A Segmented Editing and Detrending Method for Discrete Fourier Transforms

摘要

Time series data must be prepared for the discrete Fourier transform by removing wild points (fliers), as well as slowly varying trends or components, as both upset spectral validity and resolution. An editing/detrending procedure to do this (complete with a Fortran subroutine) is presented. The method is analyzed and tested on CODAR (Coastal Ocean Dynamics Applications Radar) data with good results.