• extreme rainfall;
  • stationarity;
  • eastern Europe;
  • change point;
  • trend analysis;
  • mixture distribution


Annual and seasonal maximum daily rainfall time series from 44 rain gages in Romania, Ukraine, and Moldova are used to examine questions related to the validity of the stationarity assumption, the dependence of extreme rainfall on elevation, and its upper tail properties during the 20th century. Analyses are performed both at the annual and seasonal scales. The validity of the stationarity assumption is examined by testing the records for abrupt and slowly varying changes. The presence of change-points in the mean and variance of the daily maximum rainfall distribution is assessed using the nonparametric Pettitt test, while the presence of monotonic patterns is examined using the nonparametric Mann-Kendall and Spearman tests. These results indicate that most of the violations of the stationarity assumption are related to abrupt rather than slowly varying changes, and that it is difficult to detect an anthropogenic climate change signal in these records. The generalized extreme value (GEV) distribution is used as modelling framework to investigate the dependence of extreme rainfall on elevation and its upper tail properties. Both at the annual and seasonal scales there is no strong dependence of the parameters of the GEV distribution on elevation. These records point to an unbounded-above and heavy-tail behaviour, and discussion of these upper tail properties in light of mixtures of different rainfall-generating mechanisms is provided. Copyright © 2011 Royal Meteorological Society