Spatial structure of the evolution of surface temperature (1951-2004)

摘要

The spatial structure of the evolution of the surface temperature in the GISS and HadCRU datasets are analyzed and compared, using a technique that computes the best fit to a time series by a set of continuous linear segments. The technique optimizes, in the least-squares sense, both the location of the trend changing points and the trends in each segment. When applied to the mean temperature maps, the method produces a detailed description of recent changes in the world climate, emphasizing the anomalous behavior of some important areas, namely the North Atlantic and nearby regions. A link between that behavior and the evolution of the North Atlantic Oscillation is suggested. The spatial structure of cooling and warming, and of the temporal location of trend changing points, are very similar in the two datasets.