The continuing search for an anthropogenic climate change signal: Limitations of correlation-based approaches

Abstract

Several recent studies claim to have found evidence of large-scale climate changes that were attributed to human influences. These assertions are based on increases in correlation over time between general circulation model prognostications and observations as derived from a centred pattern correlation statistic. We argue that the results of such studies are inappropriate because of limitations and biases in these statistics which leads us to conclude that the results of many studies employing these statistics may be erroneous and, in fact, show little evidence of a human fingerprint in the observed records.