Climate drift in the coupled land-atmosphere system

Climate drift in coupled ocean-atmosphere models has been recognized for many years, and many methods have been developed to address the problem. Climate drift also exists in coupled land-atmosphere models. However it is more difficult to detect and measure since important land surface state variables such as snow water equivalent, soil wetness and heat content are poorly observed, in contrast to sea surface temperature, the principal state variable at the ocean surface that affects surface fluxes. Nonetheless, drift can be quantified in the absence of reliable observations by measuring the divergence in land surface state variables between an integration of the fully coupled land-atmosphere model, and an integration where one arm of the feedback is suppressed. Similarly, the analysis "increment" in a land data assimilation system is a measure of drift, particularly in its initial stages. Drift may arise from a systematic error in one or both component models, or from an incorrect coupling (e.g., too weak or too strong) which might affect the strength of land-atmosphere feedbacks and the sensitivity of one component model to fluctuations in the other. We have attempted to establish the causes of climate drift in the coupled COLA climate model. Particular attention is paid to systematic errors in precipitation in the atmospheric model. The utility of applying some form of flux adjustment at the land-atmosphere interface will be discussed.