1990.
As a first step, it seemed to make sense to take a simple average of the data (first 3 columns of data at link) and compare it to the Hansen combined version (4th column). The Hansen combined more or less matched the rounded average up to 0.1 deg C, but the differences were not random: Hansen was always lower as shown in the first panel. In the second panel, I’ve pretty much matched Hansen by doing the following:
1) calculate the average rounding to 2 digits; 2) multiply by 10, take the floor and then divide by 10.
Typically, this sort of situation only affects pre-1990 data. GHCN made a major collection effort around 1990. In this particular case, the data hasn’t been updated after 1990. IT’s not that this station necessarily ceased measuring temperatures (as often thought); it’s may be that GHCN (and thus NASA and CRU) merely haven’t updated it since 1990. GHCN says that it updates rural stations only “irregularly” and, as it happens, the “irregular” schedule apparently hasn’t included any updates since 1990 for large areas of the world – so NASA are CRU are “forced” to rely for more recent readings primarily on the WMO airport system, which presumably is more urban-biased than the historical network. For these later readings, multiple scribal versions do not arise and “Hansen rounding” doesn’t come into play. A small effect to be sure, but Waldo is voracious and seemingly no scrap is too small.