Preliminary Investigation of Uncombusted Auto Fuel Vapor
Dispersion within a Residential Garage Microenvironment
(U.S.) Environmental Protection Agency, Research Triangle Park, NC
1993

-------
' n«r-w« % nv. 1 ?
EPA/600/A-93/123 1
3
4. title and subtitle
PRELIMINARY INVESTIGATION' OF ITSiCOMBlSTED AtTO KI/E1.
VAPOR DISPERSION WITHIN A RESIDENTIAL GARAGE MICROESVIROVMEVT
a report oati
a PERFORMING ONGANiIATion code
7 AUThORiSI
Aaedine Lansari, John J. Stretcher. Alan H. Huber,
Gtnnaro H. Cracenti, Roy B. Zw« dinger, John W. Duncan
a PE RfORMiNG ORGANI2 AT iqn Rt!*ORr NG
a PERFORMING ORGANIZATION name AND aooress
Human Expoaurt Modeling Branch
Human Expoaurt and Field IU March Divmon
AJtEAL/ORD'XJSEPA
Raacarth Tnangit Part. NC 27711
10 PROGRAM EiEME^t no
ii C6nThAet.fiAlNT Ki&
12 SPONSORING AGENCY NAME AND ADORESS
Aunoapheric Raacarch and Expoaure AiatMnwol Labonior)
Office of Riaearch and Develop m«M
U S Eovironmnul Proucuon Ajetv. \
¦Uaaarch Triangle Park, NC i77l 1
13 TYPE 0* REPORT AND PERiOD COVERED
Extended Abstract
14 SPONSORING AGENC* COOE
EPA/60C/09
IS SUPPLEMENTARY NOTES
IS ABSTRACT
Evaporative emissions from vehicle* in •n attached garage may represent • significant aource of indoor
pollution and human exposure A pilot field study was undertaken to investigate potential w-house dispersion of
evaporative emissions of uacomhusted fuels from a vehicle parked inside an attached garage In a set of experiments
using sulfur hexafluonde (SFj tracer gas. the multiional mass balance model, CONTAM88, was used to predict
interzonal air flow rat&s and SFk concentration distributions within the garage and house. Several experiments were
included to evaluate the effect of meteorology and mechanical mixing mechanisms on the dispersion of automobile
fuel vapor Measurements indicated that approximately three percent of the garage ma*•mum concentration was
measured in a room adjacent to the garage. TK. model successfully predicted garage concentrations under well
mixed conditions, but underpredicted the BX-vSu.ed concentrations within various rooms of the house, in which
mixing was incomplete Muluzonal mass br lance models sucb as COST AM 88 may be useful in approximating
contaminant concentrations at various locations wiUun the bouse.
17	H £ V WOflOS AND DOCUMENT
a OiSCiPTOXS
b CtN'iflf RS Oft N tNOlD HHVS
COSAT 1



ia oisTRia.Tion statement
19 SlCcR'T* Ci.ASi T IS
j , \ ^ c » » -i • ¦ t i
7
20 btCtXil < C. ASS Tn,ip*t-
72 0B CE
EPA Pari" 2270-1 (*•» 4 - 77 > »«[,il5ul [O'''0«.-lOIIOa'l

-------
PRELIMINARY INVESTIGATION OF UNCOMBUSTED AIJTO FUEL
VAPOR DISPERSION WITHIN A RESIDENTIAL GARAGE M1CROENVIRONMENT
Azzedine Ltmiri', John J. Stretcher*3, Alan H. Huber"1,
Gennaro H. Crcscenli"2, Roj B. Zweidinger\ John W. Duncan' •
1) ManTech Environmental Technology Inc.,
P.O. Bo* 12313, RTP, NC 27709
2} Atmospheric Sciences Modeling Division,
Air Resources Lab., NOAA; RTP, NC 27711
3) Mobile Source Emission Research Branch,
AREAL.US EPA, RTP, NC 27711
ABSTRACT
Evaporative emissions frois vehicles in an attached garage may represent * significant source of indoor
pollution and human exposure. A pilot field study was undertaken to investigate potential m-bouse dispersion of
evaporative emissions of uncombusted fuels from a vehicle parked inside an attached gartge. In a set of experiments
using sulfur hexafluoride (SFJ tracer gas, the multifocal mats balance model, CONTAM88, was used to predict
interzonal air flow rates and SF, concentration distributions within the garage and boust. Several experiments were
included to evaluate the effect of meteorology and mechanical mixing mechanisms oo lite dispersion of automobile
fuel vapor. Measurements indicated that approximately three percent of the garage maximum concentration was
measured in a room adjacent to the garage. The model successfully predicted garage concentrations under well
mixed conditions, but underpredicted the measured concentrations within vanous rooms of the bouse, in which
mixing was incomplete. Multizonal mass balance models such as CONTAMB8 may be useful in approximating
contaminant concentrations at various locations within the bouse.
Kr\y.ord index Indoor air, multizonal models, mass balance models, methanol dispersion, attached garage
INTRODUCTION
The introduction of oxygenated auto fuels and fuel additives (alcohols and ethers) into the U.S. motor vehicle
fleet has served to reduce tailpipe emissions of carbon monoxide and total hydrocarbons ' . Tailpipe emissions
represent an obvious source of ecological pollution, however evaporative emissions from vehicles in attached garages
may represent an important source of indoor (i.e. microtnvtroxmeniat) pollution.
In-bouse and attached garage concentration of evaporated (uncombusted) fuel species from an automobile's
fuel system may represent a significant component of total human exposure to these chemicals The use of alcohol
and ether additives increases the fuel vapor pressure, and hence the evaporation rate"'. The magnitude of m-house
concentration of a chemical species depends upoo the emission rate of evaporating fuel, the concentration of the
component species within the liquid fuel, and the air flow rales between garage and house. Measurement of these
critical variables enables the development of predictive models useful in population exposure assessment.
This study examines the potential for dispersion of evaporative emissions from an aulo fuel system into a
residence from an attached garage. A aeries of field experiments were conducted to obtain estimates of in-house
ambient concentration of fuel vapor components resulting from normal automobile use scenarios. A single famil)
borne with attached garage was selected as the test site. Air flo» rates between the garage and vanous zones within
the bouse were measured using SF, tracer gas. A multizonal mass balance model was used to predict the spatial
and temporal contaminant dispersion within the bouse. Previous investigation has shown that multizonal mass
balance models may be useful in designing field study monitoring strategies"'. Modeling result* were compared to
SF( measurements in order to investigate the possibility of using the model to predict methanol and or other
* Or iE£.3r"f't t c. x r,c At rr.rsp '.er j c Resfrarr' ana	Asstss'cr l»t, U.5.
t' v; ronme* t m 1 pre,*, en a c* A^e'.cy

-------
alternative fuel dispersioo in the |ini(t sad throughout the bouse Methanol amissions and concentration
distribution* were eiamiaed 10 a later phase of this pilot study. Analyse* of methanol concentration and supporting
modeling results will be preaeated in subsequent report.
materials and methods
RESIDENCE DESCRIPTION
During the summer of 1992, a house in Raleigh, North Caroliaa was choaea to study ia-fcouae dispersion of
evaporative emissions of uacombusted autotLobilc fuels from a vehicle parked taaide aa attached garage. This stud)
determined quantitative flo* rates between the garage (pollutaat source) and selected rooms of the house The first
atory of this two story house consisted of the master bedroom, a bathroom, a dec, the kitchen aad the dming room
(Fig. 1); the second story of the house includes three bedroom* aad a bathroom. One second story bod room was
located directly above the garage. The total volume of (be garage was 95 m'. The house physical characteristics
were measured during the first day of the study. These measurements were subsequently used 10 model SF,
concentration distribution ia the garage and within the house. The SF» dispersion analysis wai uaed to calculate
inlerzooal air flows and to calculate contaminant concentrations in order to determine the rooms with significant!)
different concentrations.
The multijonal mass balance model uaed u> this invettigation is the National Institute of Standards aad
Technology (N1ST) model, NBSAV15 CONTAM88" developed for die Environmental Protection Agency to
nmulate transient contaminant concentration distribution in buildings. The node! is baaed on the element assembly
approach, which assumes that a building can be represented as a combination of well-mixed woes linked by flow
and kinetic elements (contaminant mass transport and decay). CONTAM88 aolves 1 set of mass balance and flow
equations The mathematical formulation of the contaminant concentration is:
I M ] C * [ Af J " G	(1)
at
where:
C - Vector containing the discrete concentration values
[W] » System m&&<. transport matrii which contains flow rate data
(MJ « System matru which contains mass (volume) data
G m System geaeratioo vector containing kinetics data.
NBSAVIS is a preprocessor to CONTAM88 which allows the idealization of a building through the generation of
a file that describes the building configuration, including indoor and outdoor contaminant sources. Data input to
NBSAVIS is facilitated by e aenes of data entry acreens that allow the user to apecify: interior and eitenor wall
types; interior and extenor doors, windows, open passageways; filters and fans; roots descriptions; and beating,
ventilation, and air conditioning (HVAC) system descriptions NBSAVIS then calculates the interzons'air flo* rates
and system matrices. NBSAVIS was used to build an ideali2atioo of the house. The parameters that *ere measured
ia order to run NBSAVIS are:
. The house physical dimensions (including all widows, doors and other openings)
. The house HVAC system output as well as all the locations of the vents with air flow rates.
. The cootamiaant source information (name, molecular weight, emission rate).
. The location of the source (inside the garage).
. The temperatures of the various rootns in the house.
. The outdoor meteorological conditions (temperature, wind speed and wind direction).
2

-------
EXPERIMENTAL DESIGN
Throe experiments is which SF, wu released 10 the garage and t^acei inside tbe garage and the house were
designed and implemented SF, was chosen because it is non-toxic, cuble with respect 10 chemical reactions, its
removal due to deposition it negligible and it can he measured is very low coo -ect rat ions with high accuracy"
A diluted mixtv.t of SF, in nitrogen (0.362g SF, / m' ) wu prepared ard used id all experiments Sample were
tnAl\zed using gas chromatography 'electron capture detection (GC'cCTij. wi'ji • lower detection linut of 20 parts
per trillion volume (ppt) Measurement accuracy was within 5* Mft»Ajrerjcnt precision (standard deviation) was
within 5% of the measured concentration. During the eotire study t>je ouUloor temperature ranged between 19 C
and 23 C, the wad speed did not exceed 0 5 m i and the wind direction vaned The low wind speeds were mainly
due to shielding of the house and property by trees.
Tbe first experiment was designed to assess the percectag. of contaminant that infiltrates in the ho>is- from
the garage A car was parked in the garage with the front ol the car facing an interior wall of the garage L'i this
experiment five automated sequential synnge samplers'*' were used. Three samplers were placed it. the houie one
in the kitchen, one in the 4en and one in tbe bedroom above the garage Tbe remaining two samplers were located
is (he garage One samp'er was placed on the car roof, one was placed on tbe garage floor, next to the car The
ftmge door wu closed dunng the experiment SF* was released for 20 nun at a rate of 1 liter per minute (I mm)
Tbe total mass of SF, released was 7.23 mg. Tbe SF, source was located next to tbe gasoline tank on the passenger
side of the car A large box fan was used nest to the source to create s well mixed condition Samplers is the
garage were started simultaneously with tbe SF, source At the eod of the SF, release (20 mir<). the three samplers
inside the bouse were activated
Tbe second experiment wss designed to investigate bow quickJy tbe concentration in tbe garage drops after
the garage door it opened In this experiment no car was in tbe garage. Ajr sample* were taken at two vertical
locations ui the gsrage Tbe upper sample location was centered 2 feet from the garage ceiling while the lower
•ample location was centered 2 feet above the floor. The samples were analyzed u real time. SF, was released
id tbe garage for 10 mm at a rate of one I'min for a total mass released of V62 mg Tbe box fan was used to create
a wt'l mixed condition. At the end of the 10 min release time, the fan was turned off and the garage door was
opened
Experiment three was performed to test the well mixed garage assumption. Five syringe samplers were
used in the garage. one on the car roof, one next to the dmer't side, ooe next to the passenger m. e ui front
of the car (hack wall location), and ooe in the back of the car (next to the garafe door) No fan was used, and
sampling time wa& set to 12 min interval; for all the samplers SF, was released for 20 min at s rate of I I mm
The SF, source was located next to the gas tank on the passenger side of the car Tbe garage door was closed
during the experiment
RESULTS AND DISCUSSIONS
Dunng the first experiment. CONTAM8B was used to simulste the dispersion of SF, in the garage and
throughoutthebou.se Good agreement was found between simulated and measured data (Fig 2) Difk-.nces tna>
be due to a lack of complete mixing us the garage There were two samplers ui tbe garage , the one on the passenger
side (labeled LEFTGaRaGE in Fig 2) showed s sharp drop of concentration approximately 30 min after the
experiment tuned This ampler was located close to the kitchen door, therefore the drop in concentration toav
be due to local air exchange Tbe other sampler (labeled TOPGARAGE in Fig 2) agrees better with the model,
which assumes s well mixed condition.
Modeling results consistently underpredicted tbe concern'rat ion in all the rooms in the house (Fig 3). The
highest concentration was measured and modeled in the blcher., followed b\ the den and then the bedroom atxne
the garage Overall, the model predicted a behavior similar to the data However, the model underpredicted 'he
lutcheo and bedroom concentrations by approximate!) 309, and the den concentrations by approximate!) 10 9
Tbe model assumption of well mixed zones was sot satisfied in the house because of the absence of forced tmxir.g
Tbe model predicted that the maximum concentration attained in the kitchen would be 2% of the maximum
concent rati or r the garage, while measurements showed that the maximum relative concentration was approximate!)
3*.
Following the completion of SF, injection, the concentration in the garage began to decre*1*. whiU the
3

-------
concentrations in the kitchen tad the other roc .as continued to nse Tbe time delay (phase lag) between the modeled
maximum concentration in the garage and the 'utchen was approximately IS minutes; between the family room and
the garage the time delay wu approximate!) iO minutes In the bedroom the concentration was still rising SO nun
after emission stopped (Fig 3}. The measured phase lag in concentrations be/ween the various rooms of the house
is also predicted b> the roodd (Eq 1) Conservation of mass (SF,) - within the context of i box model - requirta.
that a loss of mass from one box (e g the gawge) be accounted for in a net increase of mass in tbe other boxes (e g
other rooms of the house, the outdoors)
In the a<*cond experiment, CONTAM88 was used to simulate contaminant build up and dec*) after the
garage door was open Modeling results were in agreement with the rae&suremenU (Fig. 4). After the garage door
was opened the SF, concentration decreased to less than 10* of the initial concentration in 3 minutes Similar
results were found using the decay from the modeling results This decay rate corresponds to approximately 45 air
change- per hour (ACH). During testing the wind speed was leu than 0.2 m's, therefore this high exchange rate
is likely due to mechanical mixing that occurred when the garage door was opened These results we:e supported
by observations during smoke release experiments
During tbe third experiment uteasureine,,were performed to test the well mixed box assumption Model
simulation was used to determine locations u> the garage where the well mixed condition existed Simulation results
agreed with the data collected at the car loof. tiack of the garage, and the passenger side locations The data
collected near the garage door and driver side locations were approximately half of the model predictions Total
mixing occurred between 60 and 90 min after the SF« release stopped (Fig S). These results show that sampler
locations cannot be assumed to measure avenge (well mixed) concentrations, particularly dunng the period of
mniiminmi emission However, for constant sources, a steady Male regime may develop in the garage and house,
resulting in quasi-mixed conditions
CONCLUSIONS
Good agreement was found between modeling and expenmental results in tbe garage when the well mixed
assumption was verified (box fan on) In the rooms adjacent to the garage, the model underpredictad the
concentration of SFt The model did help assess the broad trend of concentration distribution in those rooms
Approximate!) three percent of the maximum garage concentration was measured in the kitchen and I S percent in
the upstair* bedroom These experiments were performed with the door between the kitchen and garage »l»iu
closed Higher in-house relative concentrations can be expected when this connecting door is opened
When mixing was not forced (box fan off) within the garage, the well mixed assumption was not valid at
locations next to the garage door and the kitchen door, but the other locations in the garage showed more thorough
mixing It look 60 to 90 mm for total mixing to occur after the release of SF» stopped Furthermore, the *eli
mixed assumption did not bold during the contaminant release time
Preliminary results showed that multizona) mass balance models such as COSTAM88 ma) be used to
approximate contaminant concentrations within various locations of the house provided samplirg time and location
are chosen judicious!) The-se models, may help identif) the locations where mixing occurs and tbe time duration
to attain well mixed conditions The models ma) help choose sampler locations and sampling time dunng a field
study Locations of expected high concentration gradients (near locations of local air exchange, such as doors and
windows) must be sampled more intensively than locations of lou concentration gradients (remote from locations
of local air exchange) Furthermore, CO? quantify potential human exposurevS The models evaluated here are being
incorporated into this project
4

-------
DISCLAIMER
Thu paper 2uu bs a reviewed 10 accordaoce with Ibe U S EPA » peer review and •dauxustrative review
policiei wd ipproved for preweutioa tad publication. Mcouoa of (nde sumee or commercial product* doe* not
cotuuruie esdoneir^ei or recommendation for use.
R£FER£NCES
(l}Bl*ck,F.M . Contra! ofMo4or V«*kk Ervmorn 'Pn L' S taxn«nc« CnUcti K«v«»na En»»ronm««Ul Co^wJ. 31(5.6 ) 373-410 (1991)
(2)	tumt. Kmpp. km! *j> . tmon»' Imc.a of Htndint	Ontmci Owlint or Moot	Trlp*i>t ind Evipcntivt
JCHJUB£23|. ) Axr tad Warn* M*n*g*rneM, 40 I^MIO (19$0j
0) Lumn. A . A B Lmd*rom. K D T«n»l«m«n tnd J S b%a Muluaonil M«u l«l«m .a Multoonc
Buildmfi NiuomS Imuom* of Su.^Urdi ud Tachnotofy. C*i4h«nfcyrf MD	(1991)
(3)	iro»B. IM.I N DKBMdE A C«i Tht Vm of IF6 in Alwo»<>«r>( Tnii»ofl »Wd P»Wwhbb Stud.n I Gaophyt Rt. .10. 3393
J39I (19TJ)
(t) Knwiu,) f,t I	D lami>.*a4R (••«' Ai» Amount tmucn>,«l iyhMt tuinln fpr Anno»htn; Tfi;t SMIi Am
Mti Joe . 1. 372 371 (1914)
f-.or> Hjr.h
5

k,..r*r
Mr 7^«i
Floor Plan - 1st Floor
Ti g\jr« 1. Fesider.ce fleer pi ar.. A seeor.d
fleer fce-rcer (r.et s.v.owr.: was located d.rectly
ateve the garage.
5

-------
Flfltrr I. Simutaied *nd nuiund |*r*|t tiMiini coActauauooi dunaf
and tllcf ¦ 20 aun SFt uniuion
C «PL * Ml ! 20 «v- ft[:r A^i
0*6 Oi v^''t
o i;
s
0 70
0
0 i C
> 0 05
0 00
I 	 k tc
fc- ,
- - 9*3
• y k 1»
» v.fc~ ,
/
o 1 c 2c jo «c s: s:	s:
Ftgwt J, Siimikud tnd mMairtd (Mi i/tntum u houx tonc«nu«uoiu
during «nd »ft«f • 20 bub SF6 anuiwon
6
t
C I
t
i 4
Samp.rs 1 a» (s ***•":. S.' a" ; .0:*' ;k«
1
•	Hc-.-rt
•	LOC'Ct'
	
/ \ •>—
V:-.
0 ? * t e > c i r '« 11
lirrif ? rp> . * '
Fifurt 4. Simuti^ and mufurtd SF6 dtii) tfUr 10 mm SFs cminion
and sftcr iht J,rȣ' door n afxntd
' 6
I	1 4
Z.
l	'7
c	, r
I	.
o
¦;	e
*	4
C
£	J
MiN' J	Lt A$L
C * i> w3 w; *
*o t: ar
Timf (m{t
e

-------