Confusing signals in the climatic record

Five supposedly reliable data sets are used to investigate the linear trend in screen-level air temperature in Arizona over the past 50 years. The state's single national benchmark station produces a half-century cooling trend of 1.36°C, but its yearly temperature departures are poorly correlated with those of the other four data sets. Although striking in their apparent similarity, two of these latter data sets produce cooling trends while two produce warming trends; and, although not significantly different from zero, the most extreme of the cooling and warming trends differ from each other by an amount which is comparable to the apparent global warming of the last 100 years. These disparate results suggest that even the assessment of recent historic temperature trends, as contrasted with the more speculative prediction of future trends, can be a most difficult and demanding task.