호남지역 집중호우 예측을 위한 통계 모형 개발

摘要: Based on the 3-hour weather forecast system and model output statistic method, three statistical models, logistic regression with main effects, effects interaction neural networks, were applied compared. rainfall amount at 17 stations in Honam area corresponding 45 synoptic factors generated by RDAPS during 2002 to 2003 used. The training data validation contructed using cluster analysis because of short period. Some thresholds as decision critria determined receiver operating characteristic curve. According comparison models via Heidke skill scores, is recommended for prediciton occurrence heavy rain area.