Clouds, Humidity, SST and Lower Troposphere Temperature in the Tropics - More Challenges

Working under the hypothesis that clouds are a regulating feedback in the tropics seems like a slam dunk.  Warmer oceans would increase the probability of cloud formation.  In the tropics, cloud shortwave reflection amounts to about 20 Wm-2 more of CRF as compared to cloud OLR retention.  Just compare cloud cover to SST done, right?

The only marine cloud data I can access for the longer instrumental period is prepared by IOCADS and is pretty sparse.  In the Blue is the standard 30% observation default for Climate Explorer with the reach i.e. pushing interpolation limits with, 5% of observation interpolation, in yellow.  Tropical SST has warmed, there is more moisture available for cloud formation, clouds kinda correlate, but there are gaps at places that would be interesting.  Most noticeably, the divergence between 1850 and 1880.  I have run into this same divergence with  BEST land data, which on the whole has a fair correlation with SST if you allow for a land amplification factor.  There is also limited Specific and Relative Humidity (SH,RH) data available on Climate Explorer, but that starts in 1973 and ends in 2003.  The end point is likely due to lack of enthusiasm in that data set.  Not a great sign.

In any case, a warming tropical SST would increase SH and possibly RH up to a point close to saturation.  Well, 1998 might be a point close to saturation but there isn't very much data.  With the large divergence at the start of the records plus the limited data, any correlation would be questionable.  The correlated peaks around 1945 are contrasted by anti-correlated peaks around 1910 and 1998 has no peaks.  There isn't a clear explanation.  If I shorten the data set, well that would be cherry picking.

The short period for SH provides a little reference, but likely not enough to be really useful.

Tropical RH and SH have different trends, likely due to the short time period starting year.  If anything in climate should be stable it would be tropical marine air RH and SH.  So even with more data there would not be much of a reliable variation to play with.  Quite likely a reason for the limited data.

Tropical clouds versus the tropical lower troposphere though is a bit more interesting.

There are in and out of phase situations like what I would expect during regulation.  Thanks to the absolute temperature provided by RSS there is something to scale cloud anomaly to.

So even though the cloud data isn't the best there may be enough to work with.  The spike around 1946 starts around 1941 so it may be related to war data collection issues or it may just be due to the real 20th century super El Nino. The peak being equal to the recent roll down would also be encouraging as far as a regulating function would go.

In any case, a little more than half of the increase in cloud cover happened prior to 1950 and the "Year CO2 Took Control".  That would be an indication of a continuous increase in water vapor, that other green house gas.  30S-30N is not the whole globe, but right at half of the area and the majority of the energy, especially energy that produces the most water vapor.  

Climate science has an odd way of defining forcing and feedback.  If water vapor increases due to CO2 equivalent gas "forcing" then it would be a feedback to that forcing.  That assumes everything is "normal" prior to the additional GHGs. This leads to models being a "boundary" value problem because the initial conditions are assumed to be "normal".  That is a major assumption.  Anyone that disagrees or points out behavior inconsistent with something being a "feedback" as narrowly defined catches flack.  However, if there are initial value issues, the assumed "normal" isn't, then what would be a feedback response could be a forcing in it's own right.  So to humor me, consider water vapor pre-1950 as being a forcing.

Under the assumption that WV is only a feedback to other changes in forcing, the impact is approximately 2 times no-feedback sensitivity of 3.7 Wm-2.  Using the cloud data as a proxy for water vapor I can have something like this.

   A certain water vapor/cloud anomaly could have an equivalent forcing/feedback with a SWAG estimate in the right hand axis.  Remember this is a tropical data set and that would include a lot of latent energy.  With CO2 equivalent gas forcing increasing almost exponentially after 1950, one would expect a significant shift in the trend.  Yes, there are tons of issues with this developing approach, but unless the data is total crap, there should be some indication of a trend shift.  If we have one degree of warming and the original ultra high sensitivity estimate of 4.5 C per doubling, 1.5C for CO2 and 3C for water vapor and cloud feedback, even with a 1 C increase in "global" temperature, there should be something other than a straight line.  People are criticized for using linear trends for projections but until the data hints at something else that is what you go with.  With the exception of the 1940s anomaly, cloud coverage is about as linear as it gets.  If clouds were either a strong positive or negative feedback then you would expect some curve that needed to be fit.  If clouds are a regulating feedback, you would just expect a leveling off at a controlling point/temperature.  Over 28C SST, is such a controlling temperature due to deep convective triggering.

28C is a bit misleading in the tropics though.  Only small regions hit the warm pool status and they tend to move around.  NINO1.2.3.4, IPWP WPWP, etc. have pseudo-oscillations with respect to one another because of the shifting.  Those shifts lead to other weather oscillations which are great for longer term weather forecasting but not so great for "climate" forecasting.  Some of the oscillation could last for centuries and there just isn't enough data, maybe.

 There are tropical region paleo reconstructions that seem to match SST especially, tropical SST which can be useful and something a bit different, BEST land surface temperature which scales nicely to SST and tropical SST.

As a lark I scaled BEST, CET, GMSL and OHC to the IPWP reconstruction by Oppo et al. 2009 which can extend sea surface temperatures back to ~1750.   Unfortunately, none of these data sets are truly "fixed".  Every one is subject to inter-related calibration issues.   There was recently a new interpretation of the sea level data, OHC is getting another look, HADCRUT4 now has a kriged addition, ERSST has a version 4, nothing is static.  However, tropical SST does tend to be adjusted less often.  Tropical SST also currently has a high, 90% or greater correlation with the latest version of surface temperature estimates.  Correlation though vary with smoothing and the oceans are naturally smoothed accounting for most of that higher correlation.  So while the new kids hunt for higher latitude more variable locations to "adjust" temperatures, the tropical oceans are not all that adjustable.

Just comparing the Oppo et al. 2009 IPWP to Mann-O-Matics latest shows how far apart the "science" is diverging.  So there is tons of work to be done to create a "convincing" argument, but those challenges should be part of the fun, doncha know.


Moving Right Along.  

The lower troposphere temperatures provide one energy reference and the absolute SST another.

Using the ERSSTv4 data set converted to rough Sea Surface Energy (SSE) via S-B law then so massaging, normalizing the two data sets to a 1950-1999 baseline and smoothing to 27 months I get this comparison.  Obviously, if I had used a 1850 to 1910 or so baseline, the cloud data would be about the same but the SSE curve would have an exponential rise.  That could be the right way around, then again if centuries of above normal Volcanic activity depressed tropical SST, this could be the correct orientation.  Looking at the 1941 to 1946 anomaly, decreased cloud cover could have caused SST warming, which increased cloud cover, which in turn over compensated for warming cause a reduction in tropical SST.  That is pretty much how a some what sloppy controller would function, over/under shoots, with recovery, with a little less over/under shoot.  There are other inputs and responses that would keep the relationship less than "ideal", but if you have experience in older slower control systems this makes perfectly good "sense".  Experience with older slower control systems would also provide experience in run away exponential responses that tend to destroy things and are unsustainable.  That would make the early baseline a bit less likely.  So according to this "guess" reasonable control would have been establish in the 1950 range and the "hunting" begins to subside.

A lot of this would be much easier if the variance in the data sets were reliable.  You really only have three options, increased variance leading to unstable operation, decreasing variance indicating acquisition of the controller to a better control point and rock solid stable which is pretty much unheard of.  The variance though in the data sets, mainly global data sets so increasing variance, but that appears to be related to the inclusion of south pole data circa 1955.  Adding the much more variable poles to the natural smoothed SST data creates some issues including the constantly shifting "adjustment" baselines.

There is nothing wrong with trying to compile a "global" anomaly, but then there is nothing wrong with avoiding a "global" anomaly depending on the focus of your evaluation.  Approaching this from a control theory perspective would mean isolating more critical "inputs" to reduce noise.  My approach would be to start at the primary heat source, tropical oceans and primary control mechanism, tropical clouds and work towards a "global" response instead of dealing with all the noise and constant changes.

As far as CO2 equivalent gas forcing and tropical clouds, there appears to be close to zero evidence that cloud directly respond to GHG but could indirectly respond to GHG forcing of tropical SST by increasing regulation at an easier to maintain set point, a control "sweet spot" if you will.

A control "sweet spot" is a point where the control function and response function are equally matched.  There, they are able to track more smoothly reducing the time required to "settle out" after a perturbation or the system is more efficiently damped resulting in less variance.  However you like to look at the situation.

This basic control observation can lead to a variety of "novel" approaches.  Fluctuations in seemingly random tend to change variance relationship leading to "strange attractors" in Chaos theory which are analogous to control "sweet spots".  These "sweet spots" generally require non-linear relationships so you could think of them as "hyperbolic" functions.  There are plenty of ways to skin this cat and the different jargon gets in the way of communicating a fairly simple and common relationship between functions.  So if you don't like my "sweet spot" create your own terminology.

A weakly damped response like the above using the Oppo 2009 detrended is very common.  The fit is less than perfect which is also very common.  But as you can seem the amplitude decreases with time as the system approaches its "sweet spot".  The differences are likely cause by erratic volcanic forcing timing.  That volcanic forcing could be related to orbital influences or could be more random, but the system seeking control will respond with some damped function provided it has potential to be stable.
The oscillatory response curve though will change depending on if the system is above or below set point.  There is no reason to expect a continuing stable oscillation, instead you should expect changing oscillatory behavior.

 You can compare the entire Oppo et al. 2009 reconstruction with the Stienhilber TSI reconstruction and see that nature is full of pseudo-cyclic oscillations.  A "control" oscillation would be different than a cooling response oscillation which would be different than a warming response oscillation.  In phase a relatively weak "forcing" can have much more influence than out of phase.  You can jump on the "Cyclomania" bus and declare victory, but real systems are a bit more interesting.

You can also jump on the magic bullet bus.  Assume there is only one significant driver, CO2 equivalent gases for example and ignore the inconvenient "divergences" like tropical clouds for example.