Tag – you’re it! Chasing atmospheric tracers

Beijing pollution

Beijing, a megacity of ~20 million inhabitants, when I was there last September. How to keep track of the pollution being released?

This is a blog post I originally wrote for GeoLog (the most excellent official blog of the EGU), which appeared on 22 March 2013 (in a slightly edited form). 

I’ve been ruminating over the idea for this post for some time now; since last October in fact, when the EGU Twitter Journal Club discussed a paper about tagging (You can find the Storify for the discussion here). Not tagging as in the playground favourite, but the idea of keeping track of certain molecules in your chemical transport model, so you can follow them as they move through the atmosphere and undergo chemical transformations.

After deliberating, cogitating and digesting, I’ve decided to offer my opinion on tagging, which I’ve come to through reading, listening and discussing the topic with colleagues. I thought this might be of general interest, as the concept of tagging frequently sparks debates (in my experience) and seems to arouse stronger and more varied views than I would usually expect from a modelling technique. So with this post, I want to answer the question “what is tagging and what is it used for?” and at least attempt to answer “why does it generate such mixed feelings?”.

I think that the answer to the second may lie in the answer to the first, so let’s start with that. One way of describing tagging is as an accounting method. Doesn’t sound very geophysical? Well, we’re talking the model world here, and we can keep track of – or account for – every little thing we do in our model world. To try and understand the composition of the Earth’s atmosphere, people have constructed computer models to describe both the physical and chemical processes in the atmosphere. I’m particularly interested in trace gases like ozone, which is found in parts-per-billion quantities in the troposphere (roughly the lowest 10-15 km of the atmosphere). Tropospheric ozone is a popular species to study, as it is a greenhouse gas, it’s bad for human and plant health, and it’s one of the key oxidants in the atmosphere.

Another interesting thing about ozone is that it’s not emitted directly. Many other pollutants and greenhouse gases, like methane or nitrogen oxides, are emitted by both natural and anthropogenic sources, but ozone is formed through photochemical reactions, which occur in the presence of both sunlight and nitrogen dioxide. So, to study the ozone in the atmosphere using a model, it needs to include emissions of relevant gases, motions in the atmosphere that move these gases around, chemical reactions that transform one gas into another, as well as other processes like deposition of certain gases to surfaces or removal in rain.

With all this going on, it can be hard to disentangle one process from another, particularly since all the different processes are interlinked. This is where tagging can come in handy as an accounting tool. Say we are interested in how much ozone is added to the atmosphere as a result of activity in a particular city – let’s call it Mega-City One. How could you find this out? Well, one quite common way is to run the model as normal, and then to repeat the simulation but with Mega-City One removed. So you take out any emissions coming from Mega-City One and see what difference it makes. This tells you what would happen if you simply removed that city.

However, you may not want to know about this unrealistic “annihilation” situation, as a city won’t simply disappear overnight. You may be more interested in your normal run (which you think best represents the real world) and what is contributing to the ozone in that run. The non-linearly minded amongst you will see the subtle difference between the two questions. If we were talking about an inert tracer, there would be no difference, as what was emitted by Mega-City One would just stay in the atmosphere unchanged, and would be directly attributable to Mega-City One. The difference with ozone is that it’s not emitted, and the chemistry that creates and destroys it can and does behave nonlinearly.

So to answer the attribution question (in my model simulation, how much ozone is a result of Mega-City One emissions?), tagging is one method that is used. The idea is that you ‘tag’ tracers that you are interested in, and follow them through the model. So you can ‘tag’ all the emissions from Mega-City One, and when one of the nitrogen dioxide molecules from this city undergoes photolysis to make an oxygen atom, which then reacts with molecular oxygen (O2) to form ozone (O3), you know that this ozone molecule can be attributed to Mega-City One. This is done by solving a set of equations for the tags alongside the usual chemistry scheme. This does not disturb the normal running of the model, and you will find that the total amount of ozone attributed to Mega-City One will be different from calculating the difference in ozone between your normal and your annihilation runs.

Why would you want to do either of these things anyway? Well, you might want to know what effect different cities or other sources are actually having, or would have on the atmosphere if they were built. So you’d use the annihilation method. Or, you might want to know what the biggest culprit was for particularly bad air quality in a particular place. So you’d use tagging to find out if it was Mega-City One or Mega-City Two, or if it was the road transport or the power stations that was the biggest source.

I hope I’ve achieved my first aim of answering my first question, about what tagging is and what it’s used for. The second question is not quite so straightforward, but I’ll give my thoughts on the matter: I think the key to why – to put in bluntly – some people aren’t so keen on tagging, and some people aren’t so keen on the annihilation method, stems from different notions of what tagging is really being used for. Ultimately, I think this boils down to finding ways of communicating complex ideas and their applications, without being misconstrued. Existing pre-conceptions probably also play a part, as these will (rightly or wrongly) fill the gaps when someone else’s explanation is incomplete.

Successful communication of complex ideas isn’t easy! It takes time and effort, but it really pays off in the end. Just think, how often have you heard (or taken part in) a heated discussion in which you find everyone was actually in agreement all along, they just didn’t know it yet?

Paper that initially inspired the discussion:

Grewe, V, Dahlmann, K, Matthes, S and Steinbrecht, W. (2012) Attributing ozone to NOx emissions: Implications for climate mitigation measures. Atmospheric Environment. Vol. 59, pp 102-107.

Photo diary from flights around the Arctic

This gallery contains 10 photos.

Here are a few snaps from the MAMM field campaign (July 2012). We went flying on the FAAM research aircraft, kitted out to measure many gases, aerosols, and other meteorological parameters. The main aim was to search out Arctic sources … Continue reading

What do I do all day?

what i've been doing all day

Science in action! Well, maybe "in progress" would be more accurate...

I ask myself this question frequently, usually when I review my to-do list, and wonder how I’ve managed to take so much longer to complete the tasks than I originally planned. Today, for example, I spent the whole day working on a program that will automatically plot some nice histograms of my data. I hadn’t intended to spend the whole day on doing just this, but once I started, it just seemed to eat up the whole day. Worse, the program still isn’t finished! Maybe it’s a problem peculiar to IDL (it is a rather long winded programming language, but you can make pretty plots with it). But I doubt it. IDL just makes it even easier for a small task to take up your whole life. But just think of the pretty pictures it makes!

In fact, the pretty picture bit wasn’t taking the time. It’s the handling of the data, and making the code flexible so that I can feed it a whole bunch of different data, and ask it to sort it in different ways, and to plot it sensibly, all at the touch of a button. So I’m just spending my time up front, rather than later down the line.

So getting back to the question, “what do I do all day?”. Far from being a flippant remark, to anyone who is not me (or a post-doctoral researcher who works in a very similar area of work to me) it is a valid question. You might be curious about what do scientists do all day. You might even ask what scientists spend your taxes on. I can only speak for myself, but I thought I’d give it a go, seeing as I’ve had enough of IDL for one day. As my brain in almost a mush at this point in the day, I thought I’d go through the most mundane of tasks.

I run computer models. I edit the code of said models, to do little experiments and tests (this takes longer than it sounds). I find bugs in the code. I plot up the data from the models, often in comparison to observed data. Sometimes, I will process the data (eg using statistics to understand the data) before plotting the processed data. Hmm, that’s essentially it, I think. And I think that’s actually broadly similar to many other scientists in other disciplines, it’s just the specific methods that differ.

Why do I do all this sitting in front of a computer, writing code when I’m not actually a computer programmer? That’s a very good question, and one which is sometimes hard to answer when you’ve just spent hours, days or even weeks trying to get your code to work. It’s important to remember that these programs and models are just tools to use to try and understand the world around us, and to help answer scientific questions.

But what are the questions, you ask? Well, I think that’s a question for another blog post.