Helen L. Johnson and Chris Garrett, J. Phys. Oceanogr., 34, 2359-2372.
Abstract
Estimating the diapycnal mixing rate from standard CTD data
by identifying overturning regions in the water column (the Thorpe
scale approach) provides good spatial and temporal coverage, but is
sometimes limited by instrument noise. This leads to spurious density
inversions which are difficult to distinguish from real turbulent
overturns. Previous efforts to eliminate noise may have
over-corrected and hence underestimated the level of mixing. Here
idealized density profiles are used to identify the magnitude and
characteristics of overturning regions arising entirely from
instrument noise, in order to establish a standard against which CTD
data can be compared.
The key non-dimensional parameters are (1)
the amplitude of the noise scaled by the density change over the
section of profile considered, and (2) the number of data points in
the section of profile. In some cases the product of these, which is
equal to the amplitude of the noise scaled by the average density
difference between consecutive measurements, is more useful than the
second parameter. The probability distribution of ``runlength'', a
useful diagnostic, varies significantly across this parameter space.
Reasons for this are discussed, and it is shown that CTD data very
rarely lie in a region of parameter space where comparison with the
PDF of runlengths for a random uncorrelated series, or its rms value
sqrt(6), is appropriate. The distribution of Thorpe displacements
arising entirely from instrument noise, as well as the Thorpe scale
and the statistics of density inversions, are also discussed.
Analysis of CTD data from the interfaces of the thermohaline staircase
in the deep Canada Basin illustrates how the results can be applied in
practice to help distinguish between signal and noise in marginal
regimes. Density inversions seen in these data are shown to be no
different from those which would result from instrument noise.
This manuscript is available as a PDF file.