中文摘要：

以邵阳市15个雨量站逐月降雨资料为基础，采用适线法估计干旱历时和干旱烈度的分布函数，利用3种Archimedean Copula函数分别构建了两干旱特征变量之间的联合分布，并比较了3种Copula函数的拟合效果，最后分析了邵阳市干旱重现期的空间分布特征。结果表明：Copula理论能够为多变量干旱特征分析提供有效的研究途径，Archimedean Copula函数中的Gumbel—Hougaard Copula和Frank Copula函数较好地描述各站干旱特征变量之间的相互关系；适线法避免了基于数据估计分布函数参数的不合理性，使基于Copula函数的频率分析结果更客观可靠；区域干旱具有显著的空间分布特性，各站干旱程度差异较大。

英文摘要：

In this paper, monthly rainfall data at 15 stations in Shaoyang City were used to analyze the characteristics of spatial distribution of regional drought. A drought event is defined as the monthly rainfall below the drought thresholds according to the classification of meteorological drought in China, and the drought event is described by two characteristic variables, i.e., drought duration and drought severity. Here, drought duration was fitted with the exponential distribution, while the gamma distribution was used for drought severity. The distribution function of each drought characteristic variable was determined by the curve fitting method instead of maximum likelihood estimation so as to avoid possible discrepancies. Furthermore, three kinds of Archimedean copulas, i.e., Gumbel-Hougaard copula, Clayton copula, as well as Frank copula, were employed to construct the joint distribution of drought characteristic variables. Then, the copulas＇ fitting results were compared, from which the copula of the best fitting results for the estimation of drought return period was selected. Finally, characteristics of the spatial distribution of drought return periods of Shaoyang City in typical drought years were analyzed. Empirical results show that copula theory provides an alternative for multivariable analysis of drought characteristic. The Gumbel-Hougaard copula as well as Frank copula can well reflect the relationships among drought characteristic variables, both of which have similar fitting results. The curve fitting method can avoid the ineffectiveness from estimation of parameters based on data, making the results of copula-based frequency analysis more reliable, with the fitting errors of copulas using the curve fitting method much smaller than those using maximum likelihood estimation. At the Baimaoping station, the average decreasing amplitude of 3 copulas＇ fitting errors employed in this paper reached up to 60.64%. Regional drought has a remarkable characteristic of spatial distribution, resulting in di