How does one build confidence in a system dynamics simulation like C-ROADS? Above you can see results from one important test that Tom Fiddaman of Ventana Systems did.
Here’s what he did:
From the published papers and online data, we made a graph of the “CO2 in the atmosphere” results from historical data and from forecasts of two other highly-disaggregated models — MAGICC, BERN, and ISAM — when fed by three widely-divergent greenhouse-gas emissions scenarios (higher emissions, A1FI, medium emissions, B1, and lower emissions, WRE). Those results are shown above in purple.
Then we fed the same three emissions scenarios (A1FI, B1, and WRE) into C-ROADS and observed the results for CO2 concentrations in the atmosphere. Does the carbon cycle behave similarly to the other simulations? How much did we sacrifice in accuracy when we simplified the C cycle in order to make the sim run fast and be transparent?
The answer is in blue. C-ROADS’ results are consistent with those of the highly disaggregated models.
Note: Socolow and Lam make a good argument for simulation approaches such as this (decision-maker-oriented sims) here: Socolow, R.H. and S.H. Lam. 2007. Good Enough Tools for Global Warming Policy Making Phil. Trans. R. Soc. A (2007) 365, 897–934.