engleski

Modelling of extreme dry spells in Croatia – the threshold selection

In order to estimate the occurrence of long dry spells different return periods obtained from certain extreme value distribution are usually employed. When dealing with climatological extremes the block maxima approach is traditionally used. However, in recent years the peaks over threshold (POT) approach comes into the focus because of its ability to better utilize the available data. The main problem in this approach is how to choose the threshold in a systematic way. In this study we put emphasis on this issue while analysing dry spell series in Croatia. Daily precipitation data from the 1961-2010 period at 25 meteorological stations are used to construct dry spell series based on the daily precipitation amount of less than 1 mm. Simulated dry spell series based on Generalized Pareto (GP) distribution are also used. Several methods for threshold selection are examined employing a whole range of percentile values. It is found that unique threshold can not be recommended for dry spells. Instead, it seems the best option is to utilise GP distributions over a range of percentiles. The mean values of GP parameters thus obtained are subsequently used for estimation of return periods.

dry spells; Croatia; threshold; Generalised Pareto distribution

nije evidentirano