principle, a close match constitutes an independent verification of the MSU data. The two curves, shown in Figure 9.2, exhibit a number of common features and rather similar trends.
Uncertainties exist in assigning confidence levels to trends because of persistence in the data, which may or may not be due to the trend itself. There is no unique set of confidence intervals for the relatively short atmospheric temperature time series considered here. The estimated confidence intervals depend on the underlying statistical model that is used to describe the data, as well as on the exact period considered and the sampling interval (i.e., whether one uses monthly, seasonal, or annual means). One approach, following Cryer (1986), yields trends of 0.14 ±0.10 °C/decade for the subsampled MSU data and 0.04 ±0.07continue