Error Evaluation of the Fourier Analysis for the Isolation of Major Tidal Components From Coastal Groundwater Data Based on Numerical Experiments

Abstract

Sinusoidal tidal components in groundwater-level time series data collected from coastal aquifers affected by tide-level oscillations are used to determine aquifer hydraulic parameters through an analysis of the attenuation and delay during landward propagation. One method to determine the amplitude and initial phase of the sinusoidal tidal component, or to isolate the component from such groundwater observation data, is Fourier analysis. Two categories of errors are inevitable in the isolation of a sinusoidal component of a specific period from a finite-length time series using Fourier analysis. The errors are relevant to the component periods and the analyzed time series length. Therefore, appropriate selection of the analyzed time series length is essential to accurately isolate a tidal component from the observation time series data. This investigation addressed errors in the isolation of major tidal components, which have approximately diurnal or semidiurnal periods, from an artificial time series composed of the tidal components typical of coastal aquifers. The errors expected in isolation with analyzed lengths from one to one hundred days were investigated. The resultant outcome was an example selection of recommended time series lengths that should enable relatively accurate isolation of major tidal components by Fourier analysis.