Real-Time Screening Techniques for Doppler Current Profiler Data

Abstract

Computationally simple criteria for the identification of spurious Doppler profiler velocity estimates are described and evaluated. Their effectiveness is determined by changes in the confidence intervals and mean values of Doppler shear profiles. Our tests show that indicators based on measures of acoustic signal quality (signal-to-noise ratio, and spectral width) do not provide a mean reduction in confidence intervals. We show that a combination of velocity quality screening criteria can reduce the confidence intervals of velocity estimates made with a 150 kHz profiler by as much as 25 percent. After screening, both components of horizontal currents are about as accurate, almost independent of depth, to at least 240 m. More importantly, data screening also removes a depth-dependent bias (as large as 0.05 m s?1) from velocity profiles. For the indicators tested, we recommend threshold values for our system configuration; these values can serve as guidelines to suitable settings in other applications. To obtain maximum accuracy of shear profiles (and absolute profiles when accurate navigation data are available) some form of data screening is essential.