This study evaluates the accuracy of predicted mesoscale land-sea circulations over the U.S. Western Coast from Regional Climate Models (RCM) simulated outputs. Most IPCC-CMIP3 AOGCM outputs are warm biased over the eastern Pacific ocean border. Particularly, offshore California the SST biases average 3-5°C. We have found that such incorrect representation of SSTs in AOGCMs, information used as bottom boundaries, affects RCM simulations in predicting coastal climate. We used a RCM based on WRF (with resolutions of 36 and 12 km) to examine the effect of coastal circulations to wind-driven upwelling along the California coast. The RCM was integrated for 10 years (1998-2007) using NCAR-CCSM3 atmospheric forcings with and without corrections of the SSTs biases. The bias-correction of NCAR-CCSM3 SSTs was performed by a quintile mapping approach using NCEP optimum interpolation (OI) SST as observed reference fields. By isolating the effect of SST biases, we showed that coastal mesoscale circulations are affected as well as the wind-driven upwelling regime with significant differences in the seasonal scale. In this context, the results revealed notable sensitivity of RCM downscaling approaches for assessment of climate change in coastal regions.