Overcoming Chaotic Behavior of General Circulation Models (Gcms)

General Circulation Climate Models are based on non-linear partial differential equations and therefore exhibit chaotic characteristics. Each run leads to a different temperature trend value, depending on the exact initial conditions. Modelers therefore constrain the trend value by producing an “ensemble-mean” of several runs. We show here that at least 16 runs of 25-year length are necessary to achieve convergence – and at least 4 runs of 100-year length. Contrary to expectations, we find that these numbers do not depend on the magnitude of radiative forcing of the GCM.