apcc logo

극단값 모형을 기반한 국내 강우 현상 모의

저자
김무섭 박사
 
작성일
2016.01.23
조회
241
  • 요약
  • 목차

In this study, we analyze heavy precipitation of some regions in the Korea peninsula to simulate rainfalls in rainy season. We use hourly precipitation data and generate series of rainfall events that seem to be statistically independent. By following the procedure of extreme value analysis, a multivariate extreme value model is set up for extremely large rainfall amounts of several sites covering a region. Moreover, based on this model, we propose a rainfall simulation algorithm which will be a foundation for the study of multisite weather generator.

 

We have worked for the upper basins of Nakdong and han rivers and established multivariate extreme value models. Moreover, in order to completely simulate rainfalls in these basins, we have also analyzed durations of heavy rainfall events and interval times between adjacent rainfall events. as results, Gamma Generalized Linear Models are set up for durations, while random sampling turns out to be appropriate for simulating intervals. Using the combination of these resulting models, we carry out simulations and compare the synthetic data with the observations in terms of cumulated amount in the whole rainy season, maxima of daily rainfalls and 3-day cumulated rainfalls, the number of wet days.