Streamflow generation using a multivariate hybrid time series model

摘要

The long-term water resources plan in Korea is usually established on the basis of the so-called water balance analysis that estimates the water availability over an entire basin. The analysis generally uses only a single 30-year historical record of natural Streamflow. This study suggested that the water balance analysis should include the hydrological uncertainty in its results. For this purpose, a multivariate Streamflow generation model called the multivariate contemporaneous PAR(1)NT-hybrid model was proposed and applied to a multi-site monthly Streamflow generation problem for the Han River basin in Korea. The proposed model was then compared with a multivariate PAR(1) and a disaggregation model. This study showed that the proposed model reproduces cross-correlations at various lags better than the traditional parametric generation models.