Title: Stochastic simulation models for multi-site nonstationary time series using wavelets
Authors: Ying Sun - KAUST (Saudi Arabia) [presenting]
Abstract: Many meteorological variables exhibit nonstationarity in time. In particular, the time series appears to have modulated oscillations that may correspond to the recurrent but still changing set of climate conditions. A multi-scale stochastic simulation model is developed using wavelet decomposition. The multiresolution and localization properties of wavelets make them favorable for reproducing different time-varying local features of processes at different time scales. Specifically, for a given time scale, we propose to use a stochastic model based on evolving periodic functions to model the wavelet coefficients that vary in a periodic fashion, and both periodicity and amplitude are allowed to change over time. We apply this approach to analyze the monthly southern oscillation index from 1876 to 2015, and show that the proposed simulation model can successfully reproduce the interdecadal fluctuations, which have the effect of modulating the amplitude and frequency of occurrence of El Nino events. Multi-site simulation models are also developed with an application to a minute-by-minute meteorological dataset collected at multiple monitoring locations from the Atmospheric Radiation Measurement program.