Estimating higher-order moments of nonlinear time series

Abstract

This study extends the authors' earlier work that addresses the importance of bootstrap methods in computing statistical characteristics of meteorological and climatological datasets. Subsampling confidence intervals for the skewness and kurtosis are developed for nonlinear datasets, for which traditional time series techniques are not applicable. It also provides an example of how to apply subsampling to real data when only a single record of limited length is available: aircraft observations of vertical velocity in the wintertime convective boundary layer over Lake Michigan. This demonstrates the value of bootstrap methods in obtaining reliable confidence intervals for turbulent flows with coherent structures (characterized by non-Gaussian skewness and kurtosis). © 2009 American Meteorological Society.