Laboratory for Atmospheric Research

Micrometeorology Research Group


Terrain-induced flows are often highly intermittent, three-dimensional (3-D), and localized in nature as indicated by previous studies from analytical and numerical models, field campaigns, laboratory experiments, and numerical models over idealized hills with neutrally stratified background flows. Compared with flows over flat, forested terrain, flows over forested hills are characterized by topography-induced horizontal pressure perturbations, which distort the mean flows and generates turbulent eddies with different scales. The presence of a canopy generally modifies flow dynamics through altering the no-slip lower boundary condition and introducing the influence of canopy elements on turbulence generation (e.g., pressure and form drag, wake production, spectral short-circuiting of energy cascade, coupling of turbulence between above and within the canopy, etc). Our research aims to improve the Weather Research and Forecasting Model (WRF) Large Eddy Simulation (LES) for simulating canopy flows over highly complex terrain. Most of our simulations were performed on Yellowstone HPC.

To accomplish this objective:
      1. We have incorporated the immersed boundary method (IBM) into WRF-LES to eliminate the restrictions of WRF-LES simulations on gentle terrain , thus enabling WRF-LES to deal with high-resolution simulations over highly complex topography.
      2. We are developing a multi-layer canopy module in WRF-LES to provide vertical variability in sources/sinks of momentum, heat, H2O, and CO2 inside the canopy.

We will construct 3-D mean/turbulent flows and spatiotemporal patterns of the total fluxes, sources/sinks, horizontal/vertical advections, and flux divergence of momentum and scalars (i.e., all terms in the conservation equations for momentum and scalars) under neutral, unstable, and stable conditions. These spatiotemporal patterns of different quantities parameters will enable us to quantify the relative contributions of different mechanisms to momentum and scalar transfer.

WRF-LES-IBM system

Our group has incorporated and tested three IBM schemes into WRF-LES over arbitrary complex surface boundary conditions. Since the released Sub-Grid Scale (SGS) models in WRF are all constant-coefficient eddy viscosity models, they do not work well with complex situations. We also incorporated an advanced SGS model (the Lagrangian-Averaged Scale-Dependent Smagorinsky model, LASD) into WRF-LES. The precursor simulation method is used to provide realistic turbulent inflow for complex terrain simulations.

New features our group has added into WRF-LES include:

  • Three IBM schemes: Linear-IBM, Log-IBM and Stress-IBM
  • The Lagrangian-Averaged Scale-Dependent Smagorinsky model
  • Realistic turbulent inflow conditions

  • Our WRF-LES-IBM model can be used to simulate flows around very complex terrain. The surface is described by Standard Triangulation Language (STL) file format. A python script is used to convert regular gird terrain data into the STL format.

    Extreme complex surface (stanford bunny) test

    Figure 1. We have tested our WRF-LES-IBM system to show the ability of our surface algorithm to handle extreme surface situations. Linear-IBM scheme is used.

    City environment flow simulation

    Figure 2. Flows around building-like blocks are simulated with our WRF-LES-IBM system. It shows the potential of our WRF-LES-IBM system in simulating air flows in street canyons and pollutant dispersion. Our blocks are generated from an online STL editor: We generated this animation by ParaView and posted it on Youtube.

    2-D ideal hill simulation

    Figure 3. Comparisons of steamwise velocities between our WRF-LES-IBM system and the original WRF-LES over a 2D hill. The SGS model is the constant coefficient Smagorinsky model, the horizontal grid resolution is 20 m.

    Figure 4. Comparisons of mean streamwise velocities between our WRF-LES-IBM system and the original WRF-LES over a 2D hill. Stress-IBM performs best as compared with the original WRF, although it predicts smaller velocity in the leeward side.

    Validation over the Bolund hill

    The Bolund hill is 130 m long (east-west axis) and 75 m wide (north-south axis), with a maximum height of 11.7 m. The Bolund Hill experiment was conducted by Technical University of Denmark (DTU) from December 2008 to February 2009. This project was a combined measurement and modeling effort related to air flow patterns in complex terrain. Read more about the Bolund Hill Experiment. Our team has used our WRF-LES-IBM system to simulate air flow patterns over the Bolund hill and compared our results with the observations.

    Figure 5. Animation of streamwise velocity. Wind direction is 270 degrees. The horizontal resolution is 1 m. Our WRF-LES-IBM with Stress-IBM and LASD modules is used in this simulation. We generated this animation by ParaView and posted it on Youtube .

    Comparisons with observations

    We compared modeled speed-up with observations. All of these are averaged results. The default grid resolutions are dx = 2 m and dz = 0.5 m, unless otherwise stated. The results are for 270 degree case.

    Figure 6. The comparison shows that Stress-IBM with LASD SGS model gives the best prediction. When the grid resolution increases, the prediction also improves. The improvements of higher resolution are mainly in the areas where separation behavior happens (e.g. in the lee side of the hill).

    Multi-layer canopy module

    We are developing a multi-layer canopy module in WRF-LES-IBM system to provide vertical variability in sources/sinks of momentum, heat, H2O, and CO2 inside the canopy. The model is mainly based on CLM4.5. It includes a multi-layer two-stream solar radiative transfer module, a multi-layer long wave radiative module, a multi-layer photosynthesis module (Ball-Berry model), and a soil temperature and moisture module.

    Figure 7. Diagrams for our multi-layer canopy module.