New Computer Fund

Saturday, November 22, 2014

The Problem with Changing your Frame of Reference from "Surface" Temperature to Ocean Heat Content

The few of you that follow my blog know that I love to screw with the "geniuses" aka minions of the Great and Powerful Carbon.  Thermodynamics gives you an option to select various frames of reference which is great if you do so very carefully.  Not so great if you flip flop between frames.  The minions picked up the basics of the greenhouse effect fine, but with the current pause/hiatus/slowdown of "surface" warming, they have jumped on the ocean heat content bandwagon without considering the differences that come with the switch.

When one theorizes about the Ice Ages of Glacial periods, when the maximum solar insulation is felt at the 65 north latitude, the greater energy would help melt snow and ice store on land in the higher northern latitudes.  (Update:  I must add that even the shift to maximum 65 north insolation is not always enough to end an "ice age".)  Well there is more land mass in the northern hemisphere, so with more land benefiting from greater solar, what happens to the oceans that are now getting less solar?  That is right sports fans, less ocean heat uptake.  There is a northern to southern hemisphere "seesaw" because of the variation in the land to ocean ratio between the hemispheres.

So the Minions break out something like this Holocene temperature cartoon.

Then they wax all physics-acal about What's the Hottest Temperature the Earth's Been Lately moving into how the "unprecedented" rate of Ocean Heat Uptake is directly caused by their master the Great Carbon.  Earth came from the word earth, dirt, soil, land etc.  The oceans store energy a lot better than dirt.

If you want to use ocean heat content, then you need to try and reconstruct past ocean heat content.  The tropical oceans are a pretty good proxy for ocean heat content so I put together this reconstruction using data from the NOAA Paleoclimate website.  Based on this quick and dirty reconstruction, the oceans have been warming during the Holocene and now that the maximum solar insolation is in the southern hemisphere, lots more ocean area, the warming of the oceans should be reaching its upper limit.  Over the next 11,000 years or so the situation will switch to minimum ocean heat uptake due to the solar precessional cycle.  Pretty basic stuff.

With this reconstruction, instead of trying to split hairs, I just used the Mohtadi, M., et al. 2011.
Indo-Pacific Warm Pool 40KYr SST and d18Osw Reconstructions. which only has about 22 Holocene data points to create bins for the average of the other reconstructions, Marchitto, T.M., et al. 2010.
Baja California Holocene Mg/Ca SST Reconstruction, Stott 2004 Western Tropical Pacific and the two,  Weldeab et al. 2005&6 equatorial eastern and western Atlantic reconstructions.  There are plenty more to choose from so if you don't like my quick and dirty, go for it, do yer own.  I did throw it together kinda quick so there may be a mistake or two, try to replicate.

I have been waiting for a while for a real scientist to do this a bit more "scientifically", but since it is raining outside, what the heck, might as well poke at a few of the minions.

Update:  When Marcott et al. published their reconstruction they done good by providing a spread sheet with all the data.  So this next phase is going to include more of the reconstructions used in Marcott et al. but with a twist.  Since I am focusing on the tropical oceans, Mg/Ca (G. Ruber) proxies are like the greatest thing since sliced bread.  Unfortunately not all of the reconstructions used extend back to the beginning of the Holocene.  The ones that don't will need to be augmented with a similar recon is a similar area if possible or they are going to get the boot.  So far these are the (G. Ruber) reconstructions I have on the spread sheet.

As you can start to see, the Holocene doesn't look quite the same in the tropics.  There isn't a much temperature change and some parts of the ocean are warming throughout, or almost and some are cooling.  The shorter reconstructions would tend to bump the end of the Holocene up which might not be the case.  That is my reason for giving them the boot if there are others to help take they a bit further back in time.  As for binning, I am going to try and shoot for 50 years if that doesn't require too much interpolation.  Too much, is going to be up to my available time and how well my spread sheet wants to play.  With 50 year binning I might be able to do 30 reconstructions without going freaking insane waiting for Open Office to save every time I change something.  I know, there are much better ways to do things, but I am a programming dinosaur and proud of it.

Update: After double checking the spread sheet, a few of the shorter reconstructions had been cut off due to the number of points in my lookup table.  After fixing that, the shortest series starts 8600 years before present, 1950.

That is still a bit shorter than I want but better.  The periods where there aren't enough data points tend to produce hockey sticks upright or inverted which tends to defeat the purpose.  So until I locate enough "cap" reconstructions, shorter top layer or "cap" reconstructions can bring data closer to "present" and lower frequency recons to take data points back to before the Holocene starting point, I am trying Last Known Value, instead of any fancy interpolation or curve fitting.  That just carries the last available data value to the present/past so that the averaging is less screwed up.  So don't freak, as I find better extensions I can replace the LKV with actual data.  This is what the first shot looks like.

Remember that ~600 AD to present and 6600 BC and before have fill values, but from 6500 BC to 600 AD the average shown above should be pretty close to what actual was there.  The "effective" smoothing is in the ballpark of 300 years, so the variance/standard deviation is small.  Based on rough approximation, a decade bin with real data should have a variance of around +/- 1 C.  Also when comparing SST to "surface" air temperature, land amplifies tropical temperature variations.  I haven't figured out any weighting so far that would not be questioned, but weighting the higher frequency reconstructions a bit more would increase the variation.  In any case, there is a bit of a MWP indication and possibly a bigger little ice age around 200-300 AD.

Correction, +/- 1 C variance is a bit too rough, it is closer to 0.5 C for decade smoothing ( standard deviation of ~0.21 C) in the tropics 20S-20N that I am using. For the reconstruction so far the standard deviation from 0 to 1950, which has LKV filling is 0.13 C.  So instead of 1 SD uncertainty, I think 2 SD would be more appropriate estimate of uncertainty.  I am not at that point yet, but here is a preview.

A splice of observation with decadal smoothing to the recon so far looks like that.  It's a mini-me hockey stickette about 2/3rds the size of NOAA cartoon.  The Marcott "non-robust" stick is mainly due to the limited number of reconstructions making it to the 20th century which LKV removes.

Now I am working on replacing more of the LKV fill with "real" data.  One of the reconstructions that I have both low and high frequency versions of is the Tierney et al. TEX 86 for Lake Tanganyika which has a splicing choice.;  The 1500 year recon is calibrated to a different temperature it appears than the 60ka recon.  Since this is a Holocene reconstruction, I am "adjusting" the 1500 year to match the short overlap period of the longer.  That may not be the right way, but that is how I am going to do it.  There are a few shorter, 250 to 2000 year regional reconstructions that I can use to extend a few Holocene reconstructions, but it looks like I will have to pitch a few that don't have enough overlap for rough splicing.  Here is an example of some of the issues.

This is the Oppo et al. 2009 recon of the IPWP that I use very often because it correlates extremely well with local temperatures combined with the lower resolution Mohtadi 2010 recon of the same region.  They overlap from 0AD to 1950, but there is very little correlation.  Assuming both authors knew what they were doing, there must be an issue with the natural smoothing and/or dating.  Since both are in C degrees there would be about +/- 1 C uncertainty and up to around +/-300 years dating issues.  If I wiggle and jiggle to get a "better" fit, who knows if it is really better?  If I base my uncertainty estimate on the lower frequency recon of unknown natural smoothing, I basically has nice looking crap.  So instead I will work under the assumption that the original authors knew what they were doing and just go with the flow, keeping in mind that the original recon uncertainty has to be included in the end.

With most of the reconstructions that ended prior to 1800 "capped" with shorter duration reconstructions from the same area, often by the same authors, things start looking a bit more interesting.

Instead of a rapid peak early and steady decline, there is more of a half sine wave pattern that looks like precessional cycle solar peaking around 4000 BC then starting a gradual decline.  Some of the abrupt changes, though not huge amplitude changes in the tropics, appear around where they were when I was in school.  There is a distinct Medieval Warmer Period and an obvious Little Ice Age,  I am kind of surprised the original authors of the studies have left the big media reconstructions to the newbies instead of doing it themselves.

I have updated the references and noticed one blemish, Rühlemann et al. 1999, uses an Alkenones proxy with the UK'37 calibration.  Some of the "caps" are corals since there was not that many to chose from.  Since the corals are high resolution, I had to smooth some to decade bins to work in the spreadsheet.  I am sure I have missed a reference here or there, so I will keep looking for them any any spreadsheet miscues that may remain.  Stay tuned.


Since there is a revised version here is how it compares to tropical temperatures.  I used the actual temperatures with two scales to show the offset.  The recon and observations are about 0.4 C different and of course the recon is over smoothed compared to the decade smoothed observation.  Still a mini-me hockey stick at the splice but not as bad as most reconstructions.

references: No: Author

31 Kubota et al., 2010 (32)
36 Lea et al., 2003 (36)
38 Levi et al., 2007 (39)
40 Linsley et al., 2010 (40)
5 Benway et al.,2006 (9)
41 Linsley et al., 2011 (40)
45 Mohtadi et al., 2010 (43)
60 Steinke et al., 2008 (56)
62 Stott et al., 2007 (58)
63 Stott et al., 2007 (58)
64 Sun et al., 2005 (59)
69 Weldeab et al., 2007 (65)
70 Weldeab et al., 2006 (66)
71 Weldeab et al., 2005 (67)
72 Xu et al., 2008 (68)
73 Ziegler et al., 2008 (69)

Not in the original Marcott.
10(1) with Oppo et al. 2009
36(1) with Black et al. 2007
38(1) with Newton et al. 2006
74 Rühlemann et al. 1999
75(1)Lea, D.W., et al., 2003, with Goni, M.A.; Thunell, R.C.; Woodwort, M.P; Mueller-Karger, F.E. 2006

No comments:

Post a Comment