I plotted the curve from 1964 – 2012 here, with a linear fit for the data from 1964 to 1998, the beginning of the so called “hiatus”. (Data are frome here, with the NASA GISS temperature set as basis.)
As you can see, there is not much of a hiatus. But if you look sharply, you could see after 1999, the red data points, a slow-down of atmospheric (not global!) warming, resulting in a lag of about 0.1 K.
The big hiatus rumoring through the press can be seen only, when we fit our straight line to a shorter period between 1985 and 1998, which is statistically questionable.
Doing that means assuming, that atmospheric (not global!) warming did accelerate in the middle period. Now, we get a temperature lag of about 0.2 K against the expectation.
A lot of interesting work has been done to find an explanation for this fact, the most notable I know of by Foster and Rahmstorf, who calculated the deviations from the middle line caused by volcanoe eruptions and El-Nino/La-Nina events, the big southern pacific air and sea current oscillation, with linear contributions. They found out, that the “hiatus” can be explained quite well by these. Even more convincing was the simulation of Kosaka and Xie, who allowed for nonlinear feedback of El-Nino/La-Nina events to atmospheric temperature and got a theory – measurement accordance almost too good to be true. The British Met Office argued in a meticulous discussion, that a lot of the heat unaccounted for must be gone into the deep sea, which is no contradiction to the former.
But there is one striking point, where a lot of energy has been dumped during the last 18 or so years: the arctic ocean. Let’s look at another plot: the volume of the ice floating on it (from here):
Between 1995 and today more or less 8000 km³ of perennial ice melted away. Could not the heat necessary to do this be reflected in the warming lag of the atmosphere? And it could.
Here is the copy of an excel sheet, the electronic equivalent of a back-of-envelope-calculation, in which I tried to find out, whether or not the ice loss could be correlated with the warming lag.
|melting enthalpy of water||333||1,00E+03||J/kg|
|lost perennial ice volume since ca. 1995||8000||1,00E+09||m³|
|density of ice||0,9||1,00E+03||kg/m³|
|lost perennial ice mass since ca. 1995||7200||1,00E+12||kg|
|heat used to melt lost ice mass||2397600||1,00E+15||J|
|mass of earth atmosphere||5,15||1,00E+18||kg|
|specific heat of air under standard conditions||1,05||1,00E+03||J/kgK|
|total heat capacity of the atmosphere||5,4075||1E+21||J/K|
|temperature decrease of the earth atmosphere (even distribution of heat contribution assumed)||-0,44338419||1,00E+00||K|
|Difference between expected and real atmospheric temperature, ca.||0,1||1||K|
|antarctic volume gain rate||35,5||1,00E+09||m³/year|
|years since 1995||18|
|approximate gain since 1995||639||1,00E+09||m³|
The result is, that when the heat energy for the melting of the perennial arctic sea ice had been taken evenly from the hole of the atmosphere and from nothin else, and its specific heat capacity can be considered everywhere the same, a temperature lag of more than 0,4 K would have happened.
As is not so well known, the antarctic sea ice increased over the last decades, thus puzzeling once more the scientific community. Massonet and colleagues calculed this mass increase here. In the lower part of the spreadsheet, I looked at the size of this effect. It is not bigger than 10 % of the arctic loss.
This is all terribly incomplete of course. The atmosphere does not contribute evenly to the melting. A lot of the heat during summer melting comes not via the atmosphere, but directly from the solar radiation because of increasing surface absorption. It is quite possible, that a small shift in the regional radiation balance is responsible for the lions share of the melting, as the sea ice simulation team of the University of Washington warns here. The atmosphere is pretty strongly coupled to the top layers of the continents and the seas, so those should have contributed to the melting, too.
On the other hand, the observed temperature lag is only 0.25 to 0.5 of what would be expected in an atmosphere-only model. So this seems to be somehow already accounted for.
It might not be the whole truth, but it’s worth looking at it.