Is friction velocity the most appropriate scale for correcting nocturnal carbon dioxide fluxes?

The use of friction velocity u* as the turbulence scale for correcting eddy-covariance carbon dioxide fluxes in low-mixing conditions is questioned. This is done because u* is, itself, a flux and, therefore, its value is highly dependent on the temporal scale used for the analysis. The multiresolution decomposition is applied to data from three different ecosystems in Brazil, to show that u* is well behaved and related to the turbulent mixing only up to the scale that separates the turbulent mixing from the low-frequency exchange. For larger temporal scales, mesoscale fluxes may induce large variability in the friction velocity, so that time series with low turbulent mixing may show an elevated value for u*, and vice-versa. We propose, as an alternative, the use of σw, the standard-deviation of the vertical velocity fluctuations. It is shown that σw has no variability within the mesoscale range and that, therefore, it is a much better scale to quantify the turbulent exchange than u*. The relationship between the two velocity scales is shown to depend on the scale and to be universal for the scales of the turbulent exchange. It is shown that curves of the turbulent carbon dioxide fluxes as a function of the turbulence scale are smoothed when using the friction velocity. Using σw instead of u* in data filtering procedures has two main consequences: easier determination of the threshold for filtering and larger respiration rates of the series classified as turbulent. The improvement is larger for sites where very stable conditions are common.