Removal of ‘strip noise’ in radio-echo sounding data using combined wavelet and 2-D DFT filtering

• Published in: Annals of glaciology Vol. 61; no. 81; pp. 124 - 134
• Main Authors: Wang, Bangbing, Sun, Bo, Wang, Jiaxin, Greenbaum, Jamin, Guo, Jingxue, Lindzey, Laura, Cui, Xiangbin, Young, Duncan A., Blankenship, Donald D., Siegert, Martin J.
• Format: Journal Article
• Language: English
• Published: Cambridge, UK Cambridge University Press 01.04.2020
• Subjects: Aircraft; Components; Crystal oscillators; Data; Data analysis; Data processing; Decomposition; Earth science; Echo sounding; Echoes; Filtration; Flow structures; Fourier analysis; Fourier transforms; Glaciers; glaciological instruments and methods; Ice; Ice cover; Ice sheets; Ice thickness; Interference; Masking; Multiscale analysis; Noise; Radar; Radar data; Radio; radio-echo sounding; Spectral analysis; Spectrum allocation; Spectrum analysis; Strip; subglacial exploration geophysics; subglacial processes; Thickness measurement; Two dimensional analysis; Wavelet transforms; subglacial processes; glaciological instruments and methods; radio-echo sounding; subglacial exploration geophysics
• ISSN: 0260-3055; 1727-5644
• DOI: 10.1017/aog.2019.4

Radio-echo sounding (RES) can be used to understand ice-sheet processes, englacial flow structures and bed properties, making it one of the most popular tools in glaciological exploration. However, RES data are often subject to ‘strip noise’, caused by internal instrument noise and interference, and/or external environmental interference, which can hamper measurement and interpretation. For example, strip noise can result in reduced power from the bed, affecting the quality of ice thickness measurements and the characterization of subglacial conditions. Here, we present a method for removing strip noise based on combined wavelet and two-dimensional (2-D) Fourier filtering. First, we implement discrete wavelet decomposition on RES data to obtain multi-scale wavelet components. Then, 2-D discrete Fourier transform (DFT) spectral analysis is performed on components containing the noise. In the Fourier domain, the 2-D DFT spectrum of strip noise keeps its linear features and can be removed with a ‘targeted masking’ operation. Finally, inverse wavelet transforms are performed on all wavelet components, including strip-removed components, to restore the data with enhanced fidelity. Model tests and field-data processing demonstrate the method removes strip noise well and, incidentally, can remove the strong first reflector from the ice surface, thus improving the general quality of radar data.