Different aspects of the stable boundary‐layer structure are contrasted between the very stable and the weakly stable regimes from a new point of view. This study finds a limit wind speed, referred to as the crossover threshold, when the average vertical gradient of the turbulent kinetic energy switches sign at all observational levels. When the wind speed exceeds this transition, the entire stable boundary layer becomes vertically fully coupled. Consequently the very stable boundary layer in this study is considered as a decoupled regime, while the weakly stable state is referred to as a coupled regime. It is shown that the vertical profiles of other quantities, such as friction velocity, heat flux and thermal gradients are strikingly different between the two coupling states.Decomposition of turbulent kinetic energy and heat flux into temporal scales indicates overlapping of non‐turbulent sub‐mesoscale flow with turbulence in the decoupled case, while there is a clearer scale distinction between the two types of motions when coupling takes place. The turbulent kinetic energy budget is dominated by dissipation and shear production in both coupling states. However, the relative importance of the buoyant destruction term is shown to be appreciably larger in the decoupled regime. In the heat flux budget equation, buoyant destruction is larger in magnitude than production by the thermal gradient in the decoupled case, but not when there is full coupling. These results indicate that the surface heat flux plays a major role in controlling the stable boundary‐layer state, as previously proposed. For the entire dataset, the frequency distributions of turbulence quantities near the surface are shown to be bimodal. The two modes are associated with the two coupling states, each well described by independent log‐normal distributions.

Previous observational studies in the stable boundary layer diverge appreciably on the values of dimensionless ratios between turbulence-related quantities and on their stability dependence. In the present study, the hypothesis that such variability is caused by the influence of locally dependent nonturbulent processes, referred to as submeso, is tested and confirmed. This is done using six datasets collected at sites with different surface coverage. The time-scale dependence of wind components and temperature fluctuations is presented using the multiresolution decomposition, which allows the identification of the turbulence and submeso contributions to spectra and cospectra. In the submeso range, the spectra of turbulence kinetic energy range increases exponentially with time scale. The exponent decreases with the magnitude of the turbulent fluctuations at a similar manner at all sites. This fact is used to determine the smaller time scale with relevant influence of submeso processes and a ratio that quantifies the relative importance of such nonturbulent processes with respect to turbulence. Based on that, values for the local stability parameter that are unaffected by nonturbulent processes are found. It is shown that the dimensionless ratios do not usually converge to a given value as the time scale increases and that it is as a consequence of the locally dependent submeso influence. The ratios and their stability dependence are determined at the time scales with least influence of nonturbulent processes, but significant site-to-site variability persists. Combining all datasets, expressions for the dependence of the dimensionless ratios on the local stability parameter that minimize the role of the submeso contribution are proposed.