More Fun with Charts

UPDATE: More stuff at the end

Since the ocean data has less conversion issues, it is definitely a lot easier to work with.  Sea level is pretty much fixed and for the satellite era we have actual surface and layers to deal with.  So on the way to hopefully bigger and better things, this post expands on the "layers" method of teasing out sensitivities to different forcing events.

This is a bit of a rehash with some minor tweaks.  The Global Reynold's Oiv2 SST data set from 1981 to 2013 with the absolute temperature value, 291.38K is used as the reference for this comparison.  Using the standard Stefan-Boltzmann relationship, that absolute temperature is converted to effective surface energy in Wm-2.  Then using the UAH data, each temperature anomaly series is adjusted to the average absolute temperature then converted to effective Wm-2.  The actual temperature of the atmospheric layers is not equal to the SST, but using this adjustment, the effective Wm-2 variation of the atmospheric layer can be related to the sea surface energy flux variations.  The Lower Stratosphere data is inverted to simplify comparison.

Globally, there is no seasonal signal in any of the data.  Averaging globally natural smooths the season "signal".  When only hemisphere data is used, the Reynold's Oiv2 data has a strong seasonal cycle which needs to be removed for simple chart comparisons.  The satellite data doesn't have a significant season signal, so that has likely been removed before the data published online.  So the Hemisphere and other comparisons have a few more steps and more places for me to screw up.

So here is what it looks like.  For the Reynold's Oiv2 SST data the average of each month for the entire period is determined, the average of the entire period subtracted from the month averages producing a seasonal signal which is subtracted from the monthly values of the entire period.  Not as complicated as it may sound, all it is is removing the season anomaly leaving the absolute value intact.  Since that procedure was performed on the SST data, the same was performed on the others though the remaining season signal was small in the UAH data.  If the Reynold's Oiv2 and UAH had the same start date, there would likely not have been any residual seasonal signal at all.  I will try to add a spreadsheet later if anyone is extremely bored.

The result is a comparison where the actual forcing changes can be compared.  Some rescaling may be required, but it provides a reasonable picture of how the Stratosphere anomaly is an amplification of surface energy flux variation.

Here is the Southern Hemisphere done the same way.  The two volcanic events are obvious so I didn't include arrows.  In the NH and SH the peak forcing based on this reference method is about 10 Wm-2.  The Global peak is slight lower, closer to 9 Wm-2.  That difference is likely due to the seasonal signal removal either in my spread sheet or in the UAH process.  It is a indication that the accuracy of this method cannot be better than 1 Wm-2, though it is likely considerable worse than that.  What is most interesting is that the forcing of the two volcanic events is about the same in this comparison where the typical forcing estimates have the 1991 Pinatubo event being much larger than the early 1980s El Chicon event.  Estimates of direct and indirect volcanic aerosols are not proving to be very accurate and this coarsely confirms that they are off a touch.  In any case, this fairly simple comparison using an absolute temperature reference instead of temperature anomaly looks like it can get within a couple of Wm-2 of actual variations in atmospheric forcing. There are of course multimillion dollar platforms to do this "better", but it is kind of neat seeing how much can be done with some of the longer term satellite data series. 

Now don't get too excite sports fans, there is still a CO2 signature in there, it just looks a lot lower that those fat tail estimates. 


More Stuff:

Fun with Data - The Land Only Greenhouse Effect

While I prefer the Lower Stratosphere Data, the "surface" temperature data and the Lower Troposphere can be fun to compare.  The GISS "surface" for land is taken at not very uniformly distributed locations at about 3 meters (6 feet) above the actual "surface" at a variety of elevations with the "average" elevation around 680 meters.  The Lower Troposphere data is "filtered" and "weighted" from approximately the "surface" to around 5000 meters in altitude.  Despite totally different challenges they provide fairly accurate, +/- 0.2 C degree temperature anomalies for their not all that well defined "surfaces".

For GISS, the "average" temperature anomaly is the daily maximum - the daily minimum divided by two.  This has become "THE" indicator of climate change.  With the satellite era, there has been some grousing that the satellite "surface" data is not the same as the land "surface" data.  They should be different, they are measuring different "surfaces".  While they are measuring different "things", the trend relationship between the two should mean something.

The only thing that changes differently between the two is the amount of CO2, H2O and "other" stuff floating around in the atmosphere.  Since GISS is at "THE" surface and UAH is near "THE" atmospheric boundary layer, the biggest differences should be clouds, aerosols and CO2 in that order.

In the Chart above I have ploted GISS NH land only, the one with the scary linear regression, UAH NH land only, the one with the not as scary linear regression and the difference.  You should notice that the chart says Energy not temperature.  I converted the temperature anomaly to approximate Energy anomaly using 11 C as the approximate "surface" temperature for both.

There is considerable uncertainty in this comparison, but the difference should roughly approximate the atmospheric changes influencing temperature.  That 3e-02 is about 0.3 Wm-2 per decade or 3 Wm-2 per century.  If 3.7 Wm-2 produces 1C of warming, the 3 would produce less, ~ 0.8 C of warming.  Not a very accurate estimate, but in the general ballpark of the "no feedback" climate sensitivity for whatever that is worth.

The common trend would likely be due to something else.  Remember there is that whole tropospheric hot spot thing that is supposed to happening that is not.  If this trend difference is meaningful, then is should be approximately equal to the ocean energy imbalance.

Since Land only seems to be the big thing, this compares the hemispheres by the land only data.  I used UAH minus GISS this time, but note the similarity in the trends.  The trends are not pronounced enough given the noise and uncertainty to be significant, but that is a pretty strong correlation between hemispheres.

Now think about the differences between comparing the land to ocean rate and the land to lower troposphere rate.  Two different frames of references with atmosphere as the conduit. 

I could do the same with the oceans, but there is such a small difference because SST has less variation that it is not of much use.  That is were the Stratosphere comparisons have a large advantage. 

As you can see in this comparison, the surface, lower and middle troposphere are extremely close together.  Too close to confidently tell some event from instrumentation noise.  The lower stratosphere amplifies the signal naturally allowing a more realistic comparison.  This by the way was adjusted to the surface energy anomaly based on the absolute temperature.  By adjust all of the data sets to the same absolute temperature, this shows a "relative" response to a variation in surface energy anomaly.  While it is a bit noisy, you may be able to make out the "approach curve" shape in the LS data.

The values are still too coarse for much precision, but it is a neat way to make better use of the available data.

How to Splice Instrumental Data to Paleo Reconstructions?

That is a pretty deep question.  Just "eyeballing" the problem, the paleo data has a variety of natural and added smoothing methods that cause real time shifts in the data and suppress some portion of the signal.  Just splicing the instrumental period to a "best fit" seems to cause a good bit of grief since there are preconceived notions involved.   So when I looked at the Oppo 2009 Indo-Pacific Warm Pool (IPWP) temperature reconstruction, which was done the way it was done to make instrumental splicing simpler, I tried to guess which direction the Author(s) were going.

The IPWP recon was assembled in decade data points with 50 year averaging of the data points.  For sea floor sediment proxies, that is a good "standard" for the natural smoothing which is the time period required to form a sediment layer.  I was not all that trilled with the decade "pseudo data points" but on second thought, it might be an elegant solution for the splicing.

To check out how it might work I calculated the centered decade averages for the Hadley Center HADCRUT4 data set using the NH, SH and 3030 or tropical zone.  The overlap period for the Oppo 2009 data and HADCRUT4 is 1855 to 1955.  I used that as the anomaly baseline period.  That is what the instrumental data would look like in order to do the splice.

Because of the different smoothing periods, the splice is off a bit.

By shifting the IPWP recon back one decade, the splice appears to be better centered.  That shifts the anomaly baseline period a touch though.

So just to be thorough, this chart shows both the time shift and new anomaly baseline period, 1855 to 1945.  That is not a bad fit in my opinion, but I am sure there is probably a better way.

Tah Dah! There is the tropical instrumental to Indo-Pacific Warm Pool Splice.  Doesn't look much like the more famous hockey sticks for some reason.  If you are a fan of ENSO and long term persistence in climate though, it may be a lot sexier.

Data: 

WDC PALEO CONTRIBUTION SERIES CITATION: 
Oppo, D.W., et al. 2009. Makassar Strait 2,000 Year Foraminiferal SST and d18Osw Reconstructions. 

IGBP PAGES/World Data Center for Paleoclimatology Data Contribution Series # 2009-089. 

NOAA/NCDC Paleoclimatology Program, Boulder CO, USA.

http://www.cru.uea.ac.uk/cru/data/temperature/ 

UPDATE:

Here is another high resolution paleo reconstruction along with the Oppo 2009.
The Lake Tanganyila Lake Surface Temperature by Tierney et al. 2010.  This was a poster child for scary "unprecedented" warming.  Don't be too alarmed, the margin of error is pretty large in their reconstruction and "lake" is a buzz word for look out for land use impacts.  Tierney has a longer 66,000 year Lake Tanganyila reconstruction that only have a couple of data points over lapping this 1500 year recon.  There is a "gap" of about half a degree between the two.  As you can see, the variance, aka standard error, is larger than Oppo, but it still is possibly a useful data series.  

The Tierney 2010 data is not in consistent bins, the dates vary from about 18 years to over 30 years.  That means that the smoothing per data point would also be inconsistent.  So I arranged the Tierney data by date to the closest 10 year bin used by Oppo et al.  In order to Average, simple interpolation between dates with data was used.  In most cases only one date was missing data, but in a few there were two dates missing, so each received the same simple interpolation, not fancy spline fits and such.  This is the result with no temporal shift as above.  Even allowing for various things that could impact the accuracy of each reconstruction, there is evidence of a lead/lag relationship that is a little complex.  It looks like adding more reconstructions will smooth out most of the information producing a smoother shaft for a hockey stick.  The phase may shift between the reconstructions, but the range is remarkably consistent, that may be more important than the timing.