Correlation approaches to detection

Criticisms leveled at the use of pattern correlation methods in climate-change detection studies are addressed. We first explain how spatial autocorrelation effects are accounted for. We then show why it is important to consider both pattern similarity and spatial-mean results in detection studies, and evaluate a mean-square-error statistic in this context. We also show that the use of changes in climate, as opposed to absolute climate states, does not lead to spurious trends in the pattern correlation statistic R(t). In doing this, we explain how previous pattern correlation studies account for "time arrow" effects in considering changes relative to an arbitrary reference year. We conclude that criticisms of the pattern correlation method are without substance.