## Method

The aim of the study is to reveal the most promising atmospheric climate change indicators for RO accessible parameters in the UTLS domain for the 36 IPCC+ regions. Altogether 20 different realizations (runs) of A2 and B1 scenarios of the ECHAM5, CCSM3, and HadCM3 were considered (see Sect. 2).

Linear trends (least-square estimates), b, were calculated over 2001-2050 for each pressure level based on anomalies with respect to the 1961-1990 climatolog-ical mean. Given the regression line obtained, y(t) (Eq. 1), the regression residuals e(t) (Eq. 2) describe the deviations of the observed values y(t) from the regression line:

The variance of the residuals, se2, is given by se2 = (n - 2)-1£ ei(t)2, (3)

i where n is the number of samples (years) in the time series.

The significance of the trends was determined according to Santer et al. (2000) and Wilks (2005), using Students t-test and considering lag-1 autocorrelation (r1), reducing n to the effective sample size ne = n(l - rl)/(l + ri), which was used to determine the critical t-value. The test value tb is given by the ratio of the trend b itself and the standard error of the trend sb given by

The coefficient of determination, R2, was chosen as goodness-of-fit measure. It is the ratio of the sum of squares of deviations of the regression line from the mean value y (SSR) and the sum of squares of deviations from the mean value (SST):

Based on the results of the trend analysis, criteria for best trend indicators were formulated as follows:

• for each scenario (A2 or Bl) all—or all but one—trends have the same algebraic sign,

• at least 3/4 of all runs (A2 and Bl) show a goodness-of-fit of R2 > 0.5,

• at least 3/4 of all runs (A2 and Bl) show a statistical significance > 90% for the

The criteria were carefully selected to ensure the unveiling of the most robust spatially dependent trend characteristics of all RO accessible parameters in the UTLS region.

## Post a comment