干旱重现期大小是用于评价干旱事件严重程度的重要指标。干旱重现期的计算涉及给定阈值下干旱过程划分（识别）、样本系列分布函数拟合等关键环节，其中干旱阈值的确定是前提。本文提出以干旱事件的最长调查期为约束条件确定干旱阈值的思路，即根据样本计算的干旱事件最大重现期不应超过最长调查期,以此为据确定干旱阈值并从样本序列中识别干旱事件。同时,针对因干旱历时样本经验点据“平台式”过度集中而导致的频率曲线适线困难问题，建议采用基于游程理论的游程长度分布函数估计干旱历时概率分布。以青海民和县1932年－2010年的月降雨资料为例，对上述方法进行了应用研究,结合Copula函数计算了干旱事件的重现期。

The concept of return period is commonly used to evaluate the severity of drought events. To compute the return period for a given drought contains drought recognition and selection of distribution function and so on, which relies on the determination of drought threshold. This paper proposed to use the maximum survey period of historical drought as a reference to determine the drought threshold, in other words, the resulting return period based on frequency analysis should not exceed the maximum survey period of historical drought. Besides, considering that the drought duration samples tend to concentrate on several values and lead to the platform-concentrated phenomenon of empirical points, the distribution function of run length was suggested to fit the drought duration samples. The monthly precipitation series of Minhe County in Qinghai province covering the period from 1932 to 2010 was used to demonstrate the proposed methods, where the Copula function was employed to build a bivariate joint distribution of drought duration and drought severity to compute the return period of drought event.

国家自然科学基金重大项目（51190095）