Statistical Prediction of North American Air Temperatures from Pacific Predictors

Abstract

A statistical study suggests that sea surface temperatures (SST) anomalies in the Pacific can be used to forecast subsequent changes in surface air temperature anomaly over North America. The SST's generally produce higher hindcast skill than do forecasts made from sea level pressure (SLP) or from persistence. The skill associated with the hindcasts was generally low and dependent on both location and season. Over a large region of the central United States significant predictions could not be made for any season and combination of SST and SLP predictors. The results of the study were found to be insensitive to changes in model structure, treatment of the predictands and methods of skill scoring. Independent tests of the predictive relations gave results in excellent agreement with those discussed above from the dependent data set. The major predictive ability comes from water temperature and sea level pressure variations in the equatorial and tropical Pacific Ocean. Fluctuations of SST and SLP in the central North Pacific were of comparatively small value in predicting subsequent air temperature anomalies. The spatial distribution of predictors suggests the zonal SST gradients contribute most to the predictive skill while it is a coherent, near Pacific-wide fluctuation in the subtropical ridge that leads to the SLP skill. A physical hypothesis involving the interaction of the tropical ocean, subtropical and polar jet was constructed to explain these results.