Detecting anthropogenic climate change with an optimal fingerprint method

We propose a general fingerprint strategy to detect anthropogenic climate change and present application to near surface temperature trends. An expected time-space-variable pattern of anthropogenic climate change (the 'signal') is identified through application of an appropriate optimally matched space-time filter (the 'fingerprint') to the observations. The signal and the fingerprint are represented in a space with sufficient observed and simulated data. The signal pattern is derived from a model-generated prediction of anthropogenic climate change. Application of the fingerprint filter to the data yields a scalar detection variable. The statistically optimal fingerprint is obtained by weighting the model-predicted pattern towards low-noise directions. A combination of model output and observations is used to estimate the noise characteristics of the detection variable, arising from the natural variability of climate in the absence of external forcing. We test then the null hypothesis that the observed climate change is part of natural climate variability. We conclude that a statistically significant externally induced warming has been observed, with the caveat of a possibly inadequate estimate of the internal climate variability. In order to attribute this warming uniquely to anthropogenic greenhouse gas forcing, more information on the climate's response to other forcing mechanisms (e.g. changes in solar radiation, volcanic or anthropogenic aerosols) and their interaction is needed. (orig./KW)