A NEW THRESHOLD SELECTION METHOD USING L-MOMENT FOR POT ANALYSIS OF PRECIPITATION TIME SERIES

Abstract

 Torrential rains which cause disasters seem to have increased and become severer recently. Extremes of observed precipitation at a location or catchment averaged one have been used for extreme analysis. There are two methods of extracting extremes from a time series for extreme analysis; annual maximum series(AMS) and peaks over threshold(POT). AMS has been widely used from its simplicity, however, POT used practically less than AMS because of its complicated process in threshold selection and POT extraction. This paper proposes a new threshold selecting method using L-moment and shows several notes in practical use. The result shows it estimates more stable return level with accumulating observation than GEV with AMS.