Optimal Moment Determination in Pome-Copula Based Hydrometeorological Dependence Modelling
Dengfeng Liu,Dong Wang,Vijay P. Singh,Yuankun Wang,Jichun Wu,Lachun Wang,Xinqing Zou,Yuanfang Chen,Xi Chen
DOI: https://doi.org/10.1016/j.advwatres.2017.04.016
IF: 5.361
2017-01-01
Advances in Water Resources
Abstract:Copula has been commonly applied in multivariate modelling in various fields where marginal distribution inference is a key element. To develop a flexible, unbiased mathematical inference framework in hydrometeorological multivariate applications, the principle of maximum entropy (POME) is being increasingly coupled with copula. However, in previous POME-based studies, determination of optimal moment constraints has generally not been considered. The main contribution of this study is the determination of optimal moments for POME for developing a coupled optimal moment-POME-copula framework to model hydrometeorological multivariate events. In this framework, margins (marginals, or marginal distributions) are derived with the use of POME, subject to optimal moment constraints. Then, various candidate copulas are constructed according to the derived margins, and finally the most probable one is determined, based on goodness-of-fit statistics. This optimal moment-POME-copula framework is applied to model the dependence patterns of three types of hydrometeorological events: (i) single-site streamflow-water level; (ii) multi-site streamflow; and (iii) multi-site precipitation, with data collected from Yichang and Hankou in the Yangtze River basin, China. Results indicate that the optimal-moment POME is more accurate in margin fitting and the corresponding copulas reflect a good statistical performance in correlation simulation. Also, the derived copulas, capturing more patterns which traditional correlation coefficients cannot reflect, provide an efficient way in other applied scenarios concerning hydrometeorological multivariate modelling. (C) 2017 Elsevier Ltd. All rights reserved.