Carbon model pros

As with any model, simplification has both advantages and disadvantages. This article describes some of the carbon model pros.

Types of models

In computer modelling there is a continuum between two caricature classes of models:

  1. conceptual-type models that aim to simplify a system down to include only its most important elements, in particular those that govern its behaviour
  2. simulation-type models that are highly detailed and try whenever possible to avoid making assumptions about what is important; instead they try to include as much detail as possible in the expectation that nothing important will be omitted

By simplifying to only the most important fluxes and state variables, this model is very much a conceptual-type model.


One of the key advantages of such an approach is fast run times. The most detailed (fine spatial resolution) simulation models can, at the extreme, take many months of real time to simulate only a few decades. In contrast, this conceptual model can simulate hundreds of thousands of years in only a few seconds.

Leading on from this, a second advantage is the ease of experimentation and sensitivity with the quicker models such as this one. If each model run takes many months then the ability to carry out many different runs in series is limited. This is particularly important where uncertainty in the model formulation favours large numbers of sensitivity runs.

Another advantage of simpler models is the ability to understand why the model produces a particular behaviour. This is much harder, and sometimes impossible, with very complex models. It is not always possible to disentangle the large network of interacting effects in a complex model in order to work out which one is key, whereas it is easier in a simple model.

The carbon model is fairly sophisticated in terms of the number of different processes that it contains. Although physically naïve (e.g. no horizontal resolution), it is biogeochemically more advanced.

The complexities of the chemistry of dissolved inorganic carbon in seawater are calculated using a large set of equations that have been incorporated into the computer routines written by Richard Zeebe and Wolf-Gladrow.. This allows the whole of the carbonate chemistry to be calculated at each timestep from DIC and alkalinity. The values of, for instance, CO2, carbonate ion and pH are not tracked in the model. There are no state variables for any of these three. Only DIC and alkalinity are tracked, and all others are continuously updated from the current values of DIC and alkalinity, using the carbonate chemistry routines.

The main processes of the ocean’s carbon cycle are included. The model also represents the amount of carbon held in the atmosphere (as CO2) and simulates the net transfer of CO2 across the sea-surface. The program thereby allows the ocean and atmosphere carbon reservoirs to be dynamically inter-linked.

The process of carbonate compensation is properly included. Because the model incorporates carbon flowing down rivers into the sea, carbon burial at the bottom of the sea, and the dependence of carbon burial on seawater chemistry, it can represent the negative feedback mechanism of carbonate compensation. Because the model contains detailed knowledge about how much of the seafloor resides at each depth in the ocean (the ocean hypsometry), it can accurately calculate the relationship between changes in depth of the CCD and changes in global CaCO3 burial.

The model has been tuned in order to obtain balanced cycles of DIC, alkalinity, phosphate and δ 13C. This is not as easy as it might appear. When a carbon model is first set up, it is overwhelmingly likely to ‘drift’ over time. That is to say, it is likely that the values of the state variables will undergo some inexorable directional change. This makes it more difficult to work out what is happening during a perturbed model run, if the effect always has to be somehow subtracted from the drift. This model will, by contrast, remain happily in steady-state for hundreds of thousands or millions of years if not subject to perturbations. The amount of drift is very small.

In order to be of most use in assessing how the ocean has functioned in the past, the model should produce outputs that can easily and powerfully be compared to the data that palaeoceanographers and palaeoclimatologists collect. One of the most commonly collected palaeorecords is the ratio of the two main isotopes of carbon, i.e. δ 13C. The model does not just simulate the cycling of carbon as an element, but rather it simulates separately the cycling of two different isotopes of carbon: carbon-12 and carbon-13. This means that the model calculates the δ 13C of buried carbon over time (POC, bulk CaCO3 and foram shells), and these can be directly compared to the measurements made on ocean sediment cores. In this way a rapid assessment can be made of whether model results agree with data or not.

The model has been set up to allow the user to add fossil fuel emissions of CO2. This can be done using either a simplified sine wave, or otherwise from historical emissions data and from an SRES scenario. This allows preliminary investigations of how anthropogenic CO2 emissions are influencing and will influence ocean chemistry. The IPCC SRES scenarios are accessible through only a few mouse clicks.

Some models are "closed-system" in the sense of forcing all substances to remain within the model ocean. Sinking material that reaches the seafloor, for instance, will be remineralised at the bottom (or top) of the ocean, and rivers will not be modelled. The boundaries of these models are absolute. This model is more realistic in the sense of being open-boundary (open-system). That is to say, it doesn't just simulate internal fluxes (moving C and P around the ocean) but also the fluxes into and out of the ocean as a whole.

Mass-balance checks are included in the model (behind the scenes) to make sure that the model equations do not lead to any improper loss or gain of tracers from the system as a whole.