Making it all work together

Alok Gupta

Alok Kumar Gupta, researcher at Uni Climate and the Bjerknes Centre for Climate Research

Water evaporates from the ocean and forms clouds and rain. Rain that falls over the ocean influences the salinity of the water. Snow falling over glaciers is stored as ice. Carbon dioxide from the atmosphere is absorbed in the ocean and released at a later time. There is a continuous exchange between the different components of the Earth and its atmosphere, and an Earth system model must also take care of this.

The perfect climate model would be able to take every process in the Earth’s climate and represent them – and their relationships to one another – in a single huge mathematical equation. In reality, this would be too complicated to do, so climate modelers work sequentially.

Alok Kumar Gupta is in charge of running the NorESM. As other Earth system models, the NorESM consists of separate modules for the atmosphere, the ocean, land and ice. To make a full model, the components must be coupled together. This is done by the coupler – a separate program that organizes the data exchange between the atmosphere, ocean, land and ice modules.


The Coupler isn’t a real, physical thing, but rather a load of code that allows information to be exchanged between the model components. The code acts almost like a translator – letting each of the components communicate with one another.

It has some pretty important jobs to do:

Transforms and combines information – it takes simple variables from one component and performs calculations to work out things like surface fluxes – which can then be used for other components.

Deals with grids – the grid systems can vary for each component. It’s the coupler’s job to interpolate between each of them so that everything lines up.

Deals with information in time – for example, the ocean component of NorESM operates with daily averages, but most of the other components exchange information roughly every half hour


Climate models are set up into user-defined time steps, moving forward from an initial state (e.g. observed data which have been interpolated to match the spatial grid in the model). From each time step to the next, all equations must be solved.



This diagram summarises the main information that is exchanged between each of the components and the coupler. More details can be found here.





Have a think about how much information is being analysed, calculated, transformed and sent on – every half hour or so. That’s a lot of information passing through the coupler.

Concept of network connections. 3D render image.
Running an Earth System Model takes a LOT of computing power. Photo:



Computing – As you can imagine, all these calculations need some pretty special computers. The most advanced supercomputers on Earth are used to run Earth system models. These have hundreds or even thousands of processors,and special programming and computing techneques are needed so that the work can be spread out across them and they can work in parallel. The different jobs are run separately and then stitched together at the end into one global picture – known as massively parallel computing with distributed memory.

Storage – This information also has to be stored, so the coupler and or the scientists can access and analyse it.


The Stallo Supercomputer. Credit: Thilo Bubek, UNINETT Sigma2.


The Hexagon Supercomputer.Credit: Alexander Oltu, UNINETT Sigma2.