Author(s): Jang-Gyeong Kim; Seok-Hwan Jang; Hyun-Han Kwon
Linked Author(s):
Keywords: No Keywords
Abstract: Poisson cluster stochastic rainfall generators have been widely applied to reproduce the underlying distribution of the rainfall generating process, especially for generating synthetic sub -daily rainfall sequences. The existing optimization techniques are typically based on individual parameter estimates that treat each parameter as independent. However, parameter estimates sometimes compensate for the estimates of other parameters, which can cause high variability in the results if the covariance struc ture is not formally considered. Moreover, uncertainty associated with model parameters in the rainfall generator is not usually addressed properly. Here, we proposed a hierarchical Bayesian model (HBM) based Neyman-Scott rectangular pulse (NSRP) model, to jointly estimate parameters across weather stations by explicitly considering the covariance structure between a set of parameters. The proposed model was validated by using observed rainfall data obtained from weather stations data over South Korea. It was clearly confirmed that the HBM based NSRP model showed better performance in terms of the identification of parameters, leading to a significant reduction of uncertainty associated with parameters. Moreover, the proposed model showed significant improvement over the existing methods in terms of identifying the parameter space without predefined ranges for the parameters. We further utilized the proposed NSRP model to explore changes in rainfall under climate change. In this context, we have used various climate change scenarios derived from multiple regional climate models.
Year: 2018