Much, just about all of David Roses claims in his hit piece on Tom Karl has been walked back, not the least by Rose's source, John Bates. Among the statements which Bates has disowned and which now must be considered fabrications was the paragraph
Dr Bates said: ‘They had good data from buoys. And they threw it out and “corrected” it by using the bad data from ships. You never change good data to agree with bad, but that’s what they did – so as to make it look as if the sea was warmer.’
But Dr Bates said this increase in temperatures was achieved by dubious means. Its key error was an upwards ‘adjustment’ of readings from fixed and floating buoys, which are generally reliable, to bring them into line with readings from a much more doubtful source – water taken in by ships. This, Dr Bates explained, has long been known to be questionable: ships are themselves sources of heat, readings will vary from ship to ship, and the depth of water intake will vary according to how heavily a ship is laden – so affecting temperature readings.Dr Bates said: ‘They had good data from buoys. And they threw it out and “corrected” it by using the bad data from ships. You never change good data to agree with bad, but that’s what they did – so as to make it look as if the sea was warmer.’ERSSTv4 ‘adjusted’ buoy readings up by 0.12C. It also ignored data from satellites that measure the temperature of the lower atmosphere, which are also considered reliable. Dr Bates said he gave the paper’s co-authors ‘a hard time’ about this, ‘and they never really justified what they were doing.’
Changing contribution of different measurement techniques (top panel) and their timeseries relative to the average of all sources at any given time (bottom panel). Note large and systematic offsets between distinct sources that vary through time. Source: IPCC
Ship measurements were the first, actually they go back well before 1920 and buoy measurements only kick in late in the twentieth century. Earlier (and this would be pretty much anything published before 2000) would only, or massively depend on the ship measurements. Thus offsetting the buoy to the ship data would cause the least confusion for anybunny looking at older publications.
Makes sense to do it that way.