Comparing LES, Lab and Empirical Representations of a Neutral Flow over Canopy

Atmospheric flows over vegetation canopy are frequently investigated using widely varying methods, but the similarity between the different representations is tested to a lesser degree. Here, we will present a comparison between the representations of a neutral atmospheric flow over vegetation canopy by two large-eddy simulation models, wind tunnel data and recently updated empirical flux-gradient relationships. A neutral case was selected for the highest anticipated differences, due to the absence of production or destruction of turbulence by buoyancy.

Special attention is devoted to the dynamics in the roughness sublayer above the canopy layer, where turbulence is strongest. We demonstrate that the flow properties are generally consistent between the different datasets and models, thus bolstering confidence in the individual independent representations. We can explain the remaining differences by treating the divergent assumptions that are applied.

Systematic sensitivity analyses with the Dutch Atmospheric Large-Eddy Simulation model further demonstrate that the statistics of the resulting atmospheric flow, normalized by the friction velocity, are independent of the background wind velocity if, but only if, the evaluated points in time are scaled inversely proportional with that velocity. Another aspect is related to the one-sided plant-area density distribution (PAD). The PAD is shown to affect the numerically resolved flow not only through its general configuration, but through the transition from canopy to unobstructed air as well. Sharp transitions can induce anomalous spatial fluctuations in the wind velocity statistics, which can be fully suppressed by spreading the transition over four steps. Finer vertical resolutions only serve to reduce the magnitude of these fluctuations, but do not prevent them. To capture the general dynamics of the flow, a resolution of 10 % of the canopy height is found to suffice, while a finer resolution still improves the representation of the TKE. However, if only the vertical resolution is increased, keeping the original resolution in the horizontal directions, the representation of the turbulence flow does not improve.

To complete the evaluation, quadrant analyses, based on wind velocity, are performed on the simulation data. A major finding is that momentum transport is dominated by the mean velocity components within each quadrant. As a result, one can satisfactorily estimate the total vertical turbulent momentum flux without considering the fluctuations in the wind components within those quadrants.