I'm an aeronomer, I guess

My research for the last few years has been in a field I hadn't heard of until I entered it: aeronomy. There are a whole host of processes going on in various parts of the atmosphere that prodcue airglow (like the aurora but less profoundly, breathtakingly dynamic). Most of my research so far has been learning how we can turn airglow observations into concrete measurements of the atmosphere producing that glow. At least, concrete measurements are the dream, but the problem is actually very tough to solve. We can't directly measure the glow coming from the specific part of the atmosphere that interests us. We get saddled with all of the glow from the air in between us and that spot, and all of the glow behind it as well. This alone would make for a tough but not insoluble problem, but on top of that the atmosphere absorbs and scatters the airglow. Absorption and scattering basically couple all of the variables together making interpreting an airglow observation a bit like a game of pick-up sticks.

Scattering isn't all bad though: the airglow picks up information about the stuff it's scattering off of. Normally when photons are scattered multiple times you can't exactly ask them where the came from originally, so all you get is an image of the surface of last scattering. There are some special cases where you get a bit more though. Photons which are resonantly scattered around a particular wavelength will eventually doppler shift away to a wavelength where they are not scattered so heavily. This gives you a mix of photons in your detector which are scattered once, twice, three times, and so on all in reasonable proportions. The lines in the OII 83.4 nm triplet are perfect examples of this.

We can use Monte Carlo sampling to study how measurement uncertainties in the observed light translate into uncertainties in atmospheric parameters, In order to do that in a reasonable amount of time we need a good scattering model. For simple stuff you could use AURIC, but for MCMC you need to run the model thousands of times, preferably in parallel, and AURIC just isn't designed for that. I have written some radiative transport code in python for this purpose which is still in serious need of documentation over at github.com/georgegeddes/mcrt.

I have also just published a paper on this radiative transport code and some MCMC analysis it enabled. I couldn't have done it without copious amounts of help from all of my coauthors and many others. A preprint of tha paper is on arxiv.org and the peer-reviewed article is in JGR:Space Physics.