The Treacle Ocean
When you think of ocean oceanography you probably think of people diving with corals or boats deploying brightly coloured equipment. I myself am a physical oceanographer and personally don’t get my feet wet very often. So what do I do? Well, I model the ocean using a computer… a big one!
The ocean is very complex and understanding the whole ocean to predict how it will respond to changing climate is very difficult. Yet this is what we must do in order to forecast hurricanes, monsoons, ice drift , sea level rise, carbon fluxes and our future climate. This is an enormous task.
We use general circulation models (GCM’s) that have to be run on a supercomputer. They work by splitting the ocean up into millions of little boxes and for each box, the computer solves some fundamental physics equations to decide what the box output into the next box. A small time step is used and the calculations repeated over and over. The smaller the boxes are the more detail can be worked out, however the smaller the boxes ,the more boxes you need. That means more calculations and more computing power is required. For example, the 1/12 degree NEMO model (an ocean GCM that has boxes that 1/12 degree of latitude and longitude by 10-100 m thick) is run on the UK supercomputer Archer using 100000s of high-performance computer cores. To put that in perspective an average laptop has 2 – 4 average performance computer cores. These models have up to half a million lines of code written by teams of scientists. Here’s an example of what the end result is for that NEMO 1/12 degree run:
These ocean models are used in a number of ways. Sometimes they are combined with atmospheric models and fed with real data to give short-term weather forecasts. Other times they are combined with atmospheric, bio-geochemical and land models and run for long periods of time with changing CO2 concentrations. These models aren’t perfect, though, we can’t just keep making the boxes smaller and smaller otherwise, we’d need to make the computers bigger and bigger. For things like climate prediction 100s of models from all over the world are used to give the probability of different outcomes.
So what about the missing detail in those not small enough boxes. By focusing on one aspect of a problem on a smaller scale, we can begin to work out the large-scale effects of some small scale processes. People like myself run small models on just 10-100s of cores that try to work out how that part of the ocean works and how it would respond to various changes and what effect on the large-scale ocean. This is called a parameterization.
One last thing. Why I titled this, the treacle ocean? That’s because in order to model the ocean we must increase its viscosity to that of treacle so when you think of ocean models. They are actually treacle oceans!