INTERCOMPARISON OF REGIONAL BIASES AND DOUBLED CO2-SENSITIVITY OF COUPLED ATMOSPHERE-OCEAN GENERAL-CIRCULATION MODEL EXPERIMENTS

We compared regional biases and transient doubled CO2 sensitivities of nine coupled atmosphere-ocean general circulation models (GCMs) from six international climate modeling groups. We evaluated biases and res ponses in winter and summer surface air temperatures and precipitation for seven subcontinental regions, including those in the 1990 Intergo vernmental Panel on Climate Change (IPCC) Scientific Assessment. Regio nal biases were large and exceeded the variance among four climatologi cal datasets, indicating that model biases were not primarily due to u ncertainty in observations. Model responses to altered greenhouse forc ing were substantial (average temperature change = 2.7 +/- 0.9 degrees C, range of precipitation change = - 35 to + 120% of control). While coupled models include more climate system feedbacks than earlier GCMs implemented with mixed-layer ocean models, inclusion of a dynamic oce an alone did not improve simulation of long-term mean climatology nor increase convergence among model responses to altered greenhouse gas f orcing. On the other hand, features of some of the coupled models incl uding flux adjustment (which may have simply masked simulation errors) , high horizontal resolution, and estimation of screen height temperat ure contributed to improved simulation of long-term surface climate. T he large range of model responses was partly accounted for by inconsis tencies in forcing scenarios and transient-simulation averaging period s. Nonetheless, the models generally had greater agreement in their se nsitivities than their controls did with observations. This suggests t hat consistent, large-scale response features from an ensemble of mode l sensitivity experiments may not depend on details of their represent ation of present-day climate.