One facet of the RAGNARoCC project involves using computer simulations of the Earth climate system to develop our understanding of an incompletely observed system. These numerical models help us to learn about how systems such as the oceans and atmosphere might evolve in the future as a result of increasing carbon emissions. Beyond this, they can also teach us how the parts of these systems that are difficult to observe are working. Take, for example, the ocean carbon system. We know that human activities have increased the atmospheric concentrations of carbon dioxide since the onset of the industrial era, and that about a quarter of these emissions are taken up by the global ocean. But how does this rather important ‘ocean sink’ of CO2 work?

Work that has been carried out over the past few decades has allowed us to develop quite a rich understanding of the role of the oceans in removing CO2 from the atmosphere. There is now a very widespread observing network that collects measurements of CO2 in the oceans (as partial pressures of carbon dioxide gas in sea water: pCO2). This enormous effort has taught us that some parts of the ocean are net sink regions of CO2 for the atmosphere (such as the high-latitude North Atlantic, the mid latitude Pacific and Indian oceans) and others are net sources (such as the equatorial oceans). However, there are still large parts of the global ocean that are undersampled, and the time history of these observations for most parts of the world is quite limited. As a result, questions remain as to what causes the ocean carbon sink to vary in time. This is where numerical models can be valuable tools. My own research (as a part of the RAGNARoCC Project) involves using numerical models to study what causes the flux of carbon into the ocean to vary in time. But instead of going into that here, I wanted to show some output from a model that I work with that I think is interesting to look at.


(If the animation does not appear in your browser, you can watch a GIF version.)

One might be led to wonder what the invasion of anthropogenic carbon dioxide into the oceans looks like, and a numerical model can help to show us this in ways that observations cannot. The animation shows output from the NEMO-MEDUSA model, and describes, to a first order, the distribution of anthropogenic dissolved inorganic carbon (DIC) in the Atlantic Ocean. The model is run at a horizontal resolution of about one degree of longitude by one degree of latitude, and experiences strongly rising atmospheric CO2 concentrations throughout (as per IPCC RCP 8.5). The ocean basin concentrations are presented as a zonal mean from South to North, and each frame shows an annual average concentration at each point. This is a way of viewing carbon concentrations across the length and depth of the Atlantic by taking into account its entire west-east extent. The basin is broken into two subframes, with the upper panel showing the surface 1000m in detail, and the lower panel depicting the deep ocean (1000 to 6000m). The white at the bottom of the panels shows the seafloor. The animation covers a time period of the years 1980 to 2099.

The cool thing about this particular type of animation (an HTML-5 video, or ‘gfycat’ if your browser permits) is that users can control how it runs, unlike .gif animations. Start by right clicking on the plot and clicking “Show Controls”. If the controls fade out, left click on the animation to make them reappear again. Using the control buttons, you can speed up or slow down its progress (with the + and -), pause it on a single frame, or even run it in reverse. I present animated data in this format because it allows users to inspect it however they choose, and can focus on details that interest them. Another handy feature is that these gfycats is that they’re a much more compact file than a .gif animation (1Mb, compared to an equivalent 3Mb in .gif format). Now that we’re familiar with the animation, what is it actually showing us?

Since anthropogenic DIC enters the oceans via the atmosphere, we naturally see the highest concentrations in the surface ocean. Large parts of the deep ocean have very little, if any human-emitted carbon at all, especially early in the animation. What we see is that the only pathway for anthropogenic carbon into the deep ocean (greater than 1000m) is via the mid and high latitudes. We can see high concentrations of anthropogenic DIC entering the deep ocean in the North Atlantic with the formation of North Atlantic Deep Water. This is a water mass formed in when surface waters are warmed in the equatorial Atlantic, travel northward in the Gulf Stream and North Atlantic Current to the Labrador and Norwegian seas, where it loses heat (and therefore buoyancy) to the cold overlying atmosphere. As this water cools, it also takes up anthropogenic CO2 from the air, and eventually loses enough buoyancy that it sinks down into the ocean interior and flows southward. We can observe this southward flow as the deep-reaching blob of high DIC concentrations in the northern (right hand) part of the figure. It is this mechanism which makes the North Atlantic ocean a very efficient pump at removing CO2 from the atmosphere. We can also see other regions where CO2 deeply invades the Atlantic in the mid and low-latitude southern hemisphere (between 40 and 50°S, and south of 60°S).

It is worth noting that the DIC that humans have added to the atmosphere is not the only form seen in the oceans. The global ocean also hosts a comparatively enormous stock of ‘natural’ DIC that recent human activities are adding to. It is very difficult to quantify anthropogenic carbon in the oceans, since it is not readily distinguishable from the large natural pool. Methods do exist to estimate the proportion of DIC that is ‘anthropogenic’, but these are not direct. In fact, what is plotted in the animation isn’t quite ‘anthropogenic’ carbon in a model run. Actually, the values shown are the difference between two runs of an ocean model: one with climate change corresponding to the IPCC RCP 8.5 emissions scenario and another without rising atmospheric CO2 levels. The idea is that the difference between these two runs is technically “where DIC is higher in a run with rising atmospheric CO2 than in another run where CO2 is held fixed”.

However, if we compare a simulation with rising CO2 against one without, then the circulation patterns of the two runs will differ, because one run is experiencing the global warming effect of CO2 and the other isn’t. One of the neat features of models is that we can run them however we like. We could, for example, allow the radiative heating effect that would result from rising atmospheric CO2 levels, but prevent the atmospheric concentration of the gas from actually rising. This then gives us a simulation with the same circulation as the ‘real world’ run (which has a warming climate resulting from rising CO2), whose only pool of carbon is the natural one. If we subtract the ocean DIC concentrations of this experiment from the ‘real world’ run, we end up with this (quite roundabout) way of estimating anthropogenic DIC.