A discussion in comments on Warwick Hughes’ blog is on a tangent regarding the measurement of temperature extremes; the maximum and minimum temperatures reported for a day. (3rd comment, ff)
Back in the old days, brief temperature transients would have been filtered out by the long response time of e.g. the mercury thermometer; especially when compared to modern methods of measuring air temperature. The high thermal inertia presented by the glass and the mercury, combined with the low thermal capacity of (especially dry) air, likely results in short-term transients in air temperature being effectively invisible to a mercury thermometer.
Sufficient heat must be transferred between air and thermometer to expand/contract the mercury column accordingly. The air temperature has to remain steady for a sufficiently-long time to heat/cool the thermometer to the air’s actual temperature and the air has to maintain some movement because, in the process of transferring heat, the thermometer either cools or warms the air in direct contact.
On the other hand, the low thermal mass of the crucial parts of modern sensors simply because the sensing parts are very small, means that the modern sensors are intrinsically able to respond much more quickly.
The “new records” may represent no more than our ability to detect short-term, transient temperatures. If collected at all, the short-term temperature data are probably “lost” in the bureaucratic distillation processes. It’s not that the volumes of data are going to be huge if one goes from by-minute to by-second temperature logging. But it may reveal a “disturbing” fluctuation in measures of transient air temperatures; indeed larger than the alleged rise in average temperature over the past century.
Unless care has been taken to ensure that modern equipment has the time constants of the old, the “records” are simply not comparable.
Here is the WMO (sizeable PDF) in 1998, opening the can of worms dealing with the change in screens; the enclosures used to try to minimise undesirable effects of temperature measurements.
Suffice to say that the temperature readings obtained by changing anything in the way that temperatures are measured; present a discontinuity in the temperature readings. Therefore data before and after the change are not “alike”; they aren’t measuring the same thing. Which invalidates any subsequent processing which assumes that they are the same thing.
More to follow… no doubt.