Assessing the applicability of six precipitation probability distribution models on the Loess Plateau of China
Assessing the applicability of six precipitation probability distribution models on the Loess Plateau of China
复制标题
六种降水概率分布模型在中国黄土高原的适用性评价
DOI：
10.1002/joc.3699
复制
发表时间：
2014-02
期刊：
影响因子：
--
通讯作者：
Chen Jie
中科院分区：
文献类型：
--
作者：
Li Zhi;Brissette Francois;Chen Jie
日降水的随机模拟对许多水文和农业应用都很有用；但是，必须对降水发生器的能力进行评估，以确保准确的降水模拟。特别是，降水概率分布的适当选择是至关重要的。中国黄土高原属于半干旱气候，季风影响强烈，土壤是世界上最易侵蚀的土壤之一。降水的年变率很大，而且经常发生特大降雨事件，这使该地区对降水的随机产生非常具有挑战性。基于1961-2009年中国黄土高原47个站点逐日降水资料，比较了指数分布、伽玛分布、威布尔分布、偏正态分布、混合指数分布和混合指数/广义Pareto分布6种降水概率分布的表现。结果表明，使用越来越复杂的降水分布有助于更准确地模拟降水。然而，所测试的分布都不能模拟所有观测到的降水统计特征。三参数模型优于模拟观测到的均值和方差。混合指数/广义Pareto分布最能模拟降水的频率分布和年际变化，而偏正态分布最能模拟极端降水事件。综上所述，由于黄土高原的侵蚀高度依赖于极端降水，因此偏正态分布可能是最佳候选，因此建议在黄土高原上使用。
Stochastic modelling of daily precipitation is useful for many hydrological and agricultural applications; however, the ability of the precipitation generator should be assessed to ensure accurate precipitation simulation. In particular, the appropriate choice of a precipitation probability distribution is of utmost importance. The Loess Plateau in China has a semi‐arid climate with strong monsoon influence and contains some of the most erodible soils in the world. The large annual variability in precipitation and the common occurrence of very large rainfall events makes this region very challenging for stochastic generation of precipitation. Accordingly, the objective of this study is to compare the performances of six precipitation probability distributions (exponential, gamma, Weibull, skewed normal, mixed exponential and hybrid exponential/generalized Pareto distributions) on the Loess Plateau of China based on daily precipitation data of 47 stations during 1961–2009. Results indicate that using increasingly more complex precipitation distributions contribute to more accurate precipitation simulation. However, none of the tested distributions is able to simulate all the observed statistical characteristics of precipitation. The three‐parameter models are superior to simulating the observed mean and variance. The hybrid exponential/generalized Pareto distribution is the best at simulating the frequency distributions and interannual variations of precipitation while the skewed normal distribution performs the best in reproducing extreme precipitation events. Overall, as erosion on the Loess Plateau is highly dependent on extreme precipitation, the skewed normal distribution may be the best candidate and therefore is recommended on the Loess Plateau.
登录
查看更多内容
DOI：
10.1111/j.1752-1688.2008.00296.x
发表时间：
2009-04
期刊：
JAWRA Journal of the American Water Resources Association
影响因子：
--
作者：
Jie Chen;X. Zhang;Wenzhao Liu;Zhi Li
通讯作者：
Jie Chen;X. Zhang;Wenzhao Liu;Zhi Li
影响因子：
6.4
作者：
FRENCH, MN;KRAJEWSKI, WF;CUYKENDALL, RR
通讯作者：
CUYKENDALL, RR
影响因子：
5.4
作者：
D. Woolhiser;J. Roldan
通讯作者：
D. Woolhiser;J. Roldan
影响因子：
1.1
作者：
Semenov, MA;Brooks, RJ;Richardson, CW
通讯作者：
Richardson, CW
影响因子：
6.4
作者：
F. Castellví;I. Mormeneo;P. Pérez
通讯作者：
F. Castellví;I. Mormeneo;P. Pérez