Switching regime analysis for nonlinear hydrologic time series

Time series models are valuable analytical tools for assessing potential impacts of climate change on water resources systems. Nevertheless, few studies have been found applying switching regime analysis to nonlinear hydrologic time series forecasting or simulation. In this study, an approach is proposed to detect switching dynamics for hydrologic time series based on a hidden Markov model. The hidden Markov model theory and the change-point detection algorithm are first described. The applicability of the approach to hydrologic time series is then examined in a simulation study with long-term chaotic data generated from MacKay-Glass time series. A linear system change analysis method with linearization of nonlinear time series is also applied to the case data for comparison. Taking the advantage of the hidden Markov model in the algorithm of the change-point examination, the presented approach can effectively deal with the detection of switching regimes in nonlinear time series scenarios and ensure the segmentation of time series data into individual modes. The results indicate that the presented approach is a potential tool for studying switching regimes in hydrologic systems. The linearization method is suitable for trend analysis, but unable to capture system dynamics. This work is an attempt to seek a viable framework for dealing with nonlinear hydrologic time series analysis.