USING SPECTRAL-ANALYSIS TO DETECT SENSOR NOISE AND CORRECT TURBULENCEINTENSITY AND SHEAR-STRESS ESTIMATES FROM EMCM FLOW RECORDS

Electromagnetic current meters (EMCMs) are frequently used to gather t urbulent velocity records in rivers and estuaries. Experience has show n that, on occasion, the output of these sensors can be affected by co ntamination from various noise sources. These noises may be limited to narrow bands of frequencies and thus fail to produce conspicuous incr eases in observed signal variance. Such 'narrow-band' noises can be di fficult to identify from simple inspection of signal traces or varianc e levels, yet degrade estimates of turbulence statistics, in particula r covariances (used to calculate Reynolds shear stress). This paper de monstrates the usefulness of spectral analysis to detect and character ize narrow-band noise components in turbulent flow records. Statistica l principles underlying the use of spectral analysis for noise detecti on are briefly reviewed. Examples of u and v velocity spectra and cosp ectra are then presented from actual EMCM velocity records from flume and field deployments that were found to be contaminated by such noise s. The sensitivity of the shear stress estimates to even minor noise l evels is demonstrated. The use of spectral analysis to correct varianc e (turbulence intensity) and covariance (shear stress) estimates obtai ned from records contaminated by narrow-band noise is also illustrated .