Copula-based seasonal rainfall simulation considering nonstationarity

摘要

In recent years, anthropogenic and climate changes have greatly impacted the probabilistic behavior of hydrometeorological variables, and have challenged the stationarity assumption of marginal distributions of individual characteristics of variables as well as their temporal dependence structure inherent in copula-based modelling. This study developed a dynamic copula-based simulation model (DCSM) for single-site seasonal rainfall generation, observed at Beijing Station in Haihe River basin, China. The model entailed four phases: (1) nonstationary modelling of the margins by the Generalized Additive Models for Location, Scale and Shape (GAMLSS); (2) dynamical copula method to describe the nonstationary temporal dependence structure of rainfall observed at adjacent two seasons; (3) a dynamic copula-based conditional quantile function to generate simulated series; (4) performance assessment of the simulated series by the proposed DCSM model with the preservation of basic statistics of each sequence and simulation accuracy at each data point. Comparison of the nonstationary and stationary models to simulate the seasonal rainfall series shows that the incorporation of nonstationarity in hydrometeorological simulation would improve the performance of simulation accuracy.