IntroductionMODTRAN is a spectral band radiative transfer code first developed in the late 1980s by Spectral Sciences, Inc. in partnership with Air Force Geophysics Laboratories. It was essentially a higher resolution (1.0 cm-1) version of AFGL's LOWTRAN code which integrated radiation transfers over 20 cm-1 wavebands.
The most recent version, 5.2, was released in 2009 and is proprietary. A single license runs $1200, technical support and updates are extra. Too rich for this hacker.
Older versions are in the public domain with the full FORTRAN source code available for anyone to download, compile, and execute on a local machine. Too much hacking for this poseur.
Some smart folks at U. of Chicago have been kind enough to host a web-enabled copy of MODTRAN3 Version 1.3 12/1/95, which sports a simple and mostly intuitive GUI and outputs some pretty pictures:
|Figure 1 - The graphical output for a model run with default options set. Isn't it purty?|
Updates happen automagically every time an option is changes, and it's wicked quick -- it updates almost faster than I can blink. Even more exciting, the Show Raw Model Output button works as advertized, and pops open a new tab (in Firefox on Ubuntu) with plain ASCII tab-delimited output. Copypasta into a spreadsheet application is then dirt simple -- just the right amount of hackery for this Gnumeric/Excel guru/data junkie.
It simply begs for experimentation. So I did, repeatedly. The number of different things I have thought to do with it exceeds my ability to cram into a reasonably-sized post -- and likely the attention span of any putative audience -- so this post is to be a short introduction of how I use the thing and how it compares to real-world observation.
I should probably quit blathering now and get on with it ...
MODTRAN vs. RealityThe About this model link contains this pretty picture (caption from original) ...
|Figure 2 - MODTRAN results (red) compared with data (solid black) from the Nimbus 3 IRIS instrument from Hanel et al., 1972).|
... which is like, impressively good. The atmospheric window regions from 8-9 and 10-13 microns follow the theoretical blackbody emission curve between 315-320 K, implying a surface temperature between 42-47 °C or 107-116 °F ... otherwise known as damn hot.
The big dip in the 15 micron region is our friendly neighbourhood CO2 (here at 325 ppmv as it was circa 1970-1972) completely gobbling up outgoing photons from the surface ... plus everywhere between there and about 12 km above terra firma. Why 12 km? The 15 micron absorption band touches the 220 K blackbody emission curve, which is -53 °C or -65 °F. Not only is that bloody cold, it's the "standard" temperature of the tropical troposphere at about 12,000 m, or 39,000 feet above the surface ... a place where only jet aircraft thrive.
"Aha," you say. "220 K is ALSO the temperature of the stratosphere at about 70 km, are we not seeing 15 micron photons from there as well?" Short answer is yes. Long answer is the topic of Part 2 (or 3, depending on how many more digressions I chase) of this series, so hold that thought.
The modelled 10-13 micron window runs more toward the cool end of the range, implying that the retrieval in that band is biased a bit hot, the model a bit cool, or some combination of both. If the model, it could be anything from parametrization/vertical profile issues to oversimplified physics.
It could also be completely wrong physics of course, but as the Air Force presumably first developed this beast with the aim of anticipating what happens to aircraft (and/or missiles) at high altitude/velocity atmospheric penetration, I'm guessing their science/engineering teams weren't full of total cranks.
Overall, fidelity to observation warrants confidence that this transfer code reasonably represents reality.
But I want ...
MOAR MODTRAN vs. RealityThe much cited and apparently excellent A First Course in Atmospheric Radiation by G.W. Petty contains this figure, familiar to many climate warriors:
|Figure 3 - Clear Sky and Thundercloud spectra from a satellite somewhere over the Western Tropical Pacific circa 1970-72. Credit: G. W. Petty (2004)|
The Planck distribution curves use the temperature of the standard atmosphere at the relevant level, which are 302 K for the surface and 218.9 K at 70 km. The model says that surface emissivity is assumed to be 0.980, which I take into account (hopefully correctly).
The surface temperature is a free parameter in this implementation of the model, I chose 302 K to fit the observational curve. Default for the standard tropical atmosphere is 299.7 K. The input form asks for this parameter to be adjusted using an offset value, so 2.3 is the value I put there to get 302 K.
The transmittance plot represents the fraction of radiation at a given wavelength hitting the simulated sensor which was directly emitted from the source -- in this case the surface. 1 is the maximum value, which means that 100% of the incident photons were emitted directly from the surface target and arrived unimpeded. 0 is the minimum value, and means that 100% of the incident photons were emitted from something OTHER than the surface.
This is important: 0 transmittance does NOT mean that NO photons at a given wavelength are striking our hypothetical instrument. NOR does it mean that absorption at that particular waveband is "saturated" between the surface and the sensor. It only means that whatever radiation shown in the intensity plot for a given wavelength was emitted by some other layer of atmosphere, NOT the surface.
Transmittance for a skyward-pointing sensor would not make sense -- the only relevant targets would be the Sun or the cosmic background radiation, both of which are at frequencies with little overlap in the terrestrial longwave spectrum which is the sole focus of this version of MODTRAN. Somewhat confusingly, the code still returns transmittance in the data dump for an upward-looking model run, but the values are exactly the same as they'd be for a downward looking run at the same altitude.
Finally, the LW flux figure in the lower left corner is the integrated flux in W m-2 looking up, looking down, and the absolute value of the difference between them. In either case, chunking those values into the Stefan-Boltzmann relationship between radiative power and the 4th power of temperature gives -- NOT necessarily the temperature at a given atmospheric level -- but the so-called "effective" temperature of the radiative flux our simulated spectral sensor is "seeing" through all layers of atmosphere to the target.
Normally we'd expect a sensor pointed toward the surface to return a higher integrated flux value than one pointed skyward, in the case of the above plot this is NOT true. Reason being that a real sensor on the real surface looking up is picking up mostly radiation from lower, warmer layers of atmosphere than one at 70 km looking down, which is seeing more radiation from higher, and therefore (generally) cooler atmosphere. Further instalments of this series will show up/down views from the SAME altitude; however, the balance of plots in this post will hold altitude at 0 km looking up and 70 km looking down for apples-to-apples comparison.
Yes, there are MORE plots for this post, hopefully I'll remember what brevity and concision mean as I describe them.
MODTRAN vs. Sliced and Diced RealityEver wondered what an atmospheric absorption/emission spectrum might look like if CO2 were the only major LW-active constituent? Turns out there's an app for that:
|Figure 5 - MODTRAN CO2 Only, 325 ppmv, surface temperature 302 K, tropical atmosphere.|
Compare to the theoretical blackbody temperatures of 283 and 271 K. Am I really telling you that if we magically knocked out water vapour, methane and ozone from the real system, that the atmospheric temperature would fall 76 K (137 °F), or that the surface temperature would RISE 17 K (31 °F)? (!)
No, that's not what I'm saying. Remember -- what our imaginary sensor is seeing in this model is not a "real" temperature, it's just the temperature of what the target "looks" like based on the incident radiation hitting it.
The case of the flux from ground level looking up does make some sense, without (primarily) water vapour in the atmosphere, there would be significantly less "back-radiation" being emitted from the atmosphere, and much of what the sensor would be seeing is the cosmic background radiation of deep space, which has an apparent ("effective") temperature of about 2.76 K ... really chilly.
The situation from 70 km looking down may be less intuitive, but makes sense (to me) with a little thought: removing all LW-absorbers from the atmosphere except CO2 leaves huge swaths of the spectrum which were previously impeded on their way to outer space open for those photons to now take a straight, unfettered shot to 70 km. The transmittance curve confirms this: practically everything outside CO2's main 13-19 micron absorption band has a transmittance of nearly 1.
Assuming constant solar input, all that radiation once hindered from escaping would rush out like air leaving a slashed tyre, and surface temperatures would fall. Dramatically. How much? MODTRAN doesn't do that estimate for us -- the input parameter for surface temp is 302 K, and it "stubbornly" keeps it there. I'll walk through how to derive the new theoretical equilibrium temperature in the next post in this series.
Notice also that outside the main CO2 absorption/emission band we see some non-zero radiant intensity for the blue "looking up" curve. Those are NOT due to CO2, but rather to the default parameters for CFCs, aerosols, dust and what have you that this implementation of MODTRAN (somewhat annoyingly) does not allow the user to futz with.
Some say that water vapour is the most important "greenhouse" gas in our atmosphere ...
|Figure 6 - MODTRAN H2O Only, surface temperature 302 K, tropical atmosphere.|
Much ado has been made about flatulent cows, warming oceans releasing methane clathrates from sequestration and thawing tundra spewing methane from suddenly not-permafrost because methane is -- ppm for ppm -- a far more potent GHG than CO2 ...
|Figure 7 - MODTRAN CH4 Only, surface temperature 302 K, tropical atmosphere.|
Last on the list for this section is ozone ...
|Figure 8 - MODTRAN O3 Only, surface temperature 302 K, tropical atmosphere.|
Well, no, probably not, because then you'd have to use more CFC-compressed spray-on sunscreen and then we might get one of them there vicious cycles going. Adding more stuff to the atmosphere to mitigate the consequences of the CO2 we've already put there is a double-down on uncertainty for one thing ... we know what CO2 @280 looked like ... a bit chilly but livable FOR CERTAIN.
Oh look, I digress again ... and I've got, like another 6 plots slated for this post. I'll spare us all and only do 4.
MODTRAN and Alternate Realities... but hopefully not actual future realities. Seriously, you really don't want to do what I'm about to show you to OUR atmosphere:
|Figure 9 - MODTRAN CO2 3,200 ppmv, surface temperature 302 K, tropical atmosphere.|
|Figure 10 - MODTRAN CO2 only 3,200 ppmv, surface temperature 302 K, tropical atmosphere.|
Doesn't look bad you say? For the record, 3,200 ppmv is three doublings of 400 ppmv. The worst case IPCC scenario from AR5 (RCP8.5) only calls for CO2 equivalent (all GHGs and aerosols) of 2,641 ppmv by 2225 or so, and 1,231 ppmv equivalent by 2100. They're of the opinion that even 1,200 ppmv is a Bad Idea. I'll compare their projections for 2100 to what MODTRAN says we'd expect at the same level in a future post.
For now, I'll sign off with something that really isn't within the realm of possibility even if we burned all the oil and coal still in the ground:
|Figure 11 - MODTRAN CO2 965,00 ppmv, surface temperature 302 K, tropical atmosphere.|
|Figure 12 - MODTRAN CO2 965,00 ppmv, surface temperature 302 K, tropical atmosphere.|
And ... even if it didn't spread out across multiple spectral bands as shown, impeding the path of photons trying to get to TOA is a simple matter of adding more absorbers. The ONLY limit is how many of them you can pack into a given column of atmosphere, which according to the maths of the Beer-Lambert law is theoretically infinite -- just make the path longer, or put more absorbers in the same volume -- optical depth of a medium does not asymptotically approach any upper value as either of those two parameters increase.