| Abstract [eng] |
Extreme event - natural, ecological, social or technical event which cause danger to people and their property. The most usual extreme events can be caused by nature. Their strength and frequency often depends on the geographical location and global warming. It is important to assess their regularity, size, recurrence in order to avoid or defend from them in the future. This research of extreme event is focused on extreme precipitation. Data is provided by Lithuania hydrometeorology service for time period of 1961 and 2016. According to criteria of extreme events the extreme precipitation is defined as heavy rain which accumulates 50 mm or more in specific area during 12 hours time interval. Firstly, analysis was done for numbers of extreme events during a year period. In order to analyse the data, the following discrete distributions were used: geometric, Poisson, negative binomial discrete and uniform. The most accurate data distribution was found to be geometric. In research of extreme precipitation amount the following extreme values distributions were used: Weibull, Gumbеl and Generalized Extreme Value (GEV). Selection of distribution was done by applying different nonparametric methods, Chi square and other criteria. During the research, it was found that the best distribution for the data is GEV distribution. Then, using this distribution the frequencies and return periods of extreme events were estimated and confidence intervals of these estimates were calculated. Using Gumbеl, Wеibull and GEV distributions for 80 mm or higher amount of precipitation event per year the estimated frequency is: 1,00; 0,76; 0,63. Using GEV distibution the return period of such event is 2,35 years. The return period of observed anomaly, i.e. 250 mm amount of precipitation, using GEV distribution is 135 years. Calculations were made by using MS Еxcеl software, CumFrеq, ЕаsyFit, MАTLАB. |
