by J Storrs Hall
In my previous post, I argued that sea-surface temperatures hadn’t shown an inflection in the mid-twentieth century, and that the post-50’s rise was essently a land-based phenomenon. To take the analysis further, I thought I could try to find just what the climate signal from CO2 was. The method is to find a fit to the temperature record that included the CO2 forcing signature as a component, and see how big its contribution was compared to the other components of the fit.
First, the CO2. To get a curve since 1850, I got the estimated emissions from here, integrated for accumulation, scaled by matching to the Mauna Loa measured CO2 (red), and took the log for forcing. (No arguments, please; this is the bog-standard story. Let’s assume it’s true for the sake of argument.)
There’s clearly a knee in the curve ca. 1960. Also note that it’s been essentially straight since the 70’s — it’s the log of an exponential.
For components of the fit function, I used a cosine to capture the cyclicity we already know is in the record, a quadratic, and the forcing curve. I had used a second cosine before, and we know it produced two inflections in the result. The quadratic can only produce one, so the forcing curve has a better chance of matching the other one.
The idea is to find the overall best match and then look at the components to see how big the signal from the forcing is in comparison with the other components, which we will assume represent natural variability. We’ll plot each curve with the amplitude the optimizer gives it. Here’s what we get:
The blue line is the overall fit. Cyan is the 61-year oscillation, as before. No surprises here. Magenta is the quadratic, looking a lot like the sinusoid of the previous fit. Red is the CO2 forcing.
The CO2 forcing is upside down.
I gave the optimizer an initial guess for the forcing coefficient of 1; it came back with -1.67. This was, frankly, unexpected. I had seriously thought I would find some warming contribution from the forcing component.
So what on earth is going on? Here’s what we get if we add just the quadratic and the forcing curve:
For comparison, I’ve also plotted the second sinusoid from last time (green). It seems that the secular trend that the optimizer really, really wants is the shape of a Nike swoosh. If given only a quadratic to work with, it has to subtract the forcing curve to straighten out the twentieth-century rise. And it really, really wants the knee of the curve to be in 1890.
Does this mean that CO2 is actually producing a cooling effect? Absolutely not. It simply means that the secular rise in the twentieth century was a straight line, and the fit would do whatever it took to produce that shape. (This is why Pat Frank’s linear fit worked so well. As he noted, the linearity of sea-level rise would tend to confirm this.) What it does mean, though, is that there is no discernable CO2 warming signal in the HadSST temperature record. The (very real) twentieth century warming trend appears to have started about the time Sherlock Holmes was investigating the Red-Headed League.
Dave Springer says:
June 8, 2011 at 8:02 am
This has long been one of my concerns. We live in a world that wants to freeze.
John B says
“Yes, that is only an analogy. if you want more information, google “greenhouse effect”. There is a lot of science supporting CO2 as a significant greenhouse gas”.
John, I am afraid that most people here from experience do not believe google or wiki when it comes classifying CO2 as a GHG. I suggest you can start here
http://www.letterdash.com/HenryP/more-carbon-dioxide-is-ok-ok
When you are finished studying all of that, come back here, and learn, together with all of us….
J Storrs, I’ve put a comment in your previous thread showing why I think your original analysis is an irrelevant criticism of mine.
Your conclusion as stated here that, “the post-50′s rise was essently a land-based phenomenon depends on your unfit residual and on the validity of your model. Your model, in turn, is strongly influenced by extremely poorly-bounded mid-19th century SSTs.
Pardon me if I’m missing something, but what exactly did you show?
A Best fit of three functions
1= A 61-year oscillation that was derived from sea surface temperatures
2 = A logarithmic function to model CO2 forcing
3= A quadratic function to model everything else. Quadratic chosen because it models the bend at 1890.
In short, two functions created from the surface temperature data model better with a small negative logarithmic addition than with a positive. Short version, the quadratic fit of a line with two degrees of freedom fits better than a line with only one degree of freedom.
Now, while you can argue that there isn’t a discernable signal from CO2. The fact that you can get a better fit with 2-degrees of freedom rather than a 1-degree doesn’t really give you much of a proof. That’s before getting to the fact that it ignores damped signal and offsets. In fact, this is a blah result with extremely non-indicative results that I would expect from the alarmist camp more than WUWT.
Interesting idea. If this was an undergrad doing this, I would give it a good grade. However, I would expect more from a grad student, much less someone publishing on this site.
John F. Hultquist says:
June 8, 2011 at 9:04 am
Unlike water, CO2 is linear. Maybe you mean a triatomic molecule can bend, so energy can taken up in bending as well as rotation and stretching. Vibrational modes exist in molecules containing 3 or more atoms.
Why does everybody persist in using the hockeystick CO2 graph (the first one in this posting) which is cobbled together using two unrelated sets of data and promulgated from the false assumption that CO2 had been consistently low until recently? Comparing it with anything is a joke as it is false.
Why do sentient skeptics not realize that it would be totally unusual for anything like CO2 to be so constant and low over time? Why is the work of Ernst Beck and the many data sets produced by valid and reputable scientists over the last 200 years ignored and the spurious, cherry-picked low CO2 average for the 1800s by Guy Callendar given any credence at all?
Let’s try something more realistic and point out that CO2 has been much higher than now during three periods of the last 200 years, most recently in the 1940s and then temperatures crashed while CO2 was high. This most seriously repudiates the CO2-causing-warming link. High CO2 not only cannot cause warming, but it also cannot maintain it.
Dave Springer says:
June 8, 2011 at 9:14 am (Edit)
@tallbloke
Woud you agree that if anthropogenic activity did something to change the average turbidity (or rather the lack of turbidity) of the mixed ocean layer this would then alter the mixed layer energy budget? It seems like greater turbidity would cause sunlight to be absorbed in a shallower layer and then if conduction, convection, wind and wave mixing remained equal those mechanisms would be more effective in bringing the warmed water to the surface where it can cool.
Hi Dave. I think the ocean has a lot of counterbalancing things going on in it. Outside the tropics, (where the surface waters are very clear,) and most of the heat absorbed from the sun gets absorbed, the near surface water can get naturally turbid through plankton growth. This is limited by iron availability, but when a volcano goes off, the sea downwind of the plume is soon teeming with phytoplankton. These little critters fix co2 in their shells and take it to the bottom when they die, or it falls in whale poo etc. The co2 cycle in the ocean can take a thousand years or so. I suspect part of the C20th rise in co2 is down to the medieval warm period.
The ocean can only cool at the rate the atmospheric blanket lets it. Good job too, because there’s as much heat energy in the top two meters as there is in the whole atmosphere. If it could all escape quickly, we’d be prawned alive. So heat which can’t escape actually gets mixed downwards, which is why there is a linear dropoff in temp all the way down to the thermocline from the bottom of the mixed surface waters. This solar energy can stay in the ocean for millions of years, literally.
As Anthony put it a couple of years ago, the ocean is one big assed heat flux capacitor.
The interglacials are just like big scale el ninos, on a 100,000 year cycle rather than a decadal cycle. The ocean soaks up heat, and then when insolation falls, whammo, El nino, or in the case of the Milankovitch cycles, interglacial.
http://tallbloke.files.wordpress.com/2011/06/interglacial-elnino.jpg
OK S. says:
June 8, 2011 at 10:00 am
…
Thanks. I am advised to use Ctrl V .
It works fine
Ben, presuming you’re referring to my post here, recall that my analysis there depended on a prior analysis showing that a new 60-year oscillation appeared in the global air temperature anomaly record when SSTs were added to the GISS land-only global air temperature anomaly data set. This observation sparked my subsequent analysis that in turn led to the sensitivity estimate.
The presence of this net new oscillation in the anomalies, following SST introduction, provides a physical basis for inferring that a ~60-year cyclic ocean thermal signal exists in the 130-year global average surface air temperature trend. That, in turn, makes the oscillations that I did find more than just a numerical convenience.
The oscillations that turned up in the two anomaly data sets had about the same period as the net difference oscillation in the GISS data, but the phases were different. However, there’s no reason to think that parent net global ocean thermal phases will coincide with the phase of a difference oscillation.
As I pointed out, J Storr’s analysis was strongly determined by very poorly constrained mid-19th century SSTs. His 259 year oscillation has a period twice the length of the data set, which makes it also poorly constrained and hardly different from a numerical convenience. He also neglected to show us which part of the 259-year cosine phase actually played a part in the fit. Maybe it just mimics a linear rise during the 20th century.
In any case, when the (physically justifiable) oscillations were subtracted away from either of the the global temperature anomalies, all that was left was a linear rise in anomaly temperature; virtually the same linear rise in both data sets, as it turned out. The rest of the analysis followed directly from that, and in a very straight-forward manner.
So, with respect, you did miss something, and the rest of your post starting with this, “In fact, this is a blah result …” was intemperate.
When volcanos go off, I figure the simple fact that 2/3 of surface is water mean less SW hitting the oceans.
While temperature increases the capacity of air for water vapor (and therefor increases the average), IIRC light incident on water is the primary driver of actually water vapor increase.
Cloud formation also dries the air (and releases latent heat above the surface to be radiated mostly way).
So, rather than decreased air temperature, I think the primary driver of water vapor decline during volcanic activity is less SW light in the lower troposphere.
(I’m suggesting that studies of Pinatubo etc. overstate the relationship of temperature and humidity. Again, they likely get a large part of the correlation backward. The drop in humidity is probably more due to the decrease in the mechanism that heats the surface, rather than the temperature drop itself. The research papers’ adjustments are probably much smaller than the reality.)
(adjustments for volcano, not temperature, drying)
(Very important considering that normally most aerosols for cloud nucleation are over or near land. Big volcanoes scatter them over oceans where cloud nucleation particles are scarce and cloudless albedo is very, very low.)
maybe someone can ‘splain this to me, i am obviously an idiot, it must be completely obvious:
C02, based on Mauna Loa (I have not checked other sites but I am assuming this is roughly representative) is increasing at about a .5%/yr rate, and appears to be highly variable on a year-over-year basis. Methane is not increasing based on NOAA statistics.
However, global oil (and other fossil fuel) consumption is rising at 1.5% rate. So all else equal I would expect either: man-made CO2 was a large fraction of global atmospheric CO2, and so it was going up at the same % rate as fossil fuel consumption; or its a small but growing fraction, in which case not only is CO2 going up, the the rate at which its going up asympotically converges to 1.5%, the rate of growth of fossil fuel consumption. So there should be an *increasing trend in CO2 growth rate* – which I fail to see.
So given the variability of CO2 and methane, and the fact that growth does not trend up or track population growth/ energy consumption, how can i conclude its from anthropogenic activities?
This has me puzzled. Generally, I expect man-made CO2 growth to track the business cycle closely. The reason is that vehicle miles, manufacturing, etc, are all energy intensive and business cycle sensitive (for example, during a recession, steel and aluminum manufacturers shutter facilities consuming less coal and oil for electricity and transport).
But, um I fail to see that in the data, and the correlations look poor. Which makes me wonder how much of the atmospheric CO2 0.5%/yr growth is really man-made.
anyone?
So let’s see if I understand….the quadratic term can magically distinguish between CO2 forcing and all other forcings?
That is what you have claimed. That the quadratic term has accounted for ONLY non co2 forcing?
Hoser says:
June 8, 2011 at 1:22 pm
John F. Hultquist says:
June 8, 2011 at 9:04 am
Right, and I know better. A favorite link:
http://www.wag.caltech.edu/home/jang/genchem/infrared.htm
Of course CO2 causes global cooling.. only have to look at Mars.
95% CO2 and very cold.
But as we know if also causes Global warming…
Look at Venus.. 95% CO2 and very hot…
its the gas for all seasons.
John F. Hultquist says:
June 8, 2011 at 7:02 pm
Hoser says:
June 8, 2011 at 1:22 pm
John F. Hultquist says:
June 8, 2011 at 9:04 am
Right, and I know better. A favorite link:
http://www.wag.caltech.edu/home/jang/genchem/infrared.htm
No, you just think you do.
“Does this mean that CO2 is actually producing a cooling effect? Absolutely not.”
If the increase in CO2 were exponential and if the effect on temperature were logarithmic of the right magnitude, then the resulting temperature increase would be expected to be linear. But if the increase in CO2 is linear, but the effect on temperature were logarithmic then the resulting temperature increase should decelerate.
Well your four different colored lines of the “components of the fit” graphs all look phony to me. That faint jagged light bluish stuff (or izzat green) actually looks more like real observed data to me. The rest is pure fiction and tells us nothing.
Now the jaggy blue/green graph looks like it has a whole lot of data points. It is a fairly elementary mathematical exercise to artificially construct a functional fit to all of those data points using a function that has fewer parameters than the number of data points to be fitted. And of course that function is not unique, Many such functions can be created, from sinusoids (Fourier); but one could also use Bessel functions, or Legendre polynomials, or Tchebychev polynomials, or any other set of orthonormal functions.
I don’t see any value in that; it gives no insight into a causal structure. Ultimately the jaggy data is the best representation of what actually happened.
“”””” Ric Locke says:
June 8, 2011 at 6:03 am
In the field that used to be my profession, data points (surveyed ground locations, including elevation) are difficult and expensive to collect. It is therefore useful, from the standpoint of cost-effectiveness, to collect a few points and interpolate the intermediate values. “””””
Well it is usually called sampled data theory, and there generally is no reason to believe that interpolation gives good values for data values in between actual observed samples. That is specially true when the sampling regimen comes no where near complying with the Nyquist criterion for proper sampling of band limited signals. On the other hand, correct sampling in accordance with the Nyquist criterion, allows (in principle) exact reconstruction of the complete band limited function. In which case any intermediate values can be calculated.
Why sampled data theory is not taught in freshman science courses, is completely beyond my comprehension.
So much garbage is generated under the name of science; by “researchers” with no understanding of data sampling.
“For components of the fit function, I used a cosine to capture the cyclicity we already know is in the record, a quadratic, and the forcing curve. ”
The reason for your surprising result is that you don’t understand your “forcing” basics.
A forcing is a power term and will affect the rate of change of temperature , not the the temperature.
If you integrate your linear approximation the the CO2 forcing it will give you a quadratic. It is the quadratic that you need to be attributing to CO2 NOT your linear term.
Despite the enthusiastic applause from the audience, your article is fundamentally wrong. You probably should try to understand the basic physics and post a note to your article that will obviously mislead a lot of less trained people reading it.
Atmospheric CO2 concentration is much more temporally variable than we think. It was probably higher than today at the beginning of Holocene, 10,000 years ago. Since the general temperature trend for the last ~10,000 years is down, CO2 trend is also down. In 1940s it was higher than ~280 ppm, probably closer to todays values.
Ice core record is not reliable (accuracy). It cannot produce short peaks (low-pass filter). That ~800 years lag is likely some kind of an artifact and does not necessarily have any physical meaning.
What a joker.
Could J Storrs Hall or anyone else tell me where the data for the red line representing CO2 forcing came from. Or is is a fudge factor desighned to make the composite curve fit.