In a paper published by Richard A. Muller et al, a new estimate for temperatures of the last 250 years is given. Helpfully, error bars for single and double standard deviations are also provided. The graph looks like this:
The temperatures of the single year chart in the 18th century are all over the place, reaching even current levels, despite the little ice age still in progress. But even in the much smoother 10 year average, the temperature cannot be determined with an accuracy of one degree.
The somewhat darker shaded area is the single standard deviation interval, which has a roughly one-in-three chance of being wrong. The double standard deviation interval is almost twice as large – and one in twenty data points are expected to be outside of it, given the quality of the available data (or lack thereof).
To put it this way: there is a one in three chance that the land surface temperature in the decade of the declaration of independence of the USA in 1776 was either above 9.2 degree or below 8.2 degree Celsius. The data is so poor, that it is essentially impossible to tell those two possibilities apart. The data indicates that temperatures rose dramatically in the decades before and fell just as dramatically in the decades afterwards. The fall may be attributed to the Laki eruption of 1783, but nothing remotely comparable happened before that in the 18th century.
Even median values rose by 0.7 K in 20 years, perfectly comparable to the rise between 1980 and 2000 – although a much steeper or much shallower rise is perfectly possible, given the data. The incredibly imprecise data makes this impossible to tell.
To make the picture somewhat clearer, here is a version of the same graphs with adjusted black levels. There is a 68% chance that the actual temperature is within the black area. There is a 4% chance that the temperature was even outside of the grey area.
This graph goes back less than three centuries, but the claim that the currently observed rise in temperatures is unprecedented goes back over 10,000 years. Given that 250 years ago, relatively precise measurements with thermometers were possible and consistent data series were collected at least for a few places, this does not bode well for any derivation of temperatures going back even further in history.
In short, the claim that the current rise of temperatures is unprecedented cannot possibly be supported by the existing data, with any reasonably scientific framework.
(January 2015) I no longer stand by this statement, it was way over the top, especially saying the data could not support “any reasonably scientific framework”. It is, however, a fairly accurate representation of the way I thought at the time.