Addressing Human Exposures to Air Pollutants Around Buildings in Urban Areas with Computational Fluid Dynamics (CFD) Models

AMS Third Symposium on the Urban Environ merit, 14-18 August 2000T Davis GA
7.7 ADDRESSING HUMAN EXPOSURES TO AIR POLLUTANTS AROUND BUILDINGS IN
URBAN AREAS WITH COMPUTATIONAL FLUID DYNAMICS {CFD} MODELS
Alan Hirber*', Mart BolstacP, Mattiew Freeman*. Samir Ridasr Eric Blstf, and Karl Kuehfert2,
(1) NOAA, ARL, AS MO, RTF1, NC on assignment to the US EPA National Exposure Research Laboratory
(2) U.S. EPA Sdeniilic Visualization Certler, Lockheed-Martin, NTP, NC
(3) Fluent, Inc., Lebanon, NH
1.0 INrrJRODOCttON
Computational Fluid Dynamics (CFD) simulations
provide a number of unique opportunities for expanding
and Improving capabilities for nw^eUng exposures to
environmental poliutanls, The US Environmental
Protection Agency's National Exposure Research
Laboratory (NEHL) has been conducting cooperative
research with Ffuent, Iric to examine arid evaluate (he
application of CFD models for simulating air pollution
along the pathway from source to human exposures.
The basic framework of population-based human
exposure models separates a person's day into time
spent in a aeries of exposure microenvironments. The
environmental concentration in these exposure
micfoenviroriments is weighted by the time-spent In
each mteroeflvlroransnt to provide a total daily exposure.
The detailed spatial resolution of environmental pollution-
concemrailens that is possible tram CFD simulations
can provide important information that is not available
from a single point measurement. There are multiple
potential roles for CFD simulations in supporting human
exposure studies which will he presented outside of this
brief abstract. In tte study, weare examining (n detail
the utfcan buildings and roadway microsrwironrnents of
human exposure to ambient air pollutants. This detract
can present onfy a summaiy of some of the issues we
are examining and a tew examples of progress,
2.0 CFD MODEL SETUP
The commercial CFD software FLUENT
fhflpV/www.fl uentcomt along with its model geometry
and mesh generating software (GAMBIT) is beting used,
Mod&? Gwmatry and Mesh Set-up in GAMBIT
There are key elements with many options, which must
be set up before GAMBIT calculates a mesh, The tnoder
domain boundary conditions (including atmospheric
reality) are critical, The inlet and outlet positions relative
to the study focus area can be critical. The grid
resolution has to take into consideration the ralio
between large and small scales of the flow. With a wide
range of length scales a structured meshing becomes
Jess flexible, more time consuming and memory
intensive. An unstructured approach can reduce
meshing time dramatically because a laige portion of
the process is automated- Unstructured grids provide
flexibility in setting up a dense mesh where it
is needed and allow for local grid adaptation,
Consideration of the temporal scales of interest is also
critical.
Model Set-up In FLUENT
There are key elements wslh many options, which
mgst be set before FLUENT calculates a soiuHoFt- The
coupled solvers {explicit and implicit} are recommended
if a strong inter-dependence exists between denisfty,
enejfy, momentum, and/or species {i.e., finite-rate
'Corresponding author address: Alan Huber, US EPA,
MD-5G. RTF. NC 27711; e-mail: btiber.ala n@cpa.gov
reaction modeled flows). The segregated (implicit) solver
is preferred in all other eases because of lower memory
requirements and more flexibility In sofution procedure.
For the usual air pollution application the segregated
solver is the model Of Choice, The flow is
incompressible and turbulent and the pollutant transport
equation can be resolved al the post-processing level
because the dispersion does rot impact the flow. The
turbulence model is chosen according So the compiesdty
of the geometry and the iow. FLUENT offers a variety of
models ranging from a one-equation model to a Large
Eddy Simulation {LES), Recent Reynolds-Averaged
Navier-Stokes (RAWS) mcdefe seam to significantly
fmpiwe accuracy of the numerical section of turbulent
flows around buildings but have not been completely
satisfactory, LES remains the desired model of choice
and will play an Increasingly important role in ihe future
as computational capacities continue to increase.
3,0 EXAMPLES
The FLUENT software is run using 2 processors on a
S<3r Onyjf2 workstation which is 100% available to this
project and using 10-32 processor on the EPA National
EnvironmentaE Supereompuftig Center (MESC) CRAY
T3E as it Is available. The project is evaluating
fundamental features ot the software by comparing
model simulation results wfth data torn wind tunnel
studios. The set up of a range ol atmospheric boundary
layer characteristics is being studied, Evacuation with
wind tunnel measurement data is a critical step. We are
particularly Interested id examining ihe improvements
resulting from more refined simulations arid at what
costs. Our goal is to be able to apply CFD simulations
for an urban area such as Manhattan. A model for
Manhattan has been set up (Figure 1) and we are
comparing the CFD simulations of CO concentrations
from roadway emissions to four monitoring stations. A
single simulation has so far been limited to sheeting a
region of 2,5 million cells 25% o( the domain). Figure
2 shows the diurnal cyeie of CO which is dominated by
atriomobMe traffic. Morning and afternoon peak
concentrations occur duiing peak traffic perfects which is
absent during weekend mornings. Meteorotagfcai
conditions effect daily vsriabPily. The Nghast and lowest
concentrations arise between two sites that are closest,
This example demonstrated why a sinple point value is
not very reliable for representing a large area of
potential exposure. We are working to explain these
concentrators. Figure 3 presents an etsarnpte vertical
plane o) the wind vectors. Notice the strong upward
winds on the Its side of tho Empire Slate Building and
the multiple scales of interaction between the buildings
and street canyons which con-tain the pollutant sources.
The simulated wind patterns are being compared with
measuremenls in NYC Central Park and surrounding
sites, Oor progress is encouraging to us. Much more
tfeiail and conclusions will follow. Evaluation wiiji other
wind tunnel and field data is ongoing and more is
needed [coilaboralors am welcomed].
1

-------
Figure 1. Manhattan CFD model domain with CO sites (81,56,92,62) and meteorological
tower (MT) location.
3.5
cfi
o
3.0
2.5
¦fis 2.0
O
o
1.5
1.0
Q.5
Q.O
2 4 6 8 10 12 14 16 18 20 22 24- 2 4 6 8 10 121416 18 202224
Weekday	Weekend
Figure 2. Annual-averaged (1996-98) diurnal cycle of Carbon Monoxide at four sites.
Mean wind from left to right
Figure 3. Example vertical plane of wind field (vector direction and speed) for a CFD simulation
from the Manhattan domain (largest building is the Empire State Building).
DisclaimenTNs paper has been reviewed In accordance with the U.S. Environmental Protection Agency's peer and administrative review policies. Mention of
trade names or commercial products does not constitute endorsement or recommendation for use.
2

-------
MEKL-RTP-AMD-OO—15* TECHNICAL REPORT DATA i
,l. p.ipoar NC.
SPA/600/A-00/CJ7B

5. St
¦J TITI..E A WD SCFariCLK
Addressing Environmental Chalknwex wiih Computaliooaf Fluid Dynamics
:• 35FORT SATE
*>.PtR FORMING OStiAMZ.ATTON CODE
7. AL'THORfS)
lAlan H. Hober* :Maik Bolstad, ¦Mathew Freeman, ¦Samir Rid a, 'Lirie S. liishr
and 'iCarl H. Kuchlcrl
S PRRFORM WC CtRCA.Nl ZATIO* KtKtttT NO.
9, PERFORMING ORGANIZATION N/iKF; r\R> ADF.WESS
'Sarae as Block 12
2l'SEAP Scientific Visualization Center
Lockheed-Martin, RTF, NC
"'Fluent Inc.
10 Cavendish Court
Lebapos,KIH 03766
10. PROGRAM ELEMENT NO.
1 i. CONTRACTOR*NT NO
1 a. SPONSORS AGENCY NA?»fl'. A.NfJ AfJDRFSfi
National Exposure 'Research t'jsbwarory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
1X TYPF. OF R I-PORT AMD PFRlOD COVbKKO
14. SPONSOR'NC AGENCY C0OE
i 5. SUPPLEMENTARY NOTCS
16. ABSTRACT
Com pactional Fi^ki Dynamics fCFD) simulations provide a number of unique opportunities for expanding arid improving
capabilities fur madding exposures to environmental pollutants. The US Enviromncn.cal Pnncction Agency's Narioewl
Exposure Rcwarch Laboratory e spent ia a scries of exposure
:nicroc3ivitx;nTFje;UH. The en vim; imein&t: cone wlra lion in these exposure MieroctivirotirnenlK sjs weighted by She time-spent in
each inicroerivironmenf to provide 10 a total daily exposure. The detailed spasijil resolurior. of environ mentai pollution
concentrations thai is possible froo? CFD simulations can provide important information that is no? available from a singte poir
>neasn recent. There are multiple potential roles forCl-D simulations in supposing tinman exposure studies which »iJJ be
presented outside of ihis brief abstract. In this study, we are examining in detail ihc urban buildings and roadway
microenvironmtfnts of human expasisn: lo ambient air pollutants. This absuacl can prcscm only a summary of some of the
issues v-s are exacromni and a few examples of prepress.
i -i ki-:y words and document analysis
a. JiKSCKU'iORi
WC!EMT[f!Ert.S- DKbN liNDkt> I'l-JRMS
c;COS ATI

IS. D]XTFUBU7SON 5'E'ATr.MKN'F
RELEASE TO PL'BLIC

IS. SLCURII V tXASS rVTu'r Hepon)
UNCLASSIFIED
21. NO. OKFAGES
2

20. SECEfRTTY CI .ASS fTT.;; Pug ,•)
UNCLASSIFIED
11. PRICE
EPA-2120

-------