In this dissertation, the local similarity scaling approach was examined based on the multi-level measurements of atmospheric turbulence in the wintertime (December 2008 February 2009) stable atmospheric boundary layer (SBL) established over a heterogeneous surface influenced by mixed agricultural, industrial and forest surfaces. The 62 m tower (levels 20, 32, 40, 55 and 62 m above ground) was situated in the middle of some 120 m × 480 m area of hc = 18 m high walnut trees. The heterogeneity of the surface was characterized by spatial variability of both roughness and topography. In a first step local similarity theory in terms of flux-variance and flux-gradient relationships was investigated. Nieuwstadts local scaling approach was found to be suitable for the representation of all three wind velocity components. The roughness sublayer (RSL) influenced wind variances, and consequently the turbulent kinetic energy (TKE) and correlation coefficients at the lowest measurement level, but not the wind shear profile. After removing data points associated with the flux Richardson number (Rf) greater than 0.25, the observations support the classical linear expressions for the dimensionless wind shear ([phi]m) even over inhomogeneous terrain. Leveling-off of [phi]m at higher values of stability parameter was found to be a result of the large number of data characterized by small-scale turbulence (Rf > 0.25). Deviations from linear expressions were shown to be mainly due to small-scale turbulence rather than due to the surface heterogeneities, supporting the universality of the linear relationship. Additionally, the flux-gradient dependence on stability did not show different behavior for different wind regimes, indicating that the stability parameter is a sufficient predictor for flux-gradient relationships. Data followed the local z-less scaling for [phi]m when the prerequisite Rf[less than or equal to]0.25 was imposed. Further investigations focused on the combined influence of the RSL found above tall vegetation and the internal boundary layer (IBL) on the turbulence spectral characteristics and TKE budget. The traditional surface layer scaling was tested against the canopy scaling, which is generally valid for the RSL. It was found that canopy scaling can be successfully applied even within the transition layer. For the present complex site local isotropy was not found. Vertical velocity spectra were smaller than horizontal spectra. Similarly, dissipation rates ([epsilon]) determined only from vertical spectra were smaller than [epsilon] estimates based on horizontal components. Therefore, it was necessary to normalize vertical wind speed spectra with [phi][epsilon]w in order to get good correspondence with the Kansas spectral models. Extending the analysis to the Olesen approach, applied for the first time to the SBL over heterogeneous terrain, normalized spectra collapsed to one single curve. Finally, analyzing the budget terms of the TKE equation, non-equilibrium conditions were found. The non-local dynamics are considered to be the main reason for the observed imbalance of TKE in the transition layer as well as for the observed breakdown of z-less regime in the strongly stable conditions above heterogeneous surface. In the RSL, the turbulent transport of TKE above vegetated canopies is considered to be the main cause of the observed TKE imbalance in the neutral conditions. A less systematic behavior of the residual term was observed indicating that the advection term has more pronounced influence on the RSL than the upper levels.