It is up to the discretion of the modeler as to how to initialize their simulations, but the following criteria must be upheld:
As for the duration of simulations, the biotic equilibrium simulation should be continued at least until the globally integrated air-sea CO2 flux is less than 0.01 Pg C/yr. For most models, this criterion can be reached only after an integration of at least a few thousand model years.
To approach equilibrium more rapidly, some modelers use the acceleration technique where the timestep increases in deeper layers (Bryan, 1984). When modelers use this technique, one complication is that the global inventory of certain biogeochemical tracers (those without external sources or sinks) cannot be conserved. As a fix, those at NCAR suggest that modelers who use such a technique should periodically adjust global tracer inventories to their initial values. In any case, some method must be used to avoid losing mass from the system. Furthermore, we strongly recommend that the deep acceleration method be switched off and that the model be run for at least an additional 500 years, before final "equilibrium" output is stored for later OCMIP-2 analysis.