Abstract
　In the Yamato river basin, located in the Kinki district, flood risk has been increased because the middle zone of the basin has been urbanized rapidly in recent three decades. This basin suffered severe disaster from flood caused by a record heavy rainfall in August 1-3, 1982. The flood formation process is not simple because this basin has so large area as 1,070 km2.
　In this paper, the statistical characteristics of rain storms are examined by applying the techniques of both principal component and cluster analyses to the data of 12, 24, and 48-hour rainfalls observed at the 26 points in and around the basin for 19 rain storm cases, in which peak discharges at Kashiwara were over l,000 m3/s or 2 days rainfalls at Nara were over 100 mm in the years 1966-1985. The result of principal component analysis shows that these rain gauge stations are able to be summarized in 8 groups to represent characteristics of each rain storm. While, the result of cluster analysis is slightly different from the former but support it in a practical sense. Finally, the correlation of peak discharges at Kashiwara with rainfalls in 12, 24, and 48 hours are examined. The result shows that the peak discharges have strong correlation with areal mean 12-hour rainfalls, especially when the rain- fall becomes heavy in new urbanized areas along the main channel.

【キーワード】：豪雨，主成分分析，クラスター分析
Key words: heavy rainfall, principal component analysis, cluster analysis

*京都大学防災研究所
Disaster Prevention Research Institute, Kyoto University
**総合調査設計（株）
Sogochosasekkei Co., Ltd.