Guest Post by Willis Eschenbach
I’ve been messing about with the “Modtran” online calculator for atmospheric absorption. It’s called “Modtran” because it is a MODerate resolution program to calculate atmospheric infrared absorption written in ForTRAN, which calculates the result for each 1 cm-1 wide band of the wavenumber across the spectrum. Not quite a “line-by-line” calculation, but close. Here’s a sample of the input page:
Figure 1. User input page for the Modtran online calculation for infrared absorption. Left side is user input. Upper right graph shows absorption as a function of frequency. The lower right graph shows the GHG concentrations, pressure, and temperature, as a function of altitude. See here for an overview of the model. Click to enlarge
This shows the situation during the subarctic summer, with no clouds or rain.
Along the way, I ran into a curious mystery, one for which I have no answer.
Here’s the peculiarity I found. I decided to see what Modtran had to say about the “instantaneous forcing”. This is the forcing immediately after a change in e.g. CO2 or other greenhouse gas. In Table 1 of “Efficacy of Climate Forcings” , James Hansen et al. say that the instantaneous forcing from a doubling of CO2 is 4.52 W/m2.
So I tested that claim with Modtran using a variety of different locations, with different combinations of clear skies, cloud, and rain. I started by testing every few hundred PPMV increase, to see if the results were linear with the log (to the base 2) of the change in CO2. Finding that they were perfectly linear, I then tested each situation using 375 ppmv, doubled CO2 (750 ppmv) and two doublings of CO2 (1500 ppmv). I noted the absorption at each level, and compared that to the logarithm (base 2) of CO2. That let me calculate the forcing, which is typically given as the change in forcing for a doubling of CO2. Using Modtran, I get the following results:
Figure 2. Instantaneous forcing calculated by Modtran for different scenarios.
Now, this has the expected form, in that the forcing is highest at the equator and is lowest at the poles. The addition of either rain or clouds reduces the forcing, again as we’d expect, except during subarctic winter when some kinds of clouds increase the forcing slightly.
So the mystery is, according to Modtran, the absolute maximum instantaneous forcing from a doubling of CO2 is 3.2 W/m2 in the clear-sky tropics. I can’t find any combination of locations and weather that gives a larger value for the instantaneous forcing than that. And the minimum value I can find is subarctic winter plus cirrus, at 1.57 W/m2. I can’t find any combination giving less than that, although there may be one.
As a result, according to Modtran the planetary average instantaneous forcing from CO2 doubling cannot be any more than 3.2 W/m2, and is likely on the order of 2.4 W/m2 or so … but according to Hansen et al., the real answer is nearly double that, 4.5 W/m2.
So the mystery is, why is the accepted value for instantaneous forcing nearly twice what Modtran says?
Note that the answer to the mystery is not “feedbacks”, because we’re looking at instantaneous forcing, before any response by the system or any possible feedbacks.
All suggestions welcome, except those that are anatomically improbable …
w.
DATA: Excel spreadsheet here. You don’t need it, though. For any situation, simply use Modtran successively for two CO2 values where one CO2 value is double the other, and note the difference in the calculated upwelling radiation. This is the instantaneous climate sensitivity for that situation.
THE USUAL: If you disagree with me or someone else, in your comment please quote the exact words that you disagree with. This lets everyone know your exact subject of disagreement.
NOTE: I see as I finish this that they have an upgraded user interface to Modtran here … the results are the same. I prefer the older version, the graphics are more informative, but that’s just me.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Ferdinand Engelbeen says:
April 12, 2014 at 12:29 pm
In both cases (mine and Hansens) it’s from taking the current situation and instantaneously doubling the CO2, including all overlaps in both cases.
w.
FORTRAN? Now that brings back some memories. I didn’t know anyone used FORTRAN anymore. I wonder if they store the source code on punch cards? Just kidding, I loved FORTRAN.
It’s cuz de heat is hidin under de clams in de oceans, doncha know. And dats a tragedy!
tom watson says:
April 12, 2014 at 12:52 pm
Since you haven’t provided the slightest scrap of evidence for your claim that Modtran has significant errors, it seems that you are not intelligent enough to understand that on a scientific website, claims are meaningless without facts, logic, math, or observations to back them up.
Assuming that scientists might be wrong is a good thing. Assuming that they are idiots is always a mistake. Scientists have problems, to be sure … but lack of intelligence is not generally among them.
In fact, if you had taken the time to look at the manual and run the model and inspect the results, as most intelligent folks would do before commenting, or even read the caption to Figure 1, you’d have noticed that Modtran explicitly calculates the concentration of the variables at 33 of altitude levels, and uses that in the final calculation.
w.
The absorption curve should not be linear, check what it does with concentrations of 50, 100, and 200 ppmv.
daveburton says:
April 12, 2014 at 12:53 pm
Thanks, Dave. According to your data at the link, the 2xCO2 forcing is 3.2 W/m2 … which is identical to my results in Figure 1. So I’m not sure what you mean. I calculated the CO2 forcing from your data as the I_out for 285 ppmv minus the I_out for 570 W/m2.
w.
DHR says:
April 12, 2014 at 1:05 pm
Good question. It’s because the forcing is logarithmic (base 2). A doubling or a halving have inverse effects.
log2(8) = 3
log2(4) = 2
log2(2) = 1
So for example, dividing 4 in half or doubling it each change the log2 by 1 unit …
w.
Why do we have this discussion? It’s not CO2 and never has been. Wake up! It is the Sun.
Remember, in 1970, it was the Solar constant. Now with Soho, we find the EUV [10.7 cm flux] varies wildly. The Flux is used to determine the amount of atmospheric drag on satellites. The Flux puffs us the upper atmosphere creating a “insulation blanket”. The CO3 [ozone] reflects [absorbs] the infrared from both the Sun and Earth. The increase in “insulation” prevents heat from leaving the North and South pole. The decrease in the Flux allows heat to escape.
Watch the Antarctic set awesome new record ice extents due to the Flux being down [less than 140 during a Solar peak is a scary amount on low side].
Willis, try doubling CO2 from pre-industrial and see what you get.
@ur momisugly george e. smith says:
April 12, 2014 at 2:02 pm
…
And if they are assuming a log base 2 behavior, that means they believe Beer’s Law applies, and it doesn’t because Beer’s law presumes the absorbed photons stay dead, and are never re-emitted at any wavelength.
…
That isn’t an assumption of Beer’s law. If that were true, after a while the sample would transmit all light impinging upon it because all the molecules would have already absorbed photon. That isn’t how it happens – it is a dynamic equilibrium of absorption-emission.
Coldlynx says:
April 12, 2014 at 1:53 pm
You raise an interesting issue, Coldlynx. Modtran is a radiation model and doesn’t deal with heat balance and transfer other than indirectly. By that, I mean that you can adjust surface temperature, and the model recalculates atmospheric temperature (at 33 levels) assuming that the lapse rate is unchanged.
In other words, it cannot do what you propose because it is not included in the model. Note that this is not a shortcoming of the model. Every model only deals with certain things, you can’t escape it. It just means that you’ve picked the wrong model for what you want to investigate.
However, none of that matters to this study because I’m dealing with instantaneous forcing, before any atmospheric or surface temperature response.
Clue. If you have something to say, say it. Leaving clues doesn’t work.
Regards,
w.
What Roy Spencer says – 285 and 570 rather than 375 and 750. Probably a log relationship (or two) rather than linear.
Roy Spencer says:
April 12, 2014 at 2:59 pm
Interesting question, Dr. Roy. It turns out I had tested that before writing this post, and it’s the same, e.g. tropics clear is 3.2 W/m2 per doubling of CO2.
Following on from that, I greatly like this part of science on the web, which is that I get all this insightful feedback and interesting questions and suggestions. Since the question was interesting, I looked further at the bottom end of the CO2 concentrations and found the following results (again, tropics clear):
Doubling, Forcing
10-20 ppmv, 3.42 W/m2
20-40 ppmv, 3.49 W/m2
40-80 ppmv, 3.45 W/m2
80-160 ppmv, 3.36 W/m2
160-320 ppmv, 3.20 W/m2
By the time we get to the doubling from 160 to 320 ppmv, the change in forcing has stabilized at 3.2 W/m2. This agrees with my results looking at the doubling from 275 ppmv (pre-industrial) to 550 ppmv, which gave 3.2 W/m2.
w.
OK – When all else fails then either Hansen or Modtran is wrong. I know who I’m betting on.
Ralph Kramden says:
April 12, 2014 at 2:45 pm
Ah, yes, the famous Hollerith punch cards. The best part was after punching all 173 cards for your whiz-bang program, you turn them in to the resident Cerebus guarding the holy computer … and get them back three days later saying you made a mistake on card 11. A mistake which will require you to repunch perhaps a third of the cards …
But yes, of the three computer languages I learned during the 1960s (ALGOL, COBOL, and FORTRAN), only one survives, like some dinosaur who never got the news about the asteroid … I fear that since learning R, I have no interest in any of the others. Yeah, I still write in Basic and Pascal and Mathematica, but R is da bomb …
w.
Boy, does this bring back memories. I wrote a program at GSFC during 1972-1982 called the Synthetic Spectrum Program (SSP) (re: Rudy Hanel, Virgil Kunde, et al.) with parameters to describe the various atmospheres we wanted the simulate in the IR for remote sensing analyses: Earth, Mars, Jupiter and Saturn. It was written in FORTRAN. I remember a copy was sent to NCAR in the early 80’s. I verified that the program (substantively unchanged) was still in use in the mid-90’s. I wonder how much of that program is the basis for later models? The SSP did a very good job in a clear atmosphere. Clouds & aerosols were (and I think still are) a confounding problem that we were not able to solve then.
Willis,
You say “which calculates the result for each 1 cm-1 wide band of the wavenumber across the spectrum.”. I might believe 1um. Or 100nm. Or 10nm. (I know nothing of MODTRAN) But 1/1cm seems rather broad. At 1cm you are more into microwaves than light. Although really it is all light. But common nomenclature.
Willis Eschenbach says:
April 12, 2014 at 3:34 pm
I’m a Forth guy. But I play with bits. And controlling things. In real time.
Willis Eschenbach says: “Modtran is a radiation model and doesn’t deal with heat balance and transfer other than indirectly.”
Thanks for answering my Q before I could ask. And also thanks for your thoughtful inquiries. So Modtran could also mean MODEL Fortran?
Another case of “It’s Models All The Way Down.”
Has anybody actually, like, MEASURED this stuff? My default suspicion says : No, not really possible. But we sure can guess good.
P.S. I’m not a computer geek, just somebody who understands how really difficult it is to actually collect good data in the real world.
Wikipedia says:
If C/C₀ = 2, that is, for a doubling of atmospheric CO₂ concentration this formula gives 3.7 W/m². However, the term “radiative forcing” in this case is probably not equivalent to the “instantaneous radiative forcing” used by MODTRAN. The wiki value may be related to the IPCC usage.
Sure it is a rather contorted definition, because
1. the “tropopause” is not a well defined surface (see tropopause folds)
2. allowing one side to adjust while holding the other side fixed excludes empirical testing.
“Instantaneous radiative forcing“, although it can’t be tested either, is a bit better, because
This latter usage at least treats the entire atmosphere uniformly. Which is preferable, since heat capacity of the stratosphere is negligible, so the surface along which “forcing” is defined can be moved upward at leisure. Forcing at ToA (Top of Atmosphere) is the same as at the ill defined tropopause.
Now, the stratosphere is supposed to cool down with increasing GHG concentration. If this cooling is taken into account while the troposphere below is “held fixed”, it should increase radiative loss through the tropopause, because there is a larger temperature difference. Therefore the 3.7 W/m² given by the IPCC is an upper bound to instantaneous radiative forcing, which is consistent with your thought experiment using MODTRAN, but excludes higher values.
The NASA GISS ModelE Climate Simulations page you were referring to uses a somewhat more consistent termonilogy.
That is, they call “adjusted forcing” what is referred to as “forcing” by the IPCC. Now, their values are based on the output of Model E
Fi = 4.52 W/m²
Fa = 4.12 W/m²
Note the Fa > Fi relation does not hold for Model E, which is truly amazing. But hey, all models are wrong, but some are useful. And in this case the higher the values are the more fit for purpose they’ll be, provided the “purpose” is to generate scare.
Otherwise all such calculation, including MODTRAN is based on assumptions about the “average” atmospheric temperature/moisture profile in different regions and/or under different meteorological conditions, which may or may not be adequate.
I reckon even the IPCC adjusted forcing estimate of 3.7 W/m² is too large (derived for the US standard atmosphere or some such artificial constuct). Based on your fiddling with MODTRAN it should be below 3 W/m² (but probably above 2.4 W/m²).
Relation between reality and theory is always difficult. For instance we understand clearly why the stratosphere should cool down with increasing GHG concentrations. And indeed, according to RSS, during the past 35 years temperature of the lower stratosphere has decreased at an average rate of -0.28 K/decade. The only trouble is that almost all the decline happened close to the beginning of observations, during the last 2 decades this rate is only -0.03 K/decade, almost an order of magnitude smaller.
Willis;
[Loads fine for me, it’s the link in the “References” at the bottom of the page I linked to in the head post. -w]
>>>>>>>>>>>>>>
Nope, can’t load it. Even googled it, comes up with the link, can’t load it. Tried multiple browsers. I going to go all conspiracy and wonder if Canadian IP addresses are being blocked to prevent Steve McIntyre from looking at their data.
Not a big deal, but the 3.7 versus 4.x things is the real mystery to me. I don’t think they explicitly state it in AR4, I think you have to go back to AR3 or Ar2 to find the number, but 3.7 is what I’ve always seen quoted before.
I’m not sure if this has any relevance, but I notice Hansen defined fi to be the forcing at the tropopause, rather than TOA. MODTRAN in the screenshot appears to be set at 70 km altitude. Is it possibly being affected by absorption/overlap in the stratosphere?
Ralph Kramden says: “FORTRAN? Now that brings back some memories. I didn’t know anyone used FORTRAN anymore…”
I interviewed for a job about 15 years ago that involved patching and updating an extremely complex program that was still in FORTRAN.
M Simon says:
April 12, 2014 at 3:49 pm
Willis,
You say “which calculates the result for each 1 cm-1 wide band of the wavenumber across the spectrum.”. I might believe 1um. Or 100nm. Or 10nm. (I know nothing of MODTRAN) But 1/1cm seems rather broad. At 1cm you are more into microwaves than light. Although really it is all light. But common nomenclature.
———————————————————–
A reciprocal centimeter bandwidth is just an increment of the spectrum over which the calculations were performed. It is equivalent to going through the absorption bands in a fairly detailed way. At 10 microns in the NIR, once reciprocal centimeter would correspond to a wavelength band of about 1 Angstrom; i.e. 10^-10 meters.
Water vapour.