Assessment of calibration assumptions under strong climate changes

Abstract

Climate model calibration relies on different working hypotheses. The simplest bias correction or delta change methods assume the invariance of bias under climate change. Recent works have questioned this hypothesis and proposed linear bias changes with respect to the forcing. However, when the system experiences larger forcings, these schemes could fail. Calibration assumptions are tested within a simplified framework in the context of an intermediate complexity model for which the reference (or "reality") differs from the model by a single parametric model error and climate change is emulated by largely different CO2 forcings. It appears that calibration does not add value since the variation of bias under climate change is nonmonotonous for almost all variables and large compared to the climate change and the bias, except for the global temperature and sea ice area. For precipitation, calibration provides added value both globally and regionally. The calibration methods used fail to correct climate variability.