Spectral analysis of synthetically affected FG5 absolute gravimeter residuals

An instrumental or environmental disturbance (signal plus noise) in absolute gravimeter FG5 observations becomes visible by analyzing the residuals, which represent the misfit from the theoretical acceleration parabola. While spectral analysis of FG5 residuals via the classical discrete Fourier transform (DFT) is limited by the non-equispaced nature of the FG5 observations, the Lomb-Scargle periodogram can analyze non-equispaced observations and can be used to estimate (detect) the signal content of FG5 residuals. For the task of revitalization of noisy absolute gravimetry data sets it is interesting to first investigate the detectability of synthetically introduced disturbances in FG5 residuals using Lomb-Scargle periodogram analysis. Based on the performed frequency analysis, a heuristically derived formula using a Gaussian Bell Summation is used for estimating the impact on gravity and to eventually filter out identified disturbances using a modified FG5 data adjustment algorithm. The results demonstrate that the drop frequency used (equivalent to the number of fringes used) changes the sensitivity of Lomb-Scargle analysis in terms of frequency estimation and resolution. Using a different number of fringes consequently leads to different gravity values, which must be considered in FG5 comparisons.;A new wavelet-based approach for analyzing FG5 residuals is also presented. In this approach, a dyadic bundle of drop residuals is analyzed as a whole, while Lomb-Scargle analyzes single drops. Since the appearance of real signals and disturbances are not limited to the duration of a single drop, this analysis bears more potential to identify environmental and instrumental disturbances. By appropriate interpolation, an evenly spaced and dyadic time series is obtained. The discrete wavelet transformation is used to select a relevant frequency range and to denoise the time series of drop residuals. The denoised time series is then analyzed with the wavelet power spectrum using the continuous wavelet transformation. Finally, the new procedure is tested with synthetic and real data. Results of the wavelet analysis indicate superior performance over Lomb-Scargle analysis in terms of detectability and accuracy in frequency determination.