Much effort has gone into evaluating the skill of General Circulation Models (GCMs) for ‘standard’ climate variables such as surface (air and/or sea) temperature, or precipitation. Whether climate model skill to simulate standard variables translates to the performance of dynamical GCM forced wind–wave simulations is yet to be established.

We assess an ensemble of historical dynamical wave climate simulations whereby surface winds taken from GCMs participating in the Coupled Model Intercomparison Project (CMIP) are used to force a spectral wave model. The GCMs used include 8 CMIP5 models and two dynamically downscaled CMIP3 models.

The climatological properties of key integrated wave parameters (significant wave height, maximum wave height, mean wave period and direction) are evaluated, using two independent methods, relative to three historical wave hindcast/reanalysis datasets over 13 areas of the global ocean.

We identify that high performance of GCMs for 'standard' climate variables does not imply high performance for GCM forced wave simulations. We also identify there is little to no benefit in choosing a higher resolution CMIP5 GCM (with resolution of ∼1.4°) over a lower resolution GCM (∼2.8°) to improve skill of GCM forced dynamical wave simulations. With the conscious push towards developing projections of waves and storm surges to aid assessments of possible climate driven impacts to coastal communities, we stress the need to evaluate the performance of a GCM for the marine meteorological climate independently of the performance of the GCM for the ‘standard' climate variables.