Here we show that air-sea coupling significantly impacts the simulation of sea surface temperature (SST) variability in the California Current System (CCS). Previous work has shown important differences between coupled and uncoupled models in simulating coastal SST, but only using empirical coupling or idealized scenarios. We compare the output of the UCLA Mesoscale Coupled Model to the output of a similar but uncoupled model over the CCS. These high resolution, realistic regional models have identical coastlines, bathymetry, and topography. They are forced at the boundaries by reanalysis data over the historical period 1981-1990. This model setup allows us to evaluate how including air-sea coupling impacts the accuracy of our simulation by comparing our model output to buoy observations. Both the spatial patterns and the amount of variability are more realistic in the coupled model, which is likely due to improved simulation of internal variability.