E

Xa

b

a

ARRA

KRFTNFF

1

uufimitapnttb2mwi

aRfrQ

c

0h

Accident Analysis and Prevention 50 (2013) 616– 621

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

jo ur n al hom ep a ge: www.elsev ier .com/ locate /aap

stimation of number of fatalities caused by toxic gases due to fire in road tunnels

iaobo Qua, Qiang Mengb, Zhiyuan Liub,∗

Griffith School of Engineering, Griffith University, QLD 4222, AustraliaDepartment of Civil and Environmental Engineering, National University of Singapore, 117576 Singapore, Singapore

r t i c l e i n f o

rticle history:eceived 5 March 2012eceived in revised form 8 June 2012ccepted 10 June 2012

a b s t r a c t

The quantitative risk assessment (QRA) is one of the explicit requirements under the European Union(EU) Directive (2004/54/EC). As part of this, it is essential to be able to estimate the number of fatalities indifferent accident scenarios. In this paper, a tangible methodology is developed to estimate the numberof fatalities caused by toxic gases due to fire in road tunnels by incorporating traffic flow and the spreadof fire in tunnels. First, a deterministic queuing model is proposed to calculate the number of people at

eywords:oad tunnelireoxic gasumber of fatalitiesractional effective dose

risk, by taking into account tunnel geometry, traffic flow patterns, and incident response plans for roadtunnels. Second, the Fire Dynamics Simulator (FDS) is used to obtain the temperature and concentrationsof CO, CO2, and O2. By taking advantage of the additivity of the fractional effective dose (FED) method,fatality rates for different locations in given time periods can be estimated. An illustrative case study iscarried out to demonstrate the applicability of the proposed methodology.

ire simulation

. Introduction

Road tunnels are vital transport infrastructures, providingnderground passageways for vehicles and commuters, especiallyseful in compact cities like Singapore. With the increasing traf-c, as well as competing demands for land use in cities, more andore road tunnels are being constructed to enhance the accessibil-

ty and capacity of road transport systems. However, fires in roadunnels can lead to catastrophic consequences due to the enclosednd confined space in tunnel systems. For example, in 1999, 39eople lost their lives when fire broke out in the Mont Blanc tun-el between France and Italy, while another disaster in the Tauernunnel in Austria resulted in 12 fatalities (PIARC, 2008). In responseo these fatal accidents, the quantitative risk assessment (QRA) haseen included as a requirement in the European Union directive004/54/EC (EU, 2004). In Singapore, QRAs are compulsory on allajor urban road tunnels exceeding 240 m in length, in accordanceith the Project Safety Review (PSR) procedure manual for roads

n the country (LTA, 2005).A few QRA models for road tunnels have been developed, such

s the Geographic Information System (GIS)-based Transportationisk Analysis model (Bubbico et al., 2004), the TuRisMo model

or Austria, the TUNPRIM model for the Netherlands, the Italianisk analysis model, the OECD/PIARC model (PIARC, 2008), and theRAFT model for Singapore (Meng et al., 2011a,b; Qu et al., 2011).

∗ Corresponding author. Tel.: +65 6516 5494; fax: +65 6779 1635.E-mail addresses: x.qu@griffith.edu.au (X. Qu), ceemq@nus.edu.sg (Q. Meng),

eeliuzy@nus.edu.sg (Z. Liu).

001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.aap.2012.06.010

© 2012 Elsevier Ltd. All rights reserved.

All these models acknowledge that a fire in a road tunnel would beconsidered as the most disastrous initiating event (Bubbico et al.,2009; Meng and Qu, 2012). Once a fire has started, the concentra-tion of oxygen (O2) will decrease dramatically because tunnels areenclosed spaces; at the same time, the concentration of toxic gasessuch as carbon monoxide (CO) and carbon dioxide (CO2) increase.Indeed, toxic gases have been reported as responsible for most firefatalities (Babrauskas et al., 1998; Besserre and Delort, 1997). Fur-ther, in these QRA models, the consequences of various scenariosare evaluated based on the number of fatalities. Accordingly, itis essential to develop a method that can precisely estimate thenumber of fatalities resulting from fires in road tunnels.

Consequence analysis for fires in road tunnels has been studiedsince the 1990s (Modic, 2003). In general, the models for estimatingconsequences are based on empirical or semi-empirical regressionmodels or more advanced and precise computational fluid dynam-ics (CFD) models (Nilsen and Log, 2009). For example, Ingason(2001) proposed a semi-empirical model. Over a number of years,Ingason and his colleagues gathered information on heat releaserates from real fires, using small-scale experiments and full-scaletunnel fire tests. Several equations were proposed and calibrated onthe basis of the collected data. Migoya et al. (2009) developed a CFDmodel to simulate accidental fires in road tunnels. A simplified tun-nel model (UPMTUNNEL) for the simulation of accidental fires withlongitudinal ventilation was created using two simulation tools:FLUENT and PHOENICS.

However, all the existing QRA models for road tunnels applyempirical or semi-empirical models to estimate the number offatalities. This is because there is no method available to estimatethe number of fatalities taking traffic flow patterns (tunnel users)

and P

istiArtcornssmi

FtSmSSt4

2

2

atwnee

arce

N

waawt

D

wsEficl

l

N

X. Qu et al. / Accident Analysis

nto consideration. The literature review carried out for this studyhows that the CFD-based approaches simply focus on estimatinghe concentrations and/or temperatures of smoke, without takingnto account the traffic flow patterns and number of tunnel users.lthough the temperature and concentration of smoke may indi-ectly indicate the severity of fire disasters, it would be better to usehe number of fatalities or fatality rate to explicitly represent theonsequences of these fires, in alignment with the requirementsf the EU directive 2004/54/EC (EU, 2004) and the PSR Manual foroads in Singapore. Hence, a model is needed that estimates theumber of fatalities in fires in road tunnels. This paper puts forthuch a methodology that incorporates traffic flow patterns and thepread of smoke in tunnels. The methodology is then used to esti-ate the consequences of a particular fire occurring in a road tunnel

n Singapore.The contributions of this work can be summarized as follows.

irstly, we propose a methodology to estimate the number of fatali-ies, incorporating the traffic flow pattern and fire spread in tunnels.econdly, a practical application is carried out to guide decisions-akers and policy-makers at the Land Transport Authority (LTA) of

ingapore. The remainder of this study is organized as follows. Inection 2, the methodology is presented. Section 3 contains an illus-rative case study. Discussion of the results is presented in Section. Section 5 concludes.

. Methodology

.1. Estimation of number of people at risk

The people upstream of cross passage doors can easily evacu-te from one tunnel bore to the other. Thus, vehicles upstream ofhe fire site will not be affected by it while those downstream of itill be trapped. Accordingly, the area between the fire site and theearest upstream cross passage door (see Fig. 1) should be consid-red as the risk area. A deterministic queuing model is adopted tostimate the number of people in the risk area as follows.

We assume that the vehicles are generally uniformly distributedcross the risk area when they stop as a result of an incident in aoad tunnel. Assuming continuous traffic flow, the number of vehi-les per lane (N) from the time the fire starts until time t can bestimated by,[

n∑i=1

PiLi

]+ (N − 1)H = D(t) (1)

here n is the number of vehicle types (cars or trucks), Li is theverage length of vehicle type i, Pi is the proportion of vehicles thatre of type i, H is the distance between two successive vehicleshen they stop as a result of the incident, and D(t) is the length of

he risk area, which is mathematically defined as

(t) = min

{D0, Q × t ×

(H +

n∑i=1

PiLi

)}(2)

here D0 is the distance between the fire site and the nearest down-tream cross passage door, and Q is the traffic volume. According toq. (2), the length of the risk area will be D0 if the area between there site and the cross passenger door is fully occupied with vehi-les; otherwise, D(t) will be the distance from the fire site to theocation of the last vehicle that has entered the area.

Thus, the number of vehicles in the risk area in the various traffic

anes (Nv(t)) can be estimated by

v(t) = nlane × D(t) + H

H +[∑n

i=1PiLi

] (3)

revention 50 (2013) 616– 621 617

where nlane is the number of lanes in the tunnel. Accordingly, thenumber of people at risk (Npar(t)) is

Npar(t) = Nv(t)

(n∑

i=1

PiOi

)(4)

where Oi is the average number of people in a vehicle of type i.

2.2. Fire simulation model

The process of fire growth and spread can be formulated usingconservation equations for mass, momentum, energy, and species,coupled with the equation of state, which is the basis and foun-dation of the FDS program (McGrattan, 2005). Note that all theseequations are from the FDS technical reference guide by McGrattan(2005). The conservation of mass is written as:

∂�

∂t+ ∇�u = 0 (5)

where ∂�/∂t represents the density change over time while u isthe velocity vector. The following equation describes how the rateof mass storage within the control volume, due to the change indensity, is balanced by the net rate of inflow. The conservation ofmomentum is described as:

∂(�u)∂t

+ ∇puu = −∇p + �f + ∇�ij (6)

where ∂(�u)/∂t and ∇puu on the left-hand side of the equationdefine the rate of change of momentum, p represents pressure, �ijis the stress tensor acting on the fluid, and f consists of gravity plusother forces such as the drag exerted by liquid droplets (McGrattan,2005). The conservation of energy is written as:

∂(�h)∂t

+ ∇�hu = ∂p

∂t+ q′′′ − ∇q + (7)

where ∂(�h)/∂t and ∇�hu are the net rate of energy accumulationwithin the control volume while the terms on the right-hand siderepresent the heat release rate per unit volume from a chemicalreaction (q′′′), the conductive and irradiative heat flux (∇q), and thedissipation function (˚), that is, the rate at which kinetic energyis converted to thermal energy due to the viscosity of the fluid(McGrattan, 2005). The equation of state is written as:

p = �RT (8)

where � is the density, p is the pressure, R is the gas constant(287.05 J/(kg K)), and T is the temperature (K). The conservation ofspecifies is written as:

∂�Yi

∂t+ ∇�Yiu = ∇�Di∇Yi + m

′′′i (9)

where the fluid consists of a mixture of species, and the transportequations for each species will need to be solved. Here, Yi is themass of the ith species, Di is the diffusion coefficient of species i intothe mixture and m′′′

iis the production rate of species i (McGrattan,

2005).In this study, we apply the Fire Dynamics Simulator (FDS)

program to solve the above equations numerically. The program,developed by the National Institute of Standards and Technol-ogy (NIST), has been used extensively for fire simulations (e.g.Tsukahara et al., 2011). The FDS has been validated by comparingto experimental results (e.g. Cochard, 2002; Smardz, 2006; Lee andRyou, 2006; Hu et al., 2007; Trelles and Mawhinney, 2010, etc.). Itworks as follows. First, the initial pressure and temperature, the

tunnel geometry, the fire size and location, materials such as fuelsthat are present, the type of fire detection systems and ventila-tion systems that are in place, and the simulation period are inputinto the FDS program. Then, the FDS numerically solves the above

618 X. Qu et al. / Accident Analysis and Prevention 50 (2013) 616– 621

ram for the queuing model.

eppn

2

dttoastaeeaetriclp

om

F

wt�p

Pttrctf

F

cn

)

Fig. 1. Schematic diag

quations to give the densities of various toxic gases and the tem-erature of the smoke in different locations during the simulationeriod. Finally, the output module of the program graphically andumerically represents the results (densities and temperatures).

.3. Fatality rate estimation

The fatality rates1 need to be estimated at various locationsuring the time t ∈ [0, T]. The fatality rate here refers to the ratehat number of fatalities (including those died in the post-exposureime) caused by exposing to the risks for a given time period outf the number of people at risk area. It could also be considereds probability that one individual in risk area would die if helorhe exposes to the risks for a given time period. The concentra-ions of various types of gases (CO, CO2, and O2) can be obtained atny location and at any time from the FDS program. The additiveffects of combustion gases have been demonstrated in a number ofxperiments using rodents (Hartzell et al., 1985; Levin et al., 1987),nd these have since been extended to include the consideration ofxposure time. This strategy is commonly referred to as the frac-ional effective dose (FED) methodology. The FED is defined as theatio of Ct (concentration × time) for a gaseous toxicant producedn a given test to the Ct for the same toxicant that has been statisti-ally determined from independent experimental data to produceethality in 50% of test animals within a specified post-exposureeriod2 (ASTM, 2002; Hartzell, 2001).

The additivity of FEDs has become a useful property in the areaf fire toxicology, helping to estimate the number of fatalities, asathematically represented by

ED =n∑

i=1

m�t∑�=0

Ci,�

(Ct)i�t =

n∑i=1

m�t∑t=0

Fi,��t (10)

here Ci,t is the concentration of toxic component i at time �, Fi,� ishe FED caused by toxic component i over the exposure period [�,

+ �t], (Ct)i is the specific dose (concentration × time) required toroduce lethality; �t is the time increment (min) and t = m�t.

According to the Fire Protection Handbook (National Firerotection Association, 2008), CO2 is quite low in terms of its ownoxicological potency and is not, by itself, normally considered as aoxicant in fire atmospheres. However, it does stimulate both theate and depth of breathing, thereby increasing the fatality rateaused by CO. Levin et al. (1987) developed an empirical FED func-ion for an exposure of 30 min to combinations of CO and CO2 asollows.

CO&CO2 (XCO, XCO2 , 30) = m · XCO

XCO2 − b(11)

1 The focus of the consequence analysis is to estimate the number of fatalitiesaused by fire smokes in road tunnels. Thus, the fatalities due to fire or collision areot analyzed in this case.2 The equations used in this study are with a postexposure time of 30 days.

Fig. 2. KPE road tunnel in Singapore.

where XCO is the concentration of CO (in ppm); XCO2 is the concen-tration of CO2 (in volume percentage); m and b are two coefficientsdefined as follows: if the concentration of CO2 is less than 5%,m = −18 and b = 122,000; otherwise, m = 23 and b = −38,600. Theconfirmatory work using this model has been published by Pauluhn(1993).

Due to the additivity of FEDs, the FED function for exposure overthe time period [0,t] to combinations of CO and CO2 is

FCO&CO2 (XCO, XCO2 , t) = m · XCO

XCO2 − b× t

30(12)

According to Persson (2002), the FED with respect to a low con-centration of O2 over an exposure time period of [0,t] is

FO2 (XO2 , t) = t

e8.13−0.54(20.9−XO2)

(13)

where XO2 is the concentration of O2 (in volume percentage).By substituting Eqs. (12) and (13) into Eq. (10), we can obtain the

FED for the mixed effects of CO, CO2, and O2. Due to the additivityof FEDs, the fatality rate3 due to an exposure time period [0,t] atlocation m can be estimated by

Fm(XCO, XCO2 , XO2 , t) = (FCO&CO2 (XCO, XCO2 , t) + FO2 (XO2 , t)) × 50%(14

3. An illustrative case study

3.1. KPE road tunnel

The Kallang/Paya Lebar Expressway (KPE) in Singapore, shownin Fig. 2, was built to serve the growing traffic demands of the

3 FED refers to 50% fatality rate. Thus the actual fatality rate should be half of theFED.

X. Qu et al. / Accident Analysis and Prevention 50 (2013) 616– 621 619

sectio

nSwl

svesw

3

bbtidictinbiTiec

3

ta5f

i=1

i i

The distance between two successive vehicles when they stopbecause of an incident is assumed to be 3 m. Putting these figures

Fig. 3. Conceptual cross-

ortheastern sector of Singapore. It is the longest road tunnel inoutheast Asia (9 km). It is a dual three-lane underground passage-ay and has nine entry slip roads, eight exit slip roads and six

ongitudinal ventilation buildings.The vehicles going in opposite directions in the road tunnel are

eparated by a central dividing wall. Cross passage doors are pro-ided every 100 m to expedite the evacuation of tunnel users in thevent of a tunnel fire. As can be seen in Fig. 3, there is a 2.4 m widehoulder and three 3.6 m wide traffic lanes in each carriagewayhile the tunnel’s structural height is approximately 6 m.

.2. The accident response system

In the event of a fire, any traffic downstream of the fire site wille able to drive away while traffic upstream of the fire site wille trapped. Prompt detection of a fire in the tunnel is an impor-ant factor in preventing a catastrophic fire incident. The tunnelncorporates two types of fire detection systems; automatic inci-ent detectors (AID) and linear heat detectors (LHD) are provided

n the tunnel with the fire detection time set at 30–60 s. Closed cir-uit television (CCTV) and emergency telephones installed in theunnel are used to verify the occurrence of a tunnel fire. The fire ver-fication system should take around 60 s to respond after receivingotice from the fire detection systems. After that, the tunnel shoulde ventilated in 2 min and the smoke caused by the fire released

nto the atmosphere via exhaust stacks in the ventilation buildings.he timeline of the response plan in the event of a fire is illustratedn Fig. 4. We assume that most motorists and passengers will notvacuate from the tunnel bore until they are informed to do so (aonservative assumption).

.3. Concentrations of toxic gases

In this study, we simulate using the FDS a fire caused by a

ruck-to-truck collision. The fire site is located 85 m away fromn upstream cross passage door. The fire size is taken to be0 Mega-Watt (MW). As can be seen in Fig. 4, the response timeor ventilation system is 225 s. Considering human beings’ walking

Fig. 4. Timeline of incident response plan.

nal layout of the tunnel.

speed (0.64–1.2 m/s), it would take at least 295.8–357.8 s to get allpeople out of the risk area. Thus, the simulation time is set as 350 s.The initial mass fraction of CO2, CO and O2 are 0.03%, 0, and 21%respectively. The fuel materials are assumed to be diesel. The twobounds of the risk area are assumed to be “vent” type. The tun-nel size (mesh) is assumed to be 85 m × 20 m × 6 m. The simulationtime is 350 s. Default values are applied for all the other parameters(heat detectors, smoke detectors, etc.).

The temperature and concentrations of toxic gases at any posi-tion in the tunnel and at any time during this period can be obtainedfrom the simulation model. Figs. 5–7 depict the distributions of con-centrations of CO, CO2, and O2 at different times in the risk area. The(0,0) point represents the fire site.

3.4. Estimation of number of fatalities

According to historical records, the proportions of trucks andcars in the tunnel at any one time are 35% and 65%, respectively.Therefore, the average length of the vehicles is

L =2∑

P L = 0.21 × 20 m + 0.79 × 3.5 m = 6.97 m (15)

Fig. 5. CO concentrations in the risk area (t = 19.8 s).

620 X. Qu et al. / Accident Analysis and Prevention 50 (2013) 616– 621

Table 1Summary of fatality rates for the 24 vehicles.

Location (distance from fire site) Inner lane Middle lane Side lane

Vehicle no. Fatality rate Vehicle no. Fatality rate Vehicle no. Fatality rate

15 m 1 2.14e−2 9 9.64e−2 17 2.14e−225 m 2 1.77e−2 10 7.16e−2 18 1.77e−235 m 3 1.43e−2 11 6.20e−2 19 1.43e−245 m 4 1.20e−2 12 5.37e−2 20 1.20e−255 m 5 9.18e−3 13 4.27e−2 21 9.18e−365 m 6 7.41e−3

75 m 7 4.71e−3

85 m 8 2.41e−3

ial

N

fs

Fig. 6. CO2 concentrations in the risk area (t = 22.5 s).

nto Eq. (3), we find that the number of vehicles at risk would beround 24, assumed to be uniformly distributed along the trafficanes. The average number of people in each vehicle would be

¯ =

2∑i=1

PiNi = 0.21 × 1.8 + 0.79 × 2.5 = 2.353 (16)

The fatality rates for the 24 vehicles at risk in the time intervalrom 0 to 225 s can be estimated using Eqs. (6)–(8). The results areummarized in Table 1.

Fig. 7. O2 concentrations in the risk area (t = 12.6 s).

14 3.17e−2 22 7.41e−315 2.62e−2 23 4.71e−316 2.07e−2 24 2.41e−3

Thus, the total number of fatalities caused by toxic gases can beestimated as

Nf =24∑i=1

NiFi = 1.3725 (17)

4. Discussion

In this study, we simulated a fire caused by a truck-to-truckcollision in the KPE road tunnel in Singapore. The fire size is 50 MW.As per our analysis, the number of fatalities caused by the fire wouldbe around 1 and the overall fatality rate would be a bit more than2%, suggesting that the tunnel would be relatively safe even if a50 MW fire occurred in it. In reality, according to the ConceptualDesign Report on Fire and Life Safety for the KPE road tunnel, the tunnelis designed to cope with fires up to 100 MW.

In Singapore, a Hazmat (Hazardous Materials) Transport Vehi-cle Tracking System has been introduced by the Singapore CivilDefence Force. The system aims to ensure that vehicles carry-ing hazardous materials do not use road tunnels. Furthermore, inaccordance with the Road Traffic Act of Singapore, notice must besubmitted to the LTA before a truck can enter the KPE road tunnel.Thus, the frequency of truck-to-truck collisions is very rare. Indeed,no truck-to-truck collisions have occurred in either of the two roadtunnels in Singapore (KPE tunnel and CTE tunnel) since they wereopened to traffic. According to the proposed model, even if such arare event did occur, the consequences in terms of the number offatalities would generally be acceptable. The vehicle collisions withrespect to other vehicle types may also result in small size fire (thecar fire is around 5–10 MW, van fire is around 15 MW).

5. Conclusions

This paper has developed a tangible methodology to preciselyestimate the number of fatalities caused by fires in urban road tun-nels. First, the number of people at risk is calculated, taking intoaccount the impact of tunnel geometry and the incident responseplan. The temperature distribution and concentrations of CO, CO2,and O2 are estimated for any given fire size, as measured in MW,using the FDS. Based on FEDs, the fatality rates caused by the fire atany given location and in different time periods are then obtained.Our illustrative case study shows that the number of fatalities dueto a 50 MW fire in the KPE road tunnel in Singapore is a bit morethan 1 and the fatality rate is around 2%.

The estimation of the number of fatalities is essential whenmaking a risk assessment of a road tunnel, which is an explicitrequirement under EU directive 2004/54/EC and the PSR Manualfor roads in Singapore. However, it is difficult to validate this work

due to a lack of available data. In addition, in this study we do notestimate the number of people that might be injured by a fire in aroad tunnel. Future work could focus on this and also simulate thepassenger evacuation process.

and P

A

hit(iaA

R

A

B

B

B

B

C

E

H

H

H

tunnels protected by water mist systems. Journal of Fire Protection Engineering

X. Qu et al. / Accident Analysis

cknowledgements

The authors wish to thank two anonymous referees for theirelpful comments and valuable suggestions which considerably

mproved the exposition of this work. This study is supported byhe innovation fund of the Land Transport Authority of SingaporeContract No. ER 253). The project was conferred with the Min-stry of Transport Minister’s Innovation Award. Special thanks arelso expressed to Mr Kum Thong Yong from the Land Transportuthority of Singapore for the data collection.

eferences

STM E1678-02, 2002. Standard Test Method for Measuring Smoke Toxicity for Usein Fire Hazards Analysis. ASTM International, W. Conshohocken, PA.

abrauskas, V., Gann, R.G., Levin, B.C., Paabo, M., Harris, R.H., Peacock, R.D., Yasa, S.,1998. A methodology for obtaining and using toxic potency data for fire hazardanalysis. Fire Safety Journal 31, 345–358.

esserre, R., Delort, P., 1997. Recent studies prove that the main cause of deathduring urban fires is poisoning by smoke. Urgences Medicales 16, 77–80.

ubbico, R., Di Cave, S., Mazzarotta, B., 2004. Risk analysis for road and rail transportof hazardous materials: a GIS approach. Journal of Loss Prevention in the ProcessIndustries 17 (6), 483–488.

ubbico, R., Di Cave, S., Mazzarotta, B., Silvetti, B., 2009. Preliminary study on thetransport of hazardous materials through tunnels. Accident Analysis and Pre-vention 41, 1199–1205.

ochard, S., 2002. Validation of the Freeware Fire Dynamics Simula-tor Version 2.0 for Simulating Tunnel Fires. Zurich. Available athttp://lhe.epfl.ch/pageperso/Cochard/images/FDS2.pdf (accessed 21 April2012).

U, 2004. Directive 2004/54/EC of the European parliament and of the council of 29April 2004 on minimum safety requirements for tunnels in the Trans-EuropeanRoad Network. Official Journal of the European Union (April), L167/39–91.

artzell, G.E., 2001. Engineering analysis of hazards to life safetyin fires: the fire effluent toxicity component. Safety Science 38,147–155.

artzell, G.E., Switzer, W.G., Priest, D.N., 1985. Modeling of toxicological effects offire gases: V. Mathematical modeling of intoxication of rats by combined carbonmonoxide and hydrogen cyanide atmospheres. Journal of Fire Sciences 3 (5),

330–342.

u, L.H., Huo, R., Wang, H.B., Yang, R.X., 2007. Experimental and numeri-cal studies on longitudinal smoke temperature distribution upstream anddownstream from the fire in a road tunnel. Journal of Fire Sciences 25,23–43.

revention 50 (2013) 616– 621 621

Ingason, H., 2001. An overview of vehicle fire in tunnels. In: Proceedings of the 4thInternational Conference on Safety in Rail and Road Tunnels, April, pp. 425–443.

Land Transport Authority (LTA), 2005, April. Design Safety Submission for KPE RoadTunnels. LTA, Singapore.

Lee, S.R., Ryou, H.S., 2006. A numerical study on smoke movement in longitudinalventilation tunnel fires for different aspect ratio. Building and Environment 41,719–725.

Levin, B.C., Paabo, M., Gurman, J.L., Harris, S.E., 1987. Effects of exposure to single ormultiple combinations of the predominant toxic gases and low-oxygen atmo-spheres produced in fires. Fundamental and Applied Toxicology 9, 236–250.

McGrattan, K., 2005. Fire Dynamics Simulator (Version 4) Technical Reference Guide.National Institute of Standards and Technology.

Meng, Q., Qu, X., 2012. Estimation of rear-end vehicle crash frequencies in urbanroad tunnels. Accident Analysis and Prevention 48, 254–263.

Meng, Q., Qu, X., Wang, X., Yuanita, V., Wong, S.C., 2011a. Quantitative risk assess-ment modeling for non-homogeneous urban road tunnels. Risk Analysis 31,382–403.

Meng, Q., Qu, X., Yong, K.T., Wong, Y.K., 2011b. QRA model based risk impact analysisof traffic flow in urban road tunnels. Risk Analysis 31, 1872–1882.

Migoya, E., Crespo, A., Garcia, J., Hernandez, J., 2009. A simplified model of fire inroad tunnels. Comparison with three-dimensional models and full-scale mea-surements. Tunnelling and Underground Space Technology 24, 37–52.

Modic, J., 2003. Fire simulation in road tunnels. Tunnelling and Underground SpaceTechnology 18, 525–530.

National Fire Protection Association, 2008. Fire Protection Handbook, 20th edition.National Fire Protection Association, Boston, MA.

Nilsen, A.R., Log, T., 2009. Results from three models compared to full-scale tunnelfire tests. Fire Safety Journal 44, 33–49.

Pauluhn, J., 1993. A retrospective analysis of predicted and observed smoke lethaltoxic potency values. Journal of Fire Sciences 11 (2), 109–130.

Persson, M., 2002. Quantitative Risk Analysis Procedure for the Fire Evacuation of aRoad Tunnel—An Illustrative Example. Lund University, Sweden [Report 5096].

PIARC Technical Committee C3.3, 2008, May. Road Tunnel Oper-ation Division. Risk Analysis for Road Tunnels. Available athttp://publications.piarc.org/ressources/publications files/4/2234,TM2008R02-WEB.pdf (accessed on 19th January 2012).

Qu, X., Meng, Q., Yuanita, V., Wong, Y.K., 2011. Design and implementation of aquantitative risk assessment software tool for Singapore’s road tunnels. ExpertSystems with Applications 38, 13827–13834.

Smardz, P., 2006. Validation of Fire Dynamics Simulator for Forced and NaturalConvection Flows. Master Thesis. University of Ulster.

Trelles, J., Mawhinney, J.R., 2010. CFD investigation of large scale pallet stack fires in

20, 149–198.Tsukahara, M., Koshiba, Y., Ohtani, H., 2011. Effectiveness of downward evacua-

tion in a large-scale subway fire using Fire Dynamics Simulator. Tunnelling andUnderground Space Technology 26, 573–581.