Statistical Principles for Climate Change Studies

Abstract

Abstract Statistical principles underlying “fingerprint” methods for detecting a climate change signal above natural climate variations and attributing the potential signal to specific anthropogenic forcings are discussed. The climate change problem is introduced through an exposition of statistical issues in modeling the climate signal and natural climate variability. The fingerprint approach is shown to be analogous to optimal hypothesis testing procedures from the classical statistics literature. The statistical formulation of the fingerprint scheme suggests new insights into the implementation of the techniques for climate change studies. In particular, the statistical testing ideas are exploited to introduce alternative procedures within the fingerprint model for attribution of climate change and to shed light on practical issues in applying the fingerprint detection strategies.