Analysis of turbulent fluxes with the means of the continuous wavelet transform on the Belgian ICOS sites of Lonzée and Vielsalm

Abstract

The eddy covariance and Fourier decomposition are the standard methods for the calculation and analysis of the turbulent flux. Nonetheless, their applicability is restricted by the necessity of steady- state flow conditions, rendering them unsuitable for analyzing mesoscale motions, which are the most prominent cause for energy balance unclosure according to recent review papers. Consequently, the wavelet transform has gained traction for turbulent flux analysis in the past few years. Originally developed for signal analysis, this tool effectively captures non-stationary power by decomposing the signal in both the temporal and frequency domains, and it can be used to derive flux cospectra and calculate the turbulent flux. The present paper makes use of this tool in the scope of two research objectives. First, the parameters of the wavelet covariance formula are adapted to establish equivalence between wavelet and eddy covariance are equivalent for stationary half-hour records. Through a sensitivity analysis, the influence of parameters such as padding (introducing artificial values at signal boundaries), the shortest analyzed period, the scale step (δ j), the threshold for the cone of influence (qcoi) in covariance computation are explored, along with the choice of mother wavelet and normalization. The second research objective aims at investigating components of the turbulent flux that oscillate in periods larger than 30 min on two ICOS sites, namely Lonzée and Vielsalm, and that could explain energy balance unclosure on these sites.To achieve this, CO2, H2O, and temperature measurements are utilized to derive wavelet flux cross-scalograms, cospectra and ogives. These different representations revealed the presence of flux components in the low-frequency range that are much important that the energy balance residuals, which suggests that the flux aggregated over 3.2 hours is overestimated, as a consequence of a too short averaging time or as the consequence of biased observations of quasi-stationary motions. In the conditions of mean wind and with the nature and scale of heterogeneities on Lonzée and Vielsalm, the presence of quasi-stationary motions is possible but could not be ascertained.