LES of a three-dimensionally resolved forest using terrestrial laser scanning data

Within the field of meteorology the atmospheric boundary layer (ABL) is an important object of investigation for researchers. Canopies and the highly turbulent region above them, the roughness sublayer, have a strong influence on the ABL. The structure of a tree with its trunk, branches and leafs and the ability of changing its shape under different wind conditions causes a very complex turbulent flow field containing many different scales. Therefore most of the research on canopy flows assumes a homogeneous forest and neglects the details and dynamic behavior of trees. Sites for field experiments to measure those flow fields and the influence of inhomogeneities (cutouts and clearings) on advective fluxes and source area distributions, are often chosen to appear as homogeneous as possible. However, even under these conditions, airborne laser scanning data shows a strong variation of the vegetation.

Such a field site is operated by the Institute of Hydrology and Meteorology of the Technische Universität Dresden since 1959. Located about 25km southwest of the city of Dresden (Germany) it has been part of various European projects, e.g. EuroFlux and CarbonEurope. Details about the investigated domain, which is aligned according to the predominant wind direction of West to East and includes a clearing of 50m x 90m could be found e.g. in Gruenwald and Bernhofer (2007). The velocity profiles recorded with up to 25 ultra sonic anemometers on four measurement towers within two measurement campaigns (June 2007 to May 2009) give detailed insight into the flow field at various positions (Queck et al., 2012). For the analysis of a measurement campaign and to improve dispersion modeling near and within tall vegetation the knowledge of the three-dimensional turbulent wind field within the whole domain would be useful.

Numerical simulations are at present the only way to achieve the desired velocity field, and due to the huge hardware requirements of direct numerical simulations, only Large-Eddy simulations (LES) are capable of this task. LES solves the large scales directly and uses a subgrid-scale model (SGS) for the unresolved scales. The commonly used SGS model for flows through canopies is based on the work of Deardorff (1980) and Shaw and Schumann (1992). The dynamic tree behavior is so far neglected and the plant details (branches, trunk, leafs) are merged into an aerodynamic resistance by using a plant area density (PAD) and a drag coefficient. Within many recent publications the PAD is assumed as a vertical generic profile leading to a homogeneous forest, e.g. Finnigan (2009) or with forest edges e.g. Dupont et al. (2011). Bohrer et al. (2009) presented the first LES using a three-dimensional forest generated by their Virtual-Canopy Generator.

Recently, Schlegel et al. (2012) used the method of terrestrial laser scanning (TLS) to derive a three-dimensional plant area density distribution of 300m x 160m based on a point cloud of approx. 50 million points (Bienert et al., 2010). Due to the poor quality of the data towards the boundaries of the scanned domain the TLS data was restricted to an area of 191m x 30m and a spanwise averaging was performed, leading to a two-dimensional PAD. The remaining forest within the domain of 760m x 380m x 210m was assumed to be homogeneous. Nevertheless, the study revealed a significant influence of small-scale plant distribution on the mean velocity field as well as on turbulence data. The varying PAD generates a complex pattern of sustained up and downward motions inside and above the forest stand.

The proposed contribution for the 20th Symposium on Boundary Layers and Turbulence will extend our previous study by using a realistic, three-dimensional representation of the same field site for a computational domain of 600m x 600m x 240m. The results of the computations using a spatial resolution of 2m and approx. 11 million grid points will be presented. The vegetation will be represented by a PAD with 1m spatial resolution within an area of 326m x 170m. Obtained by TLS, this PAD is based on a point cloud of 150 million points. The remaining forest is generated with the Virtual Canopy Generator of Bohrer et al. (2007) using a a crown model gained by airborne laser scanning, a forest patch map of 46 patches and two generic plant area density profiles.

References:

A. Bienert, R. Queck, A. Schmidt, Ch. Bernhofer and H.-G. Maas, Remote Sensing and Spatial Information Sciences 38, 92-97 (2010).

G. Bohrer, M. Wolosin, R. Brady and R. Avissar, Tellus B 59, 566-576, (2007).

G. Bohrer, G. Katul, R. Walko and R. Avissar, Boundary-Layer Meteorology 132, 351-382, (2009). J. Deardorff, Boundary-Layer Meteorology 18, 495-527, (1980).

S. Dupont, J.M. Bonnefond, M. Irvine, E. Lamaud and Y. Brunet, Agricultural and Forest Meteorology 151, 328-344, (2011).

J. Finnigan, R. Shaw and E. Patton, Journal of Fluid Mechanics 637, 387-424, (2009).

T. Gruenwald and Ch. Bernhofer, Tellus B 59, 387-396, (2007).

F. Schlegel, J. Stiller, A. Bienert, H.-G. Maas, R. Queck and Ch. Bernhofer, Boundary-Layer Meteorology 142, 223-243, (2012).

R. Shaw and U. Schumann, Boundary-Layer Meteorology 61, 47-64, (1992).

R. Queck, A. Bienert, H.-G. Maas, S. Harmansa, V. Goldberg and Ch. Bernhofer, European Journal of Forest Research 131, 165-176, (2012).