Evaluation of ocean and climate models using present‐day observations and forcing

Abstract

The most common method used to evaluate climate models involves spinning them up under perpetual present‐day forcing and comparing the model results with present‐day observations. This approach clearly ignores any potential long‐term memory of the model ocean to past climatic conditions. Here we examine the validity of this approach through the 6000‐year integration of a coupled atmosphere–ocean–sea‐ice model. The coupled model is initially spun‐up with atmospheric CO₂ concentrations and orbital parameters applicable for 6KBP. The model is then integrated forward in time to 2100. Results from this transient coupled model simulation are compared with the results from two additional simulations, in which the model is spun up with perpetual 1850 (preindustrial) and 1998 (present‐day) atmospheric CO₂ concentrations and orbital parameters. This comparison leads to substantial differences between the equilibrium climatologies and the transient simulation, even at 1850 (in weakly ventilated regions), prior to any significant changes in atmospheric CO₂. When compared to the present‐day equilibrium climatology, differences are very large: the global mean surface air and sea surface temperatures are ,0.5°C and ,0.4°C colder, respectively, deep ocean temperatures are substantially cooler, Southern Hemisphere sea‐ice cover is 38% larger, and the North Atlantic conveyor 16% weaker in the transient case. These differences are due to the long timescale memory of the deep ocean to climatic conditions which prevailed throughout the late Holocene, as well as to its large thermal inertia. It is also demonstrated that a ‘cold start’ global warming simulation (one that starts from a 1998 equilibrium climatology) underestimates the global temperature increase at 2100 by ,10%. Our results question the accuracy of current techniques for climate model evaluation and underline the importance of using paleoclimatic simulations in parallel with present‐day simulations in this evaluation process.