Origins and Levels of Monthly and Seasonal Forecast Skill for United States Surface Air Temperatures Determined by Canonical Correlation Analysis

Statistical techniques have been used to study the ability of SLP, SST and a form of persistence to forecast cold/warm season air temperatures over the United States and to determine the space–time evolution of these fields that give rise to forecast skill. It was found that virtually all forecast skill was due to three climatological features: a decadal scale change in Northern Hemisphere temperature, ENSO-related phenomena, and the occurrence of two distinct short-lived, but large-scale, coherent structures in the atmospheric field of the Northern Hemisphere. The physical mechanisms responsible for the first two signals are currently unknown. One of the large-scale, coherent features seems largely independent of the ENSO phenomena, while the second is at least partially related to ENSO and may be part of a recently discovered global mode of SLP variation. Both features resemble various combinations of known teleconnection patterns. These large-scale coherent structures are essentially stationary patterns of SLP variation that grow in place over two to three months. The structures decay more rapidly, typically in 1 month, leading to a highly asymmetric temporal life cycle. The average forecast skills found in this study are generally low, except in January and February, and are always much lower than expected from studies of potential predictability. Increase in the average skills will require new information uncorrelated with any of the data used in this study and/or prediction schemes that are highly nonlinear. However, the concept of an average skill may be misleading. A forecast quality index is developed and it is shown that one can say in advance that some years will be highly predictable and others not. Use of the classical definition of “winter” in forecast work may not be advisable since each of the months that make up winter are largely uncorrelated and predicted by different atmospheric features.