First the link to this week’s complete list as HTML and as PDF.
***
Urgolites et al. repeat a common and potentially devastating mistake, when they state “performed similar to the controls P > 0.1”
. That a difference is too small to show up as significant in only 5 subjects, means it can’t be demonstrated to be present. This is by no means the same as claiming it can be shown to be not present. If they meant “not meaningful”
, i.e. if they compare real cohorts and not idealised standards, then they should not have used the standard error but the standard deviation in their diagrams.
***
Cuffey et al. claim to measure ice temperatures to a precision and resolution of Millikelvin. This is easy to do, if all you want to measure is the temperature of the bulk of your thermometer, but nearly impossible even under ideal laboratory conditions for the undisturbed temperature of a surface or medium. What they actually do is drill a miles long 10 cm hole while frequently withdrawing and reinserting their drill and its thermal capacity. They then fill the borehole with an unspecified fluid, wait a while, and measure that fluid’s temperature while lowering their thermometer. Not only that, they claim the temperature they measure at a depth of thousands of meters to be a residue of the temperature that snow had, when it fell tens of thousands of years ago – somewhat attenuated by conductivity and thermal diffusion perhaps, but still there nonetheless. And calculating back they arrive at that original surface temperature with a climatologically meaningful precision. Who knows, it might even be true, but what I see is a rather simple and smooth curve with just three maxima and two turning points and a model with many tuneable parameters.