The aim of this study is to investigate the low-frequency characteristics of diurnal turbulent scalar spectra and cospectra near the Amazonian rain forest during the wet and dry seasons. This is because the available turbulent data are often nonstationary and there is no clear spectral gap to separate data into "mean" and "turbulent" parts. Daubechies-8 orthogonal wavelet is used to scale project turbulent signals in order to provide scale variance and covariance estimations. Based on the characteristics of the scale dependence of the scalar fluxes, some classification criteria of this scale dependence are investigated. The total scalar covariance of each 4-hour data run is partitioned in categories of scale covariance contributions. This permits the study of some statistical characteristics of the scalar turbulent fields in each one of these classes and, thus, to give an insight and a possible explanation of the origin of the variability of the scalar fields close to the Amazonian forest. The results have shown that a two-category classification is the most appropriate to describe the kind of observed fluctuations: "turbulent" and "mesoscale" contributions. The largest contribution of the sensible heat, latent heat, and CO2 covariance contributions occurs in the "turbulent" length scales. Mesoscale eddy motions, however, can contribute up to 30% of the total covariances under weak wind conditions. Analysis of scale correlation coefficient [r(Tvq)] between virtual temperature (Tv) and humidity (q) signals shows that the scale patterns of Tv and q variability are not similar and r(Tvq) <1 for all analyzed scales. Scale humidity skewness calculations are negative during the dry season and positive during the wet season. This suggests that different boundary layer moisture regimes occur during the dry and wet seasons.
- wavelet analysis