Measurements and Large Eddy Simulations of Plume Dispersion in an Urban Boundary Layer

1. Implement, test, and use new generation subgrid-scale models for simulating pollutant transport in urban environments in the JHU – Large Eddy Simulation (LES) code.
2. Measure aerosol profiles in the atmospheric boundary layer with the JHU lidar in eastern Baltimore collocated with point aerosol sensors to identify pathways and sources of aerosols.

To address potential exposure pathways in urban environments from airborne particles both computational simulation tools and instruments are being developed and deployed. Air pollution is affected critically by wind that transports pollutants from the emitter to other locations. Computer simulations of air movement and pollutant transport in the urban environments are especially challenging due to the complex ground topology typically found in cities. The goal of this year’s research has been to implement and test a new-generation physical model of turbulence to improve on the state-of-the-art of computer simulations of flow and transport within urban environments and to undertake a series of field observations on airborne particles in southeastern Baltimore using a suite of air quality instruments.  Measurements are taken at this location, in part, since a prevailing concept is that the buildings of downtown direct particulates from the hazardous stacks in south Baltimore toward the east neighborhoods. 
     The major fundamental issue in developing improved computer simulations is the parameterization of unresolved small-scale turbulent motions (vortices and eddies smaller than the computer mesh-size). Classical parameterizations rely on adjustable parameters (e.g. the so-called Smagorinsky coefficient, c_s) that can only be tuned in a fairly ad-hoc fashion. This state-of-affairs has greatly limited the predictive powers of computer modeling of atmospheric turbulent phenomena. A major breakthrough occurred in the 90’s in the field of computer simulation of turbulent flows, when Germano et al. (1991) recognized that one could use the turbulent eddy dynamics that are being computed in order to determine numerical values of unknown model parameters. These same parameters could then be used for modeling the unknown scales of motion, under the assumption of scale-invariance. This new approach, the so-called “dynamic model”, has already been applied with success to a number of fairly simple flow conditions, mainly in engineering flow applications. Our group has been involved in generalizing this new paradigm to atmospheric flows and our prior work has shown how to relax the basic assumption of scale invariance (Porté-Agel, Meneveau & Parlange 2000). The approach is based on a statistical analysis of the resolved motions during the computations, i.e. the resulting fields must be averaged over directions of statistical homogeneity.
     For applications to urban environments, the geometries are too complex to allow finding such directions of statistical homogeneity; in fact such directions rarely exist. Hence, a more general form of the dynamic model is needed. The so-called Lagrangian dynamic model (Meneveau, Lund & Cabot, 1996) provides a workable alternative, since it evaluates the statistical averages “on the fly” by following fluid particles during the simulation. The parameter thus obtained is a function of position and time. We will denote it as c_s(x,t). The main advantages of the Lagrangian approach are: applicability to flows with no homogeneous directions, preservation of Galilean invariance, computation of a local c_s at every point, ability to handle complex geometries and unsteady flows. However, while quite successful in complex engineering flows, this model has not yet been applied and tested in the context of atmospheric flows where the length-scales are quite different. One of the objectives of our research has been to implement, test, and use this new model for simulating pollutant transport in urban environments.
      Our modeling progress to date has been (i) to implement the scale-dependent Lagrangian dynamic model in the JHU LES code, (ii) to perform validation tests on flat surfaces (the tests were successful), (iii) in the context of another project (funded by NSF), to simulate flow over patches with varying roughness scales (those tests have demonstrated the capability of the model to capture local variations in coefficient), and (iv) in the context of this EPA grant, to implement the model in wind-flow over a building topology. For this purpose the JHU-LES code was modified to allow prescription of complex-geometry boundary conditions (to represent buildings, using the so-called embedded boundary method), and the Lagrangian dynamic model was implemented in conjunction with these new boundary conditions.
     Preliminary results for flow around a representative building shape  (see Fig. 1) confirm that physically realistic flow patterns are obtained (including the existence of the characteristic horse-shoe vortex around the base of the building, the separated wake, etc..). In terms of the ability of the dynamic model to predict spatially-varying values of the coefficient, in Fig. 1 we present spatial distributions of the predicted coefficient field. Interestingly, the coefficient increases almost five-fold in regions of rapid straining on the sides of the building where the flow is being deflected in an irrotational fashion, whereas the coefficient is decreased along the shear-layers downstream of the building. There are good reasons to believe that these trends are physically realistic. As comparison, simulations with the standard Smagorinsky model would impose a spatially uniform value of near 0.03 thus damping the shear layers too much and possibly not damping enough in the high-strain regions.

Figure 1

(a)
Lagrangian cs distribution in atmospheric boundary layer flow over single building, along representative horizontal plane.

(b)
Two three-dimensional iso-surfaces of Lagrangian dynamic coefficient showing complex spatial structure

     These preliminary results are highly encouraging. As next steps we will introduce transport equations for concentration of pollutants, test the predictions of flow around building shapes with available data, and generalize to the case of several buildings to approach the level of complexity typically found in urban environments.
     On the experimental side we have improved the capabilities of the JHU lidar that can be used to assess certain features of the LES simulations (e.g. ABL entrainment) and measure the transport patterns of aerosols. In this project, and in the context of another EPA project (Ondov PI – EPA Baltimore Supersite), aerosols and their chemical properties have been measured during the intensive summer 2002 measurement campaigns at the JHU Bay View Hospital to assess sources of particulates in that community.  We have taken advantage of the related nature of the experimental work in using the eastern field site for safety concerns and that the operation and maintenance of a lidar is extremely time consuming as it requires the use of safety spotters at all times.   An example of lidar aerosol profiles obtained during the Canadian forest fire event of July 7 is presented in Figure 2. Boundary layer structures - strong downdrafts – are clearly evident bringing large amounts of upper atmosphere (Canadian smoke) into the urban atmosphere. Additional air quality and meteorological data were taken and will be analyzed to answer questions regarding aerosols pathways in the Baltimore environment as the next steps in this project.

Figure 2

Lidar time series of relative aerosol concentration for a height range z from 250m to 2200m (arbitrary units) demonstrating ABL entrainment.

• To date no peer reviewed papers have been published funded by this project. 
• A presentation on the LES model development was given this summer at the American Meteorological Society meeting: The 15th Symposium on Boundary Layers and Turbulence, Wageningen University, Wageningen, The Netherlands 15-19 July, 2002, Title: LES of the atmospheric boundary layer over heterogeneous surfaces using dynamic lagrangian models, Authors: Elie Bou-Zeid, Marc Parlange, and Charles Meneveau.: Johns Hopkins University, Baltimore, MD 21218.
• Additional presentations will be given at 1. The American Physical Society – Fluid Dynamics Conference in November (Dallas) by Elie Bou-Zeid updating progress on the LES and 2. The American Association for Aerosol Research by Markus Pahlow at the NC October annual meeting to present results on the eastern Baltimore aerosol measurements.

1. Introduce transport equations for concentration of pollutants, test the predictions of flow around building shapes with available data, and generalize to the case of several buildings to approach the level of complexity typically found in urban environments.
2. Analyze the field data collected at the eastern Baltimore site to identify Baltimore city and long distance sources of aerosols and their movement from point sensors and lidar measurements.

• Turbulence research group, Charles Meneveau http://pegasus.me.jhu.edu/~meneveau/

• Environmental Fluids group, Marc Parlange, http://www.jhu.edu/~dogee/mbp/ 

• Johns Hopkins University Center for Environmental and Applied Fluid Mechanics, http://www.jhu.edu/~ceafm/