Trends of precipitation variables on different datasets

It is a well-established fact that different types of data (station, gridded, reanalysis) possess different statistical characteristics, e.g. for higher-order moments, extremes, and trends. In this contribution we examine the long-term changes in precipitation characteristics on different data sources over Europe. We calculate and display differences between the datasets and attempt to identify causes for the differences and for specific behavior of the datasets. We used data from stations across Europe (ECA&D project), gridded data (E-OBS) and reanalysis (NCEP/NCAR, JRA-55). We mainly analyze the trends of the seasonal total amount, intensity and probability of precipitation. Long-term trends of seasonal values of precipitation variables and their statistical significance are calculated by non-parametric methods (Mann-Kendall test, Kendall statistic). The analysis is conducted on a seasonal basis, with emphasis on winter and summer. We found that each of the datasets has its advantages and drawbacks. Trends in reanalysis deviate considerably from the other datasets mainly because the type and amount of data assimilated into them change in time. The weakness of the grid data sets is the unstable number of stations entering the interpolation in time, and the lack of representativeness of some climate stations is the main disadvantage of the station data.