The lag between SST and TLT was discussed here, as it would seem to have important consequences when trying to estimate the overall climate feedback (and hence sensitivity) from surface temperatures. Here I present the estimates of lag time using the SST and 600mb AMSU data available from the UAH Discover website. The script is available here.
Anyhow, here is a plot of the daily anomalies:
Clearly, there is a lag between the peaks of SST and those of the atmospheric temperatures at the 600 mb layer, which is also confirmed with the chart below (as you might recall, 1-3 months was our estimate from TLT vs. SST, but at this layer it could be even close to 4 months):
For those interested in what these graphs would look like if we did not center based on daily values, and left the annual cycle in there instead, here’s what we get:
Troy's Scratchpad 
I show here http://vixra.org/abs/1108.0032 that SST lags TSI by precisely one quarter of a solar cycle as expected the energy balance model CdT/dt=S where SST is integrating the solar anomaly. There is the same shift between Spencer’s TOA data and surface temperature, shown here http://landshape.org/enm/phase-shift-in-spencers-data/.
Looking at your red and blue time series figure above I think the lag should also be exactly one quarter of the annual cycle if you take peak-2-peak shift. The cross-correlation function may be shifted by the secondary peak.
In this case, the atmosphere also follows the dynamics of the basic energy balance model, in integrating the annual periodic from the SST, leading to the exact shift of one quarter of the cycle length. The one quarter cycle is easily seen by integrating a sine wave to give -cosine, which has a one quarter cycle lag.
Nir Shaviv mentioned this shift in his calorimetry paper, and notes also that it will be slightly changed by diffusion.
Comment by laterite — October 3, 2011 @ 10:38 pm
Interesting, have a look at CO2 and SST here http://www.biomind.de/realCO2/realCO2-1.htm
cementafriend
Comment by cementafriend — October 4, 2011 @ 4:12 am
An interesting question would be why it is close to four months at this particular level: is the lag much less closer to the surface?
One of our reviewers believed we were looking at a layer too high to be more important to clouds than surface temperatures (to which he expected low level clouds to be magically quantum entangled I guess) Perhaps in the future we can use daily AMSU data (I believe the Science of Doom fellow suggested at CA that a mainstream paper said using daily data would reduce bias due to random cloud variations) and different layers, to determine which layers dominate the responses. That way, if it were really true that the LT data were too high in altitude overall to be connected to the clouds that “must” dominate the feedback, we could at least determine what layer it should be.
I remain convinced that the clouds do not magically know what the surface temperature is, but respond to their immediate environment.
Comment by timetochooseagain — October 5, 2011 @ 7:48 pm