Testing the Statistical Significance of the Response of the NCAR General Circulation Model to North Pacific Ocean Surface Temperature Anomalies

The response of the six-layer NCAR atmospheric general circulation model (GCM) to mid-latitude North Pacific Ocean surface temperature anomalies is examined. In particular, January and winter season averages of the 1.5 km temperature fields generated by the GCM with prescribed ocean surface temperatures (i.e., the control experiment) are subtracted from January and winter season averages of three separate prescribed change experiments in which ocean surface temperature anomalies are superimposed on the ocean surface temperature patterns of the control case. These differences between prescribed change and control experiments are compared at each grid point to the standard deviations of the 1.5 km temperature field constructed from five independent GCM generated January simulations. The statistical significance of the results is evaluated using a methodology developed for the GCM by Chervin and Schneider based on the classical Student's t-test. The results suggest that a statistically significant signal can be detected in the immediate vicinity of large ocean surface temperature anomalies, but that teleconnections of this signal downstream over the United States can be identified only with a small statistical confidence. Some implications of these results and suggestions for further work are presented.