Quantile spline models for global temperature change

The Hansen and Lebedeff data set on global surface air temperature change is reanalyzed using smoothing splines designed to estimate the conditional quantile functions of global temperature over the last century. It is assumed only that the quantiles are smooth functions of time. The smoothness of the fitted quantile functions is determined by a data driven version of the Schwarz criterion. The estimates offer statistical evidence of a break in the generally upward sloping trend of the temperature series during the period from 1940 to 1965, a finding originally suggested by Hansen and Lebedeff (1987).