Parametric versus non-parametric estimates of climatic trends

Two methods of estimation of the trend magnitude are compared: the parametric one (least-squares regression) with the non-parametric one (median of pairwise slopes). The comparison is carried out for seasonal and annual trends of ten climatic variables at a network of stations in the Czech Republic. We show that the difference between the two trend estimates is very small, falling well within the 95% confidence limits of the parametric estimate. The magnitude of the difference does not depend on the degree of normality of the distributions, with the exception of two variables, maximum temperature and precipitation, for which a slight dependence is observed. As a by-product, the normality of seasonal and annual means of the climatic variables is evaluated by the Kolmogorov-Smirnov one-sample test.