Estimating Variability of Surface Air Temperature in the Northern Hemisphere

Abstract

The space/time scales of surface temperature anomalies over much of the Northern Hemisphere have been analyzed for winter and annual average data. The variability over land is 2–6 times larger than over the oceans. This fact and the latitudinal dependence of the land/ocean area ratio can explain many results obtained by other authors. Regions of high variability were found to be “local” in both space and time, thus suggesting that a rather dense spatial network of sampling points is required to adequately estimate changes in hemispheric temperature. The “local” nature of the anomaly field is related to the meridional advection of heat by long waves, a result suggested by van Loon and Williams (1976). If the results of this study hold to first order for the Southern Hemisphere, then it may be concluded that reliable estimates of temperature change in that hemisphere, and hence also for the entire planet, have yet to be made. The covariance matrix of annual average data had an eigenvector whose components had virtually the same sign over the entire hemisphere. This implies there is a pattern of temperature change that is coherent over nearly the entire hemisphere. The principal variance in this pattern occurred in the oceans and not over land masses. The variance in the wintertime hemispheric temperature field was essentially constant between 1950–62. Since 1963 the hemispheric variance has experienced large oscillations with an apparent three-year cycle. This remarkable change in variability began before the eruption of Mt. Agung.