STATISTICAL MODELLING OF EXTREME PLUVIOMETRIC EVENTS BY MEANS OF GENERALIZED PARETO DISTRIBUTION

The extreme events are one of the most important aspects of the climate, especially regarding its impacts on the society and the environment. A rainfall episode that exceeds over a relatively high threshold is one of the most interesting extreme events. In general, the extreme events are characterised by a low probability of occurrence and long return periods so that the study of distribution tails of special interest in dealing with several problems. In order to model extreme hydrological events, such as floods, the Generalised Pareto Distribution (GPD) has been chosen for its adequacy in modelling that peaks over a threshold. In this work, both time series of daily precipitation of the Santa Cruz de Tenerife observatory, with a record from January of 1938 to April of 2002, and time series of daily precipitation of the Badajoz observatory, with a record from January 1880 to December 2001 have been modelled by means of the GPD. For comparison, the General Extreme Value (GEV) distribution is used to model the extreme maximum precipitation series obtained from the annual values in the same period. In statistical terms, the threshold is usually expressed as the event corresponding to a specified exceed probability, i. e., the level exceeded on the average once in T years. This level is calculated for both distributions and its empirical value is estimated from the series. The comparative analysis shows that the GEV distribution substantially underestimates the observed occurrence frequencies, to return period higher two years, while the fit of GPD to the empirical results are better.