Abstract Turbulence anisotropy was recently integrated into Monin-Obukhov Similarity Theory (MOST), extending its applicability to complex terrain and diverse surface conditions. Understanding which processes drive anisotropy over a variety of surfaces and stability conditions still remains a challenge. This study therefore employed random forest models trained on measurement data from both flat and complex terrain and including upstream terrain features, to understand the drivers of turbulence anisotropy. Two approaches were compared: using dimensional variables directly or employing non-dimensional groups as model input. To address correlation among features, we developed a new feature selection method, Recursive Effect Elimination. Finally, interpretability methods were used to identify the most influential variables. Contrary to expectations, variables directly related to terrain influence were not found to significantly impact turbulence anisotropy. Instead, non-dimensional groups of common turbulence length, time and velocity scales proved more robust than dimensional variables in isolating anisotropy drivers, enhancing model performance over complex terrain and reducing location dependence. A ratio of integral and turbulence memory length scales was found to correlate well with turbulence anisotropy in both daytime and nighttime conditions, both over flat and complex terrain. During the day, a refined stability parameter incorporating both the surface and mixed layer scaling emerged as the dominant driver of anisotropy, while at night, parameters related to rapid distortion were strong predictors.

Abstract Urban canopy models (UCMs) are routinely used to diagnose or predict the temporal and spatial variations of urban surface-atmosphere exchanges and associated phenomena. Traditionally, UCM evaluations rely on in-situ measurements, which are inherently local and encompass a wide range of physics that are challenging to quantify comprehensively. This study assesses the strengths and weaknesses of a UCM, Urban Tethys-Chloris (UT&C), by comparing its predictions against highly controlled large eddy simulations (LESs) and observations in Phoenix, Arizona. Simulations are performed over an idealized urban geometry for ten clear sky days. Due to the inability of the UCM to accurately account for buoyancy-driven transport mechanisms, surface temperature errors for the considered days can be significant when compared against LES. A key limitation in the UCM resistance parameterization is the inability to capture the stability-dependent variations in vertical heat flux, resulting in discrepancies with the LES results. Further, while the UCM captures the primary cooling effect of radiative shading by street trees, it fails to fully represent the enhanced tree-induced turbulent heat transfer beneath the foliage, thereby underestimating the net cooling impact of trees.

Abstract Describing the turbulent mixing of heat in the stable boundary layer (SBL) has been a long-standing difficulty for similarity theory. At three sites impacted by topography, we investigated the connection between turbulent mixing of heat, the thermal structure of the near-surface SBL using Distributed Temperature Sensing, and the universal decoupling parameter, which describes the degree of vertical coupling for turbulent eddies. Three categories of thermal structures were found: logarithmic, sublayered, and quasi-logarithmic profiles. The logarithmic type is mostly associated with vertically-coupled turbulence but exists for a range of stability and vertical coupling values, the sublayered types are almost never well-coupled, and the quasi-logarithmic SBL type exhibits a mixed behavior between logarithmic and sublayered. Existing similarity scaling relations are shown to be a consequence of aggregating across these SBL types and degree of vertical coupling and, critically, none of the existing similarity scaling relations are physically consistent with the profile types or degree of vertical coupling. Several other frameworks of the SBL are found to be a similar result of aggregating across these SBL types. Similarly, methods for selecting data consistent with similarity theory are only partially successful in distinguishing between sublayered, uncoupled and logarithmic, coupled cases. Finally, we show that the universal decoupling parameter may be a more appropriate choice for scaling the turbulent mixing of heat in the SBL than the non-dimensional temperature gradient as it better encodes the physics driving the turbulent mixing processes and has a more robust scaling relationship, without the problem of self-correlation.