POWGRAF2: a program for graphical spectral analysis in cyclostratigraphy

Abstract

Spectral analysis is a widely used statistical tool in cyclostratigraphy. The aim is to detect hidden periodicities in noisy temporal (or spatial) series derived from stratigraphic sequences at outcrops or boreholes. A graphical program to assist the geologist with the spectral analysis is presented in this paper. The program follows a sequential approach to spectral analysis in which the automatic graphical displays of the results guide the user through the analysis. Features of the program are the imputation of missing data, linear drift detrending, selection of white noise or red noise from the correlogram, possibility of evolutive spectral analysis (spectrogram), and a variety of spectral estimators and confidence bands for the power spectrum estimates consistent with the selected model for the underlying noise (red or white). The analysis is enhanced by graphical screens with the original series, detrended series, experimental correlogram, theoretical model of an AR(1) correlogram fitted to the series, confidence bands for the estimated correlogram, estimated power spectrum, underlying white or red noise spectrum and 90%, 95% and 99% upper confidence bands for the estimated power spectrum. The performance of the program is illustrated using a well-known time series of δ 18O at site 607 of the Deep Sea Drilling Project.