Modified singular spectrum analysis for despiking acoustic Doppler velocimeter (ADV) data

Abstract Spectral investigation of spike contaminated velocity time series recorded by acoustic Doppler velocimeter (ADV) remains a challenging task. In this paper, we propose a singular spectrum analysis (SSA) method with novel grouping criteria to remove spikes from highly contaminated velocity time series. The proposed technique is tested on the highly contaminated experimentally observed velocity time series data in the vegetated sand bed channel. The performance of this method is assessed by the power spectral density of clean ADV time series data. Special importance is placed to produce the accurate power spectral density in streams with high energy fluctuation. The trends show that the reconstructed velocity time series is free from spikes and noise components. Furthermore, the velocity power spectra of the filtered AVD time series data satisfy the Kolmogorov −5/3 law in inertial sub range.