Large-eddy simulations of surface-induced turbulence and its implications to the interpretation of eddy-covariance measurements in heterogeneous landscapes

Abstract:

We used the RAMS-based Forest Large Eddy Simulations (RAFLES) to evaluate the sensitivity of eddy covariance measurements to land-surface discontinuity. While the sensitivity of eddy covariance measurements to surface heterogeneity is well known, it is, in most cases, no feasible to restrict measurements only to sites where the surface include undisturbed and homogeneous land cover over vast distances around the observation tower. The common approach to handle surface heterogeneity is to use a footprint model and reject observations obtained while the source of observed signal is from a mixture of land-use types, and maintain only measurements where the signal originates mostly from the land-use type of interest.

We simulated two scenarios - measurements of fluxes from a small forest-surrounded lake, and measurements near a forest edge. These are two very common scenarios where measurements are bound to be affected by heterogeneity – measurements in small lakes, will, by definition, be in some non-negligible proximity or the lake edge; forest edges are common in any forest, near the forest patch edge but also around disturbed patches and forest gaps.

We identify regions where the surface heterogeneity is creating persistent updraft or downdraft. A non-zero mean vertical wind is typically neglected in eddy-covariance measurements. We find that these circulations lead to both vertical and horizontal advection that cannot be easily measured by a single eddy-covariance tower. We identify downwind effects, which are well known, but also quantify the upwind effects. We find that surface-induced circulations may affect the flux measured from a tower up to several canopy heights ahead of the discontinuity.

We used the High-resolution Volatile Organic Compound Atmospheric Chemistry in Canopies (Hi-VACC) model to determine the actual measurement footprints throughout the RAFLES domain. We estimated the land-cover type distribution of the source signal at different virtual tower locations using a standard footprint model, and by explicit advection/diffusion simulations of a virtual trace scalar with Hi-VACC. We use the deviation between these two estimates to evaluate which locations are the most reliable for flux measurements and for footprint-based filtering.