C-ROADS Carbon Cycle model

This is a reimplementation of the carbon cycle components of Climate Interactive's C-ROADS model for exploring possible futures with man-made climate change. The original model includes data about the various economic blocs involved in climate negotiations, and simulations predicting their likely growth in population and GDP, allowing the effects of implementing various absolute or relative restrictions on greenhouse gas emissions to be estimated. Here we reproduce only the biophysical components of the model, namely the carbon dioxide, methane and temperature change components. The parameterization file attached here includes time series data for CO2 and CH4 emissions generated by the full C-ROADS model for the baseline "business-as-usual" scenario. The model can be run with different values for various aspects of climate sensitivity to anthropogenic GHG emissions to produce a range of results.

This Simile model makes use of units in math to provide consistency checking of units throughout, as well as automatic numerical conversions where required. For instance, in the carbon dioxide component of the original C-ROADS model, there are various explicit conversion factors such as "ppm CO2 per GtonC" which are used in the equations. In the Simile version however, only natural constants such as the volume of the atmosphere and the molecular weights of the various species need be included -- all the rest, including things like conversion between ppm and ppb, are provided by the modelling tool.

The Simile model also includes a multi-instance submodel to represent ocean layers. In the original, these are represented by a collection of array variables, using special subscripts to access values from adjacent layers. In the Simile version, there are no explicit arrays, but the whole submodel is duplicated implicitly to stand for identical processes in multiple entities. The in_preceding() function is used to allow a calculation in each instance to access a value from the preceding (higher layer) instance. This avoids the need for a separate association submodel, which would allow values to be transferred between arbitrary pairs of instances.