Analysis of a long record of annual maximum rainfall in Athens, Greece, and design rainfall inferences

An annual series of maximum daily rainfall extending through 1860-1995, i.e ., 136 years, was extracted from the archives of a meteorological station i n Athens. This is the longest rainfall record available in Greece and its a nalysis is required for the prediction of intense rainfall in Athens, where currently major flood protection works are under way. Moreover, the statis tical analysis of this long record can be useful for investigating more gen eralised issues regarding the adequacy of extreme value distributions for e xtreme rainfall analysis and the effect of sample size on design rainfall i nferences. Statistical exploration and tests based on this long record indi cate no statistically significant climatic changes in extreme rainfall duri ng the last 136 years. Furthermore, statistical analysis shows that the con ventionally employed Extreme Value Type I(EV1 or Gumbel) distribution is in appropriate for the examined record (especially in its upper tail), whereas this distribution would seem as an appropriate model if fewer years of mea surements were available (i.e., part of this sample were used). On the cont rary, the General Extreme Value (GEV) distribution appears to be suitable f or the examined series and its predictions for large return periods agree w ith the probable maximum precipitation estimated by the statistical (Hershf ield's) method, when the latter is considered from a probabilistic point of view. Thus, the results of the analysis of this record agree with a recent ly land internationally) expressed scepticism about the EV1 distribution wh ich tends to underestimate the largest extreme rainfall amounts. It is demo nstrated that the underestimation is quite substantial (e.g., 1:2 for large return periods and this fact must be considered as a warning against the w idespread use of the EV1 distribution for rainfall extremes.