Summary

This study concerns a comparison of the ECHAM3/T42 simulated series of daily extreme temperatures and series observed in southern Moravia (a part of the Czech Republic). ECHAM climate model was developed from the ECMWF model (the former part of its name EC) and parametrizations were created at the Max Planck Institute in Hamburg (the latter part of the abbreviation HAM). Simulated (1CO2) times series of daily variables have rarely been validated against the real datasets. In this paper, attention is focused on autocorrelation coef®cients whose estimates are computed by the jackknife method and differences in the estimates between the simulations and observations are examined. It is shown that for the average simulated series (4 gridpoints) the jackknife autocorrelation coef®cients are substantially larger in all seasons than those computed for the average series in Moravia (5 stations). The daily extreme temperature variability is underestimated in the simulations, the persistence of the simulated series being much higher. In order to gain an additional insight into this ®nding trimmed means and trimmed sample variances are computed. An examination of frequencies of day-to-day changes (absolute values) calculated from the observations and simulations shows that small day-to-day temperature changes are clearly preferred in the model at the expense of larger changes which are recorded in Moravia. It is obvious that the largest changes observed are not captured in the simulations. 1