Panel and Multivariate Methods for Tests of Trend Equivalence in Climate Data Series

We explain panel and multivariate regressions for comparing trends in climate data sets. They impose minimal restrictions on the covariance matrix and can embed multiple linear comparisons, which is a convenience in applied work. We present applications comparing post-1979 modeled and observed temperature trends in the tropical lowerand midtroposphere. Results are sensitive to the sample length. In data spanning 1979 to 1999, observed trends are not significantly different from zero or from model projections. In data spanning 1979 to 2009 the observed trends are significant in some cases but tend to differ significantly from modeled trends. Short title: Panel and multivariate trend equivalence tests