I have a comment on these temperature record pie charts:

[caraher]

17. I think that's right

It's easier to see that the number of new records should decrease with time if all further data does is add to the number of measurements made on an essentially stable expected temperatures - which could lead to some spikes in the ratio if you wind up taking the quotients of small numbers.

But for the most part, if you're seeing random fluctuations only and not some warming or cooling trend yeah, you should be getting closer to 1:1 on average subject to the kind of amplification of noise that can occur when the absolute number of new record temperatures per decade gets small. Even then, if you're talking about something like a few hundred records in a decade that should yield fluctuations on the order of 10%.

It's also noteworthy that even in a "stable climate" scenario the decadal records should still have some nontrivial excursions from the mean climate. For instance, in the '90s we had the Mt. Pinatubo eruption, which should have pushed the ratio down somewhat. It's interesting that even so, there's still a notable excess of record highs recorded.