The data clustering methodology is a recent method based on a bounded variation, finite element, vector autoregressive factor method (FEM-BV-VARX) and is applied to the SnoHATS dataset of near-surface stable boundary layer (SBL) turbulence. We use the FEM-BV-VARX methodology to characterize the influence of non-turbulent, submesoscale motions on the turbulence in the SnoHATS dataset. Regimes are thereby identified, two of them weakly stable and two very stable turbulence states. In each identified regime, the variability of turbulent momentum fluxes is characterized here using an extended multiresolution flux decomposition methodology. We will show that the transport properties in each regime of near-surface SBL turbulence differ. The same methodology is used to investigate the scales of motion responsible for shear generation of turbulence.
In regimes identified as little influenced by submesomotions, traditional weakly stable turbulence behavior is highlighted by the multiresolution flux decomposition analysis. In the regimes that are identified as having a strong influence of submeso forcing, the results suggest a likely direct transfer of energy from the submesoscale horizontal velocity fluctuations to turbulent vertical velocity fluctuations. Those strongly stable regimes are further separated in two clusters that show different dynamics of the scales of motions. Mainly in one of them, the analysis suggests that a scale gap separates submesomotions from turbulence, whereas flux variability is more continuous in scale in the other one without a scale gap. In the latter, the turbulence is probably not in equilibrium with the submesoscale motions because the lack of scale separation implies that the turbulent adjustment time may not be short compared to the time scale of the submesomotions.