There is a wonderful monograph by the late John Bell, The Speakable and Unspeakable in Quantum Mechanics which deals with the issue of what can and cannot be described on many levels. Roger Pielke Sr. is pursuing such a chimera, demanding that we use metrics of climate change which are simply either not available, or available only for short periods of time, or, as we have seen for ocean heat content, or are not (one hopes yet) reliable.
Now, Roger Pielke Senior is a fan of the IPCC using such metrics such as “the global-averaged and regional-averaged patterns of changes in heat content in units of Joules” instead of surface temperatures. Eli wished him great good luck in explaining that to your average policy maker.
Well, how about enthalpy or energy content. Eli agrees, the perfect measure would be the energy change of the oceans, surface and atmosphere. One could settle for lower atmosphere and upper ocean. Multidecadal records of these too are not currently available and, as yesterday's post shows, many of the measurements that we have are not as good as we would wish. Worse from the standpoint of Pielke, the best way of reconstructing such profiles is through models which can be validated by comparison with the data we do have, such as records of surface temperatures and precipitation.
To paraphrase Thomas Knutson: Should we use surface temperature observations or energy content profiles? In reply we note that if we had multidecadal observations of energy content throughout the oceans and atmosphere, we obviously would trust them more than the surface temperature records, but unfortunately we cannot go back in time with modern instrumentation and recapture the needed profiles.
Eli wishes to emphasize that the only way to get those profiles is through inproved postdiction with models. Those who demand unknowable perfection before doing anything have little interest in action. They don't much like proxy measurements either.
Yesterday, when describing a few of the problems with ocean heat content measurements, Eli forgot to point to this post by Tamino which showed that even without corrections, the differences between the models and the observations were within natural variability. The figure is from Hansen et al. (2005, Science, vol. 308, pg. 1431).
It will be interesting to replot the models against corrected observations.