关键词: 油菜, 连阴雨日数, 持续降水量, K-均值聚类分析, 评估指标

Abstract:

Based on daily meteorological data of 44 observation stations and rape yield data from each county (city) from 1980 to 2014 in Anhui province, yield losses of rape and relevant characteristic quantities of continuous rain such as continuous rainy days, continuous precipitation and sunshine duration during anthesis were calculated by use of mathematical statistics. According to the principle that the factors with the greatest impact on the yield are induced and the correlations between the factors are lowest, the disaster causing factors were chosen and their critical values were determined. In the last, the grade indices of continuous rain disaster were established by K- means clustering analysis method and the index’s veracity was verified by means of the yield losses and investigated disaster date during 2010-2014.The results showed that the main influence factors of continuous rain disaster to rape during anthesis were continuous rainy days and continuous precipitation. The threshold values of continuous rainy days were 3 days and the threshold values of continuous precipitation were 20mm, 50mm and 70mm respectively in Jianghuai, Yanjiang and South Anhui. Based on this, the grade indices of continuous rain disaster was established respectively for Jianghuai, Yanjiang and south Anhui by using K- means clustering analysis method. The damage could be defined as slight when the yield losses was between 5%-10%，moderate when the yield losses was between 10%-20%, severe when the yield losses was between 20%-30% ,extra severe when the yield losses was more than 30%.The testing results indicated that the accurate rate of slight and moderate indices were higher（88%-100%）, while the accurate rate of severe and extra severe indices were relatively low（65%-70%）.

Key words: Rape, Continuous rainy days, Continuous precipitation, K-means clustering analysis, Evaluation indices