Some very strange statistical jiggery pokery from Jim Hansen’s outfit
Frank Lansner has been a busy man, and he’s asking some very thought provoking questions.
The Northern Hemisphere has a ratio of 40% land to 60% oceans, and the Hadley Met Centre seems to use a similar ratio (NH HadCrut Series: 58% ocean, 42% land). But Frank Lansner wondered why, when he graphed the GISS land-data-set alongside the combined-sea-surface-temperatures (CSST), GISS comes up with an “averaged” line that runs much closer to the land data set and not the sea surface set. If it were weighted 60:40 (ocean:land) the combined Northern Hemisphere line ought to run slightly closer to the ocean based temperatures.
So Lansner mixed the land and sea temperatures in different ratios and graphed them and an odd thing occurred. Perhaps there is some good reason for it, but the GISS NH average line is currently running close to a mix that could be almost 70% land, and only 30% ocean. Back in 1985 the NH Average was closer to the sea temps as would be expected. In fact as late as 1995, the NH line still ran at around 40% land area. But somewhere post 1995 – 1999 for some reason (see the update at the bottom for some good suggestions) the average tracks closer to the 70% line. According to Frank, this effect does not occur with the HadCrut average.
Frank is looking for feedback and suggestions, and wondering if there could be any other explanation. So am I. The effect is clear also in this graph. The land-based datasets are the brown ones near the top. The blue ones in the middle are GISS and then below that Hadley, then finally the black line is the satellite measured average for land and sea, and the combined sea surface temperatures. It’s interesting how closely the satellite set compares with the sea surface data.
Lansner points out that the key difference between the brown lines at the top and the blue-black lines underneath is that Urban Heat Island only affects the lines at the top (i.e. there is no urban heat island over the oceans, and not too much “urbanity” near the satellites either). Lansner further split up the satellite measurements into satellite-land versus satellite-ocean and what’s especially interesting is how the satellite-land values agree better with the ocean measures after 2001 than they do with the surface-land thermometers. Are we eyeballing the effects of UHI and siting problems in the thermometers on the land, and are GISS somehow inadvertantly amplifying these artificial effects with weighting, homogenisation, gridding or averaging proceedures that rely more on these land measurements than on the independently consistent satelites and sea surface measures?
Much more HERE (See the original for links, graphics etc.)