Evacuation models are running out of time

Fire Safety Journal 78 (2015) 251–261

Contents lists available at ScienceDirect

Fire Safety Journal journal homepage: www.elsevier.com/locate/firesaf

Evacuation models are running out of time Peter Thompson a,n, Daniel Nilsson b, Karen Boyce c, Denise McGrath d a

Autodesk Ltd., Farnborough, UK Lund University, Lund, Sweden Ulster University, UK d University College Dublin, Dublin, Ireland b c

art ic l e i nf o

a b s t r a c t

Article history: Received 23 December 2014 Received in revised form 10 August 2015 Accepted 19 September 2015

The representation of crowd movement in existing evacuation models is typically based on data collected in the 1950s to 1980s, i.e., data that are more than 40 years old. Since the 1970s, population characteristics have changed dramatically around the world. Reports show that the percentage of elderly and obesity rates have increased significantly and this trend is predicted to continue into the future. Recent research [1–3] illustrates the magnitude by which different age cohorts of a population group can reduce the general speed and flow rates. In addition, well established studies have quantified the impact of body dimensions on speed and flow [4]. However, many existing evacuation models fail to take the changing characteristics of populations into account. This paper aims to review existing knowledge of population demographics and crowd dynamics, derive an indicative flow reduction factor for future populations, and consider the implications for computer models and building design in the future. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Evacuation Model Population Crowd Walking Flow Safety Simulation Demographics Emergency Standards

1. Introduction The standard analyses of life safety in buildings and transportation systems use simple flow rates and walking speeds to calculate evacuation times. Current design guidance documents typically use a basic flow rate for a single uniform population, which has not changed since door and passageway sizes were initially regulated in the mid-20th century. The reference datasets used for current guidance documents were published more than 50 years ago [4–6] and therefore potentially pose an unquantified risk to life for groups with different mobility characteristics. In recent years, these long-established data and relationships between speed or flow and density have been questioned. Indeed, the originators of what are widely considered some of the most significant North American datasets (Fruin [6], and Pauls [7]) have stated that their datasets are no longer applicable and have asked them to be removed from future design guides [8]. The loss of confidence in the use of older ‘uniform’ data is due to the recognition of the ever increasing proportions of elderly, obese and mobility impaired in our society (United Nations [9,10],

OECD [11]). These proportions have increased significantly since the original observations were made of the egress and circulation flows of office-workers and commuters in the 1950s–1970s. Mixed populations may have a dramatic effect on optimal crowd flow movement and ultimately safe escape. Some recent studies have reviewed flow data and formulae for people of different ages on stairs [12] and level surfaces [1–3] and the indications are that, while basic flow values for uniformly ‘healthy adult’ groups may not have yet changed significantly, there is a very significant drop in flow rate for primarily ‘elderly’ populations. A recent study by the National Institute of Standards and Technology (NIST) indicates that the mean walking speed for elderly adults in staircases was 0.28 m/s, while multiple sources [6,13,14] quote between 0.48 and 1.7 m/s for ‘healthy adult’ groups descending staircases. Reduced speeds and mobility will have a consequential impact on flow rate. We know that the following parameters affect individual walking speeds, but no account is taken of these factors in crowd flow analysis:


Changes to age distribution – (United Nations [9], Ando et al. [15]).


Population physical size: overweight and obesity rates (OECD n

Corresponding author. E-mail address: [email protected] (P. Thompson).

http://dx.doi.org/10.1016/j.firesaf.2015.09.004 0379-7112/& 2015 Elsevier Ltd. All rights reserved.

[11]).


Presence of disability (Boyce et al. [16]).

252

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

Prescriptive regulations and design guides [17–23] use one uniform standard flow rate (typically 80 people/metre-width/min, sometimes more for sports stadia), and even modern computer simulation models use these unified parameters for generic populations. Homogenous approximations to populations are less valid with the increasing mixture of abilities and it is the assertion of the authors that the current modelling approaches are ‘running out of time’, as population demographics continue to change but the underlying mathematics do not. New approaches, matched by corresponding data are needed in order to consider the life safety implications now and in the future. Given the changing characteristics of building users, i.e., increasing age (leading to greater incidence of disability [24], reduced walking speeds and increased use of walking aids) and higher obesity rate (reduced walking speed and less people per unit area), it is imperative to re-examine the representation of crowd movement in evacuation models. The present paper aims to take an important step in this direction by: (1) reviewing the population demographic trends over an 80 year timeframe; (2) reviewing the increasing sophistication of the analytical capability of evacuation models, but illustrating the simplicity of the underlying mathematics; (3) suggesting some first mathematical accommodations for the effects of the demographic changes; (4) discussing a possible way forward for the future study of crowd movement taking biometrics and people interaction into account.

2. Evacuation calculations: basic flow rates over 40 years The majority of calculations for evacuation analyses are still based on simple flow rates, i.e. the rate at which a given population will ‘flow’ through an evacuation element (passageway, doorway or staircase), expressed in people per unit width, per unit of time. Our present understanding of pedestrian movement in populated spaces is based on relatively old data on mainly ablebodied people. The most significant datasets used in the analysis of people movement and evacuation [5,25,4,6,7,15] are derived from research conducted between the 1950s and 1980s, see Table 1. These datasets are derived from observations of the movement of able-bodied commuters (Hankin and Wright [5], Fruin [6]; Ando et al. [15]), pedestrians in normal circulation in a range of building types (Predtechenskii and Milinskii [4]) or during evacuation drills in buildings (Pauls [7]; Predtechenskii and Milinskii [4]). This early research has formed the basis of our understanding of flow phenomena and indeed formed the basis of design guidance documents worldwide, e.g., the Green Guide (Home Office [17]), SFPE Handbook of Fire Engineering (Nelson and Maclennan

[18]), PD 7974-6 (BSI [19]). Some sample flow rates from the second half of the 20th century are identified in Table 1. A commonly used value which underpins current guidance documents in the UK (Approved Document B [21]), US (NFPA 101: Life Safety Code [22]) and those of the International Maritime Organisation (IMO [23]) is 80 people/m/ min, also expressed as 1.33 people/m/s. The simple flow rate approach and figure has changed little since the 1970s. For example, several earlier versions of the US and UK standards from the 1970s used the unit-width calculation of 40 people/21 in./min, which equates to 74.99 people/m/min. In fact, the standardised flow rates for this basic calculation approach have increased slightly in later documents, but for no apparent scientific reason other than the expediency of unified numbers. The doubts cast upon some of the original data [27,6] has led on to more recent data gathering and reviews. For example, the NIST study [12] combined the collection of new staircase movement data for elderly and mixed-ability occupants, and office populations, with a review of egress movement on staircases It highlights significant differences in walking speeds for elderly populations. However, there are currently no plans to reassess or change the ‘standard’ flow rates adopted in current guidance, while population demographics continue to change. Terms such as ‘obesity epidemic’ and ‘ageing society’ are increasingly common. We know that factors such as age and physical attributes (e.g. body size etc) affect individual walking speed but because there are no dedicated scientific analyses of such effects on a group basis (i.e. for ‘crowds’), there is currently no anticipated change in the majority of computer ‘flow’ models that inform evacuation standards, now or for the next 40 years.

3. The fundamental physics of ‘flow’ The basic equation of crowd ‘flow’ is expressed in Eq. (1) below:

q=v×d

(1)

where q ¼flow per unit width (p/m/s), v ¼velocity (m/s), d¼ density (p/m2). We should not forget that, like any physical model, crowd ‘flow’ is an expression of the rate at which quantities (of people) pass through a spatial unit (in the evacuation system) per unit of time. If parameters change which affect the crowd velocity (walking speeds) or density (concentration and size), then there is an inevitable physical implication for the flow. It is important to consider basic physics as we consider the implications of complex demographic changes and physical interactions.

Table 1 Summary of historic crowd flow rates. Year

Source

Max. design flow (p/m/ sec)

Ultimate flow (p/m/ sec)

Scope of data/analysis

1958 1969 1972 1973

Hankin and Wright [5] Predtechenskii and Milinskii [4] SCICON report [20] Guide to Safety at Sports Grounds [17]

1.48 1.70 1.37 1.82(unit exit width method) 1.37

1.92 2.06

Commuters under normal conditions Peak flows at high density for adults in summer dress. Data from football crowds Based on Japanese data and derived from 1.0 pers/0.55 m/s unit exit width calculation Max. flow is ultimate regimented, 'funnelled' soldiers flow under pressure Data collected in Israel, sidewalks Commuters under normal conditions

1971 Fruin [6] 1983 Polus et al. [26] 1988 Ando et al. [15]

1.25–1.58

4.37 1.56 1.7–1.8

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

253

Fig. 1. The proportion of ‘over 65s’ relative to the adult population, now and future [28].

4. Evacuation analyses: using ‘flow’ and time The essential aim of life safety calculations (involving evacuation analyses) is to ensure that the occupants escape the building before conditions become untenable and there is a clear risk to life. The basic approach is to ensure that the time taken by occupants to reach a place of safety (the Required Safe Escape Time) is less than the time expected for them to become exposed to unsafe conditions (Available Safe Escape Time). The ASET vs RSET calculation (BSI [19]) is well established and used extensively amongst the fire safety community. A very early assessment of an appropriate ASET was a nominal two and a half minutes, which was derived from the review of the real evacuation of the Empire Theatre in Edinburgh in 1911. In that incident, the presence of a fire was alerted at around the start of the playing of the national anthem. The theatre was evacuated in approximately the duration of the anthem (as the orchestra continued to play) although there was some loss of life. It is important to note that, even now, this is the figure used in UK guidance documents [21] as a target flow time through exits. This duration of 150 s, combined with the assumption of a flow rate of 1.33 p/m/ s (80 p/m/min) results in the most commonly used exit width per person (5 mm/person) that is used for the calculation of exit width capacity in current design guidance (Eq. (2)).

q=

Np (people) w (metres) t (sec .)

∴

⎛ 80 ⎞ Np ⎟ × 150 = 200 =q×t=⎜ ⎝ 60 ⎠ w

1 w ∴ = = 0.005 200 Np

design approval. Performance based design can involve the use of simple hand calculations or more complex computer modelling. Simple hand calculations are dependent upon reliable assumptions of speed and flow (often a density has to be assumed) and can only be as robust as the underlying data. Computer modelling, although evolving significantly over the last decades, is also dependent upon the underlying data. If the base data or movement parameters for a model are flawed or out of date then that has a real impact on the reliability of the results, and hence the confidence with which the results can be used to assess life safety.

5. Population demographic trends The standard flow rates and design curves, which are in use today were collected mostly for healthy, fit, office or commuter populations between the 1950s and1970s (some of which are presented in Table 1). Since then, the accessibility to buildings for people with disabilities has changed significantly, as have the demographics of populations globally. There is a much higher proportion of elderly people in most regions, and this is predicted to become much more evident heading towards 2050 (United Nations []). Additionally, the proportion of overweight and obese people in the adult population is predicted to continue to increase significantly (OECD [11]). 5.1. The ‘ageing society’

(2)

where q ¼flow per unit width (p/m/s), Np¼ number of people, w¼ exit width (m), t ¼total time (s). Therefore, the width per person for ‘safe’ escape ¼ 0.005 m (5 mm). This ‘standard’ number of 5 mm exit width per person to allow ‘safe’ evacuation used by building codes in the UK [21] and US [22] therefore represents an approximation for a uniform, standard population (from the mid-20th century) to escape within the time duration of the UK national anthem. Clearly both the 2.5 min target and the flow of 80 people/m/min can both be challenged regarding appropriateness for any given occupancy. The current scope for change is that most modern regulations now also allow for more advanced ‘performance-based’ approaches to be adopted, where sufficient scientific rigour must be demonstrated to the authority having jurisdiction for building

The predicted differences in the proportion of elderly in society in different regions of the world, both now and in the future are illustrated in Fig. 1, and noted in multiple sources [9,28]. There are several reasons for such a significant change in the demographics of the population: 1. The ‘baby boom’ generation is getting older: the higher birth rate in the 1960s has produced, to some extent, a population ‘bubble’ [29]. For example, the birth rate for the USA has nearly halved, per mother, since 1960. 2. People are also living longer (medical science, diet); life expectancy rose from 70 to nearly 80 in the US over the last 40 years [30]. 3. Mothers are having children at an older age [31], producing a delay in the impact of births on society, contributing to a temporary increase in the ‘ageing’ proportion of society.

254

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

statistical tendency for less exercise as part of people's daily lives [37]. In addition, childhood obesity is rising [38], indicating that the observed trends are due to continue into the future. If we look at the predicted trend for the USA (Fig. 3), then adult obesity is due to rise to 50% by 2030 (Levy at al. [39]).

6. The impact of demographics on speed and flow

Fig. 2. Adult obesity rates in multiple OECD countries, over 40 years[28].

The population trends create a significant statistical change across the OECD countries (Fig. 1). The average proportion of ‘elderly’ in adult society is predicted to change from just over 20% to nearly 50% by 2050, with some countries (Japan, Spain, Italy) predicted to have a significant majority of the adult population in the over-65s category. It should be noted that over 65's may also be somewhat healthier than in previous generations, which may affect flow rates, but the magnitude is unclear. The magnitude of the predicted demographic changes will have a significant impact on society as a whole. Governments are carefully considering the financial impacts and also the impact on the caring professions [32], and are putting in place a number of mechanisms intended to cope with these changes. It is important also that building designs, and the data and tools that we use in engineering analyses also adapt to reflect and accommodate our changing society. 5.2. Larger bodies and the ‘obesity epidemic’ In most countries around the world, the rate of obesity and ‘overweight’ proportion of people has been increasing steadily for several generations, and indeed this change has become more rapid and evident in the last 25 years (OECD [11]) (Fig. 2). Health care systems are making plans to accommodate larger and more obese people in terms of bed sizes and even the patient lifting equipment in ambulances have to be modified [36]. The trends are becoming well documented, and are due to a combination of higher calorific intakes per head of population combined with a

Crowds consist of people in moderately or highly congested space, and it is important to understand the implications that slower, mobility-impaired, or larger occupants may have on the overall crowd movement: restricted space limits the ability of individuals to overtake, and the less mobile crowd members will therefore begin to dictate speed and flow. It is suggested that even a minority of slowly walking individuals may have a dramatic effect on crowd flow rates. At the sustained ‘design’ flow rates between 3 and 4 people/m2 (shown in the range illustrated, Fig. 4, by Fruin [6] between the “touch” and “no-touch zones”), overtaking is not possible so the slower members of the crowd will dictate the overall speed and hence flow rate. In addition, for escape elements such as doorways and stairs, the slower, and possibly larger members will begin to dictate flow rate as they transition through that restricted-width space. The increasing prevalence of elderly and obesity are of such magnitude that they are likely to have a significant impact on the population’s mobility and body size, and subsequently on crowd movement parameters. 6.1. The impact of age on crowd speed and flow Age is directly related to a deterioration in physical, mental and neurological functions (Reeves et al. [33], Kang and Dingwell [34]), which impacts negatively on individual movement, e.g., speed and stride length. The chart illustrated by Ando et al. [15], shown in Fig. 5, shows that ‘elderly’ (over 65s) walk approximately 20–25% slower than adults in the age range 18–40 years. Other research by Fujiyama and Tyler [35] suggests that 60–84 year old descend stairs at unimpeded speeds of between 0.60 m/s and 1.11 m/s depending on stair geometry and urgency, which is much lower than typical unimpeded descent speeds achievable by younger adults. The slightly more recent NIST study [12] indicates that the rate at which elderly people descend stairs (assisted or unassisted) can be much much lower – a 50% reduction compared to healthy adults. If the predictions from the OECD reports [11] are correct, the future proportion of ‘elderly’ in society will outweigh the rest of

Fig. 3. The prevalence of obesity among US adults 1960–2030 [39].

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

255

Fig. 4. Fruin's “Touch” and “No-Touch Zones”reproduced from [6].

1.2 1.1 1

Walking Speed (m/s)

0.9 0.8 0.7 0.6

Adults (mid-age)

0.5

Children (pre-school)

0.4

Elderly

0.3 0.2 0.1 0 0

1

2

3

4

5

6

7

8

9

Density (people / m )

Fig. 5. The relationship between walking speed and age on a level surface (Ando et al. [15]).

the adult population in some countries. It follows that the average walking speed of the population will decrease and it is therefore important to understand the impact that this may have with respect to the flow in different situations. If we consider the statistical implications of Figs. 1–5, then there may be a 20–25% reduction in crowd speed and flow rates even for groups of people with a significant minority of elderly or infirm people where naturally faster occupants can no longer overtake at these higher densities. This effect is likely to be more when the flow is clearly ‘dominated’ by the less able occupants, and also when you consider the walking speeds of assisted and non-assisted elderly in the NIST [12] study, with walking speeds of 50% or less, compared to healthy adults on stairs. Recent studies by Kholshenikov et al. [1–3] have explicitly quantified the impact of population demographics, on crowd speed and flow rates, see Figs. 6 and 7. The studies referred to previous tests and more recent data gathering, and provide a very useful insight into the trends and implications for general crowd movement on level surfaces. The three population types included in the study, namely adults, children and elderly, can reasonably be assumed to have varying age and body size characteristics, but we are using the mean values quoted in their findings. The data used for charts in Figs. 6 and 7 were plotted using the threshold

Fig. 6. Walking speed vs density for different age demographics, from Kholshenikov data [1–3].

Fig. 7. Crowd flow rate vs density for different age demographics, from Kholshenikov data [1–3].

values between ‘calm’ and ‘active’ for the adult and pre-school children populations. The calm/active threshold was used, because they were already slightly on the lower end of the relaxed walking speeds quoted by other sources [6,15] for normal egress conditions, and seemed disproportionately lower if we’d used the midband of ‘calm’. Only ‘calm’ values were available for the elderly. Therefore the unimpeded walking speeds used for these formulae were 1.1 m/s (adult), 0.63 m/sec (elderly) and 0.75 m/s (pre-school

256

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

children). The peak flow rate (in Fig. 7) for the elderly demographic (1.06 p/m/s) is 41% less than that for the healthy adult flows (1.80 p/m/s). In addition, this peak elderly flow is 20% less than the typical design value of 1.33 p/m/s (80 p/m/min) discussed in the “introduction” section. Interestingly, although the pre-school children have a slower unimpeded walking speed, they achieve the highest flow rate: they have much smaller body sizes and are generally fit and healthy. Therefore, a crowded population, with a significant proportion of elderly members are likely to have at least a 20% lower flow rate, potentially higher. 6.2. The impact of body size on crowd speed and flow The impact of significantly higher proportions of obese people has yet to be properly considered with respect to crowd movement, however the findings from one study carried out by Spearpoint and Maclennan [40] stated that “…total evacuations times will increase as populations age and trend towards higher proportions of obese occupants. These increases may be of the order of 5–8% when comparing 2006 and 2031 New Zealand scenarios and up to 20% when comparing the 1971 Canadian with the 2031 New Zealand scenario…”. An estimated increase in evacuation time of up to 20% over 60 years begins to highlight the potential order of magnitude of impact of obesity on crowd flow. If we refer to the basic equation of flow in Eq. (1), where flow is considered to be derived from speed and crowd density, then we could begin to understand the effect of increasing the space occupied by each person, and the consequential influence on the measured crowd density. If we consider the findings from Paul's studies observing crowd movement with and without coats [7], this may be considered similar (from a spatial analysis perspective) to comparing ‘overweight’ people with slim people (but is likely to underestimate the full effect given additional issues of fitness, freedom of movement, secondary ailments, body sway, etc.). This comparison indicated that there is a 10–20% reduction in flow rate for people with coats on compared to those without coats. Fig. 8 contains a rather limited number of points but does suggest a difference between the mean evacuation flow (persons per second per meter of effective stair width) for those evacuations where coats were used. Between an evacuating population of approximately 200 and 450 persons per metre of effective width, the mean evacuation flow for a given evacuating population where

Fig. 8. Pauls' flow rates with/without coats reproduced from (Pauls [7]).

Area of horizontal projection of a person

Fig. 9. Flow ‘concentration’ for emergency / ‘normal’ conditions, and the area per person [4].

coats were used was lower than the respective mean evacuation flow where coats were not used. It should be added that obesity might not only change the body size, but potentially also other aspects of movement, e.g., body sway, movement speed, etc. It is, however, difficult, at this time, to estimate the influence of these related movement parameters. Predtechenskii and Millinskii [4] also observed the effect of flow, related to effective body size (expressed in the form of winter clothing), based on analysis of over 5000 trials. They also observed an opposite effect for smaller body sizes (children). However, in order to automatically cater for the size of people, these Russian researchers expressed density in terms of m2 occupied space per m2 free space, shown in Fig. 9. The flow rate was therefore also in different units, i.e. m/min. Predtechenskii and Millinskii also measured the ‘horizontal projection area’ of each person as being 0.1 m2 for summer dress and 0.125 m2 for winter dress. Therefore, if we choose to assess flow in terms of people/m/s, then the increase in body size from summer to winter dress may be concluded to result in a 25% drop in flow rate due to the larger overall dimensions of people in winter clothing. We should, however, note that the main limitation of this approach is that, while body size is taken into account to some extent, there is no explicit accommodation for related changes in speed, or gait or cadence. The indicators from Pauls [7] and Predtechenskii and Millinskii [4], who looked at differences between summer and winter clothing, are really the only indicators we have that illustrate the potential effect of an increase in body size on flow. In fact Pauls has questioned the validity and applicability of some of his own data (Pauls et al. [8]) due to population change and the rising rates of obesity. In attempting to quantify the effects of obesity in the population, rather than the moderate effects of changing from summer to winter clothing, our first approximation is to use the 25% indicator from the Russian studies. Clearly this should be studied further, but it is important to choose an empirically derived, indicative figure rather than none at all. The absolute impact of differing body sizes on flow is difficult to quantify since the dynamics of movement are more complex. Obesity, for example, has been shown to correlate with reduced movement speed (Hulens et al. [41]) and increased walking sway (He and Baker [42]), both of which are undoubtedly important parameters related to individual movement in a populated space. It is difficult therefore to understand fully the impact of body size per se on the flow dynamics since it is impossible to disaggregate the impact of larger body size and reduced speed on the flow dynamics without further research. Other studies involving children, i.e. small body sizes compared to adult populations, have produced flow/density relationships that show relatively higher

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

flows for given ‘densities’ than the standard adult curve (e.g. as suggested in the Society of Fire Protection Engineers handbook [43] and shown by Khoshelnikov et al. [1]). However, the ‘children’ demographic encapsulates different body dimensions and higher walking speeds (sometimes running) speeds, increasing the flow rate at higher densities. The demographic represents a good example of the combined effects of changes in body sizes and speed, and is provides a useful comparison for the dynamics of mixedpopulation groups.

7. Evacuation analyses: the ‘evolution’ of computer models The sophistication of the simulation models used for evacuation analysis has evolved over time. A number of detailed reviews of simulation models and their characteristics exist in the literature (Kuligowski and Peacock [43]; Gwynne and Galea [44]). The modelling approaches have taken 3 chronological forms: 1. Flow network models (or networks of ‘nodes’) – where each room or space ‘node’ in a building has a given capacity of people, and each node is connected with passageway or door ‘arcs’ with a set capacity and maximum sustained flow rate. These models were first developed in the 1980s, and grew in complexity and capability from the 1980s to the 1990s (BFIRES (Stahl [45]), EVACNETþ (Kisko and Francis [46]), EXIT89 (Fahy, [47]) etc.). 2. Grid models – where the building spaces are subdivided into a grid or ‘mesh’ of approximately body-size plan area elements (0.5  0.5 m2). Movement happens on a ‘stepping stone’ basis as each simulated person transits from one grid cell to another. These models were first developed in the 1990s and continue to be developed and used today in the more sophisticated forms, e.g., Egress (Ketchell and Cole [48]), building EXODUS (Galea et al. [49]), STEPS (Mott McDonald [50]) etc. 3. Continuous models – where occupant movements through building spaces are not restricted to a simple mesh form, but instead movement is ‘continuous’ through the geometrical forms of the building, and movement ‘vectors’ represent occupant progression through the Cartesian coordinates of a model. These models originated in the late 1990s and most new models use different mathematical approaches to represent ‘continuous’ movement models (Pedgo [51], Legion [54], Simulex [55], PathFinder [56], MassMotion [58], and FDS þEvac [59]). It should be noted that whilst there are a wide range of calculation approaches in these ‘continuous models’, most movement algorithms are based on localised assessments of speed or flow vs density derived from research published in the 1960s and 1970s. The evacuation simulation packages tend to use the speed/ density or flow density relationships derived from work published by researchers such as Hankin and Wright [5] and Togawa [60] in the 1950s, Predtechenskii and Millinskii [4] in the 1960s, Pauls [61] and Fruin [6] in the 1970s and Ando [15] in the 1980s. In essence, the basic ‘flow’ approach still forms the basis of the simulation approaches (grid and continuous) although in a number of cases the flow rates are varied with the density of the standard ‘crowd’ over smaller unit areas. Nearly all of the movement algorithms in these three ‘generations’ of computer models assume uniform crowd flow parameters, except the spatial-proximity models (e.g. Simulex [52] and Legion [54]) and the ‘social forces’ models (e.g. PedGo [57] and FDS þEvac [59]), which still tend to be calibrated against the traditional flow data. It should also be noted that the more advanced computer simulation packages have the ability to show very

257

impressive 3D visual representations of the egress or escape movement. The user may therefore assume a more sophisticated level of rigour in the underlying movement algorithms and data than is actually implemented.

8. The detailed analyses of movement It is important to note the main points of reference for these simulation models, other than fixed flow rates. The ‘grid models’ and many ‘continuous models’ use the relationship between crowd speed and density (such as those shown in Fig. 10) or crowd flow and density (Fig. 11). The spatial proximity models, on the other hand, use a relationship between inter-person distance and speed. Fig. 12 shows how the approximated ‘circular’ spacing (observed by Fruin [6] in aggregated crowd groups) may be used to mathematically (and geometrically) derive average linear inter-person distances at given densities. The inter-person distances are then used to derive the velocity/distance curves illustrated in Fig. 13. Continuous models can use these types of relationships between velocity and inter-person distance, and one example of such a model is Simulex [22]. In Simulex, each person in a given simulation is assigned a normal unimpeded walking speed (representative of age, gender or other effect). The relationship between walking velocity and inter-person distance is scaled such that the unimpeded walking velocity is achieved at a given threshold distance, beyond which the presence of one (obstructing) person does not quantifiably affect another person further behind. Implementations of the Simulex approach are illustrated in Fig. 14. When this relationship is plotted as density, using the equations described in Fig. 12, the velocity/density relationship is very similar to that plotted by Ando et al. [15] and shown in Fig. 10. This approach was also verified by video-based data collections at 4 different sites in Edinburgh in 1994 [55]. The implementation of the relationship between walking velocity and inter-person distance (Fig. 14) enables Simulex to make some attempt to model the effects of variations in walking speed amongst a constituent population. Fig. 15 illustrates the plan representation of the physical size of people as used in the Simulex model (other continuous models might use similar body representation). This means that Simulex can adjust the radii of body ‘circles’ in order to attempt to reproduce some of the effects of different body dimensions e.g. ‘obese’ adult bodies or children. Only ‘continuous’ models such as Simulex can, in their current

Fig. 10. The relationship between crowd velocity and density (from Thompson [55]).

258

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

Fig. 13. Reducing aggregated flow to velocity and inter-person distance (from Thompson [55]). Fig. 11. The relationship between crowd flow and density reproduced (from Thompson [55]).

form, be used to easily accommodate the effects of population demographics, such as body size, on movement and flow. The results of simulations of some of these prototype mixed-population groups are presented in the following section.

9. Simulating different population demographics with Simulex Simulex can, similar to many other continuous models, accommodate individual population size and speed definitions because of the way in which movement within crowded spaces is assessed. Thompson et al. [62] have previously used the computer simulation package Simulex to incorporate explicit definitions of different population demographics to quantify the effects of elderly people (as stated in the IMO regulations [23]) and body size, requested at the time by the Swedish Maritime Authority (Sjöfartsverket or ‘SFV’) for lifejacket tests. Table 2 shows the flows generated in the simulations in Simulex for the pre-defined populations, i.e., combinations of people with associated movement speeds and body sizes. Each population group is ‘discrete’ i.e. they only contain characteristics within the defined parametric bands. As can be seen in the table, the flow changes as a function of the

Fig. 14. Relationship between walking velocity and inter-person distance in Simulex [55].

population composition. It should be pointed out that this type of simulation exercise could potentially have been done with many other continuous models, and Simulex is used here mainly to exemplify. We should be cognisant of the limitations of these simulations,

Converting circular ‘spacing’ to inter-person distance and pedestrian density… Lateral Spacing

Fig. 12. Reducing approximate radial spacing to inter-person distance (from Thompson [55]).

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

259

accommodation of the mixed ability movement that is associated with elderly passengers in particular. The increase in body size produces a slightly lessened drop in flow rate when compared to Pauls, and Predetechenskii and Millinskii because only the torso is adjusted, and only by the width of a modern life jacket. There is also no allowance for a possible reduction in walking speed associated with obese and morbidly obese physiology.

10. Discussion: calculations, models and design We have considered two primary demographic trends that are predicted to affect crowd flow rates and general speed of movement:

Fig. 15. The representation of body size (on plan) and ‘contact distance’ in Simulex [55].

Table 2 Simulex results for International Maritime Organization (IMO) exit flow tests [62]. IMO Population Group (s)

Peak flow (people/s)10 sec sample

Sustained flow (people/ m/s)

Commuter IMO Crew Elderly Passengers SFV Adults SFV Adults(þ life jacket)

1.4 1.6 1.0 1.4 1.2

1.36 1.58 0.88 1.28 1.12

namely that they are not paired with corresponding ‘real-life’ studies, but they do give some indication of the mathematical implications of changing ‘body size’ and also trends associated with implementing the reduced walking speeds for ‘elderly’ passengers in these simulations. The “Commuter”, “Crew”, and “Elderly Passengers” demographic definitions are defined in the International Maritime Organization (IMO) document “MSC Circ 1238: Guidelines for a Simplified Evacuation Analysis for New and Existing Passenger Ships” [23]. Outcomes of these simulations that are of particular interest are: a. The baseline “Commuter” population set recreated a flow rate of 1.36, which was very close to the regulatory ‘standard’ of 1.33 for fit, healthy adults. b. The “elderly” population defined by the IMO (with a small accommodation for mixed-ability) showed a 35% reduction in flow rate when compared to the baseline “Commuter” population. c. The “SFV Adults” population set showed an 11% reduction in flow rate by just adding the indicative thickness of a life jacket to the torso of each person, compared to the unit tests carried out with the same population in normal clothing. The effect of the “elderly” population demonstrating a 35% reduction in flow rate may be more than that projected from Ando's data because the defined demographics also take some

a. The ‘ageing society’, which is predicted to lead to a much higher proportion of elderly across the range of OECD countries, with some countries predicted to have over 60% of the adult population over 65 years of age by 2050. If we use the numbers from the analysis of the increase in the elderly proportion of adults in society, combined with Ando's analysis of walking speeds for the over 65s then we might use the 20% reduction in speed, as an indicator for an impact on flow rate discussed in the earlier sections. The data from Kholshenikov [2,3] indicates a potential drop of around 40%, but gives an actual peak value which is 20% less than the typical value of 1.33 p/m/s. The Simulex simulations produced a more significant reduction in flow rate of 35%, but we would suggest that further tests are required, and the IMO population definitions compared to the general population before using the results from these theoretical computer simulations more widely. For egress models using specified flow rates, we would therefore nominate 20% as a figure to consider for a general reduction in speed and flow parameters to accommodate an ‘elderly’ population. b. The ‘obesity epidemic’ and the tendency across many modern societies for an increasing level of ‘overweight’ population members. We used three points of reference for reduction in walking speeds: Pauls’ winter clothing tests (10–20% reduction for wearing winter clothes), Predtechenskii and Millinskii’s analysis of larger body size for heavier clothing (25% reduction in flow), Spearpoint and Maclennan [40] figure of 20%, and Simulex tests with life jackets around the torso (11%). For egress models using specified flow rates, we would therefore nominate 20% as an indicative figure to accommodate an ‘obesity-dominated’ crowd flow, as we would expect the effects of obesity to be at least as much as wearing heavy clothes, and it is likely that there would also be a reduced walking speed for the more pronounced cases of obesity. We may consider the compound effects of these two demographic trends in the following equation:

qd = q × ( Fe × Fo )

(3)

Where qd is a nominal flow rate for design or simulation purposes, q is the normal unadjusted flow rate for a ‘standard’ population, Fe is an adjustment factor for ‘elderly dominant’ flow, and Fo is an adjustment factor for ‘obesity-dominant’ flow. From the numerical values suggested above, i.e. Fe ¼1–0.2 ¼0.8 and Fo ¼ 1  0.20 ¼0.8, and the most widely accepted ‘standard’ flow rate of 1.33 p/m/s, therefore:

qd = 1.33 × (0.8 × 0.8) = 1.33 × 0.64 = 0.85

(4)

From these calculations, it is suggested that we could consider a nominal flow rate figure of 0.85 p/m/s (36% less than 1.33 p/m/s)

260

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

in order to accommodate the population demographic trends predicted by the OECD for 2050. Indeed, some reduction factor should be applied to current design practices as a large degree of this change has already occurred. The above calculation has been carried out not to suggest yet another simple, single ‘catch-all’ number, for all designers for all buildings to use going forward. It is merely an example calculation, which is used to identify potential implications in simple terms. In reality, the population of different building demographics should be considered. It could be argued that similar factors should be applied to buildings with mixed-ability populations or mobility impaired occupants at present. Such factors could be used to adjust the algorithms of computer models with pre-defined flow characteristics, simply by applying reductions to the speed or density curves. The simulation would then demonstrate comparative trends in flow. This would not apply to ‘continuous’ movement models, which would instead have individual characteristics updated, and incorporate algorithms which realistically implement the effects of those characteristics.

11. Additional factors It is recognised that in future, additional factors that may affect crowd flow rates, and the ability to successfully navigate to an exit, will need to be considered. This paper has focused primarily on the need for computer models to accommodate demographic change, but the general ethos is to adopt more flexible calculations and step away from a single figure which is supposed to “fit all” cases. The following aspects may also be considered in the future: 1. Gait Biomechanics: a more detailed analysis of the biomechanics of walking is required to more fully understand the essential mechanisms at play. Factors such as leg length, stride length, body sway, gait patterns, walking aids etc. could all potentially be considered in future. Also, in a population biased towards elderly ages and mixed-ability, it may be prudent to consider that an increasing number of occupants may need assistance (either with walking aids or assisting staff) to reach an exit. The parametric quantification of such factors would be required before simulation is attempted, but the inherent complexities should not by any means lead to the avoidance of such considerations. It is likely that the computer simulation approaches need to further advance the ‘continuous models’ approaches to look at the gait patterns of inter-person behaviours and potential points of contact at high densities where ‘crushing’ may become a risk. 2. Fatigue: the potential reduction in walking speed encountered after navigating a long escape route, potentially encountered in a high rise building may be important, especially if the people involved are either elderly, obese, or suffering from other motor-sensory impairments. 3. Smoke Toxicity: some computer simulation models do incorporate the Fractional Effective Dose Model [63] and a potential effect of smoke toxicity on walking speed. However, such effects on health could also be considered in the context of the demographics of the population. For example, elderly people may be much more susceptible to the effects of smoke inhalation. There are a number of implications for building design. If this form of numerical analysis provides a reasonable indication of the performance of building elements to accommodate egress flows, then we will need wider passageways and doors to allow for the modified flow capacity of those elements. Such adjustments should consider the nature of the building populations now and in

the future. Simply increasing the width of passageways should not be the only consideration – we should be aware of the different cognitive and mobility aspects associated with the changes in age, obesity levels, and increased proportions of people with ‘mixedabilities’ for navigating a built environment and arriving at a place of safety. The design guides should be reviewed and the computer models updated, and indeed we may also need to consider how procedural measures can be implemented to ensure the safe evacuation of all.

12. Conclusions Today, the representation of crowd movement in existing evacuation models is typically based on old data that do not accurately account for the increasing proportion of elderly and obese people. Therefore crowd ‘flow’ models and design approaches need to be re-examined. As highlighted in this paper, the proportions of elderly and obese have increased in the last 40 years and these trends are expected to continue 40 years into the future. It is argued that these changes will results in people moving more slowly and, in the case of obesity, reduce the achievable density of crowds in terms of people per square metre. Flow, which is the product of speed and density, may therefore have been reduced in the past 40 years and is expected to reduce further in the coming 40 years. This has significant implications on the fire and life safety of existing buildings, which have been designed based on old data and may in the future take a longer time to evacuate than initially intended. Also, old data in evacuation models affect the accuracy of present design calculations for buildings. This paper has highlighted the problem of an increasing proportion of elderly and obese people in society, but more research is needed to resolve this potential problem. It is argued that the best way forward is to re-examine crowd movement taking people interaction into account, which is a bottom-up approach for a fundamental understanding of crowd movement. It is suggested that commonly used flows might be over predicted by as much as 36%. This reduction factor (or a derivation thereof) could be used to modify design or regulatory guidance in the meantime, until we have developed a more accurate understanding of the complex interactions of biomechanics and predicted population demographic change.

References [1] V.V. Kholshchevnikov, D.A. Samoshin, A.P. Parfyonenko, I.R. Belosokhov, Study of children evacuation from pre-school education institutions, Fire Mater. 36 (2012) 349–366. [2] V.V. Kholshchevnikov, D. Samoshin, R. Istratov, The problems of elderly people safe evacuation from senior citizen heath care buildings in case of fire, in: Proceedings of the 5th International Symposium "Human Behaviour In Fire" 2012, Cambridge, UK, pp. 587–593. [3] V.V. Kholshevnikov, T.J. Shields, K.E. Boyce, D.A. Samoshin, Recent developments in pedestrian flow theory and research in Russia, Fire Saf. J. 43 (2008) 108–118. [4] V.M. Predtechenskii, A. Milinskii, Planning for Foot Traffic in Buildings (translated from the Russian), Stroizdat Publishers, Moscow, 1969, English translation published for the National Bureau of Standards and the National Science Foundation, Amerind Publishing Co., New Delhi, India, 1978. [5] B.D. Hankin, R.A. Wright, Passenger flow in subways, Oper. Res. Q. 9 (1958) 81–88. [6] J.J. Fruin, Pedestrian planning and design, Metropolitan Association of Urban Designers and Environmental Planners Inc., New York, 1971. [7] J. Pauls, Movement of people, in: P.J. DiNenno, C.L. Beyer, R.L.P. Custer, W. D. Walton, J.M.W. Watts, D. Drysdale, J.R. Hall (Eds.), The SFPE, Handbook Of Fire Protection Engineering, 2nd ed., National Fire Protection Association, Quincy, MA, 1996, pp. 3-263–3-285. [8] J. Pauls, J. Fruin, J. Zupan J, Minimum stair width for evacuation, overtaking movement and counterflow, Technical bases and suggestions for the past, present and future, in: Waldau Gatterman, Knoflacher Schreckenberg (Eds.), Pedestrian and Evacuation Dynamics, Springer Verlag, Berlin, 2005, pp. 57–69. [9] United Nations, “World Population Aging 2013”, United Nations Department of

P. Thompson et al. / Fire Safety Journal 78 (2015) 251–261

[10]

[11] [12]

[13]

[14]

[15] [16]

[17]

[18]

[19]

[20]

[21]

[22] [23] [24]

[25] [26] [27] [28] [29]

[30]

[31]

[32]

[33]

[34] [35]

Economic and Social Affairs Population Division, 2013, accessed at: 〈http:// www.un.org/en/development/desa/population/publications/pdf/ageing/World PopulationAgeing2013.pdf〉. United Nations, “World Population Prospects: The 2012 Revision, Key findings and Advance Tables”, Department of Economic and Social Affairs, Populations Division, United Nations, New York, 2012. OECD, “Obesity Update 2012”, Organisation for Economic Co-operation and Development, 2012, accessed at 〈http://www.oecd.org/health/49716427.pdf〉. E.D. Kuligowski, R.D. Peacock, P.A. Reneke, E. Wiess, C.R. Hagwood, K.J. Overholt, R.P. Elkin, J.D. Averill, E. Ronchi, B.L. Hoskins, M. Spearpoint, “Movement on Stairs During Building Evacuations”, NIST Technical Note 1839, National Institute of Standards and Technology, USA, 2015, available online at: http:// dx.doi.org/10.6028/NIST.TN.1839. J.L. Pauls, Building evacuation: research findings and recommendations, in: D. Cantor (Ed.), Fires and Human Behaviour, John Wiley & Sons, New York, 1980, pp. 251–275. R.D. Peacock, B.L. Hoskins, E.D. Kuligowski, Overall and local movement speeds during fire drill evacuations in buildings up to 31 stories, Saf. Sci. 50 (8) (2012) 1655–1664. K. Ando, H. Ota, T. Oki T, Forecasting the flow of people, Railw. Res. Rev. 45 (1988) 8–14 (Japanese). K.E. Boyce, T.J. Shields, G.W.H. Silcock, Towards the characterization of building occupancies for fire safety engineering: capabilities of disabled people moving horizontally and on an incline, Fire Technol. 35 (1) (1999) 51–67. Home Office/Scottish Home and Health Department, “Guide to Safety at Sports Grounds (Football)” (also known as the ‘Green Guide’), third ed., HMSO, UK, 1985, ISBN 011 3407610. H.E. Nelson, H.A. MacLennan, Emergency movement, in: P.J. DiNenno, et al., (Eds.), The SFPE Handbook of Fire Protection Engineering, 2nd ed., National Fire Protection Association, Quincy, MA, 1996, pp. 3-286–3-295. British Standards Institution, “PD 7974-6:2004, The application of fire safety engineering principles to fire safety design of buildings. Human factors. Life safety strategies. Occupant Evacuation, behaviour and condition (Sub-system 6)”, British Standards Institution, London, UK, 2004. Home Office, “Report by SCICON Ltd. on Safety in Football Stadia: A Method of Assessment, Lord Justice Taylor's Final report, Technical Working Part on the Guide to Safety at Sports Grounds”, Home Office, UK, 1989. DCLG, “The Building Regulations 2000 Fire Safety Approved Document B Volume 2 Buildings Other than Dwelling Houses, 2006 Edition”, Department of Communities and Local Government, 2007 (amended). NFPA 101, Code for Safety to Life from Fire in Buildings and Structures, U.S. Dept. of Commerce, Washington DC, 2012. International Maritime Organisation, Guidelines For Evacuation Analysis For New and Existing Passenger Ships, MSC.1/Circ.1238, London, October 2007. Cornell University, “2012 Disability Status Report”, Employment and Disability Institute, 64 pp. found at: 〈http://www.disabilitystatistics.org/StatusReports/ 2012-PDF/2012-StatusReport_US.pdf〉, 2012. S.J. Older, Movement of pedestrians on footways in shopping streets, Traffic Eng. Control 10 (1968) 160–163. A. Polus, J.L. Schofer, A. Ushpiz, Pedestrian Flow and Level of Service, J. Transp. Eng. 109 (1983) 46–47. J.L. Pauls, in: D. Canter (Ed.), Fires and Human Behaviour, John Wiley & Sons Ltd., London, 1980, pp. 227–275 , Chapters 13 and 14. OECD, Society at a Glance: OECD Social Indicators, OECD Publishing, France, 2006. J.A. Martin, B.E. Hamilton, M.J.K. Osterman, S.C. Curtin, M.A. Matthews, T.J. Matthews, Births: final data for 2012, National Vital Statistics Reports volume 62, number 9, US Department of Health and Human Services, National Center for Health Statistics, 2013. Elizabeth Arias, United States Life Tables, 2006, National Vital Statistics Reports, vol. 58, number 21, 2010, accessed at 〈http://www.cdc.gov/nchs/data/ nvsr/nvsr58/nvsr58_21.pdf〉. UK Parliament Research publications, Have kids, settle down: marital and maternal age since 1938, 2013, accessed at 〈http://www.parliament.uk/busi ness/publications/research/olympic-britain/population/have-kids-settledown/〉. European Commission, “2015 Ageing report: economic and budgetary projections for the 28 EU Member States (2013-2060)”, 2015, ISBN 978-92-7944747-1. N.D. Reeves, M. Spanjaard, A.A. Mohagheghi, V. Baltzopoulos, C.N. Maganaris, The demands of stair descent relative to maximum capacities in elderly and young adults, J. Electromyogr. Kinesiol. 18 (2008) 218–227. H.G. Kang, J.B. Dingwell, Effects of walking speed, strength and range of motion on gait stability in healthy old adults, J. Biomech. 41 (2008) 1567–1573. T. Fujiyama, T. Tyler, Free walking speeds on stairs: effects of stair gradients and obesity of pedestrians, in: Peacock, et al., (Eds.), Pedestrian and Evacuation Dynamics, Springer Science and Business Media, 2011, pp. 95–105.

261

[36] B.B.C. News, Fat patients 'prompts ambulance fleet revamp', in BBC News/ Health 2011, online at: 〈http://www.bbc.co.uk/news/health-12287880〉. [37] World Health Organization, Obesity and Overweight, WHO Fact sheet N°311, 2015, available online at: 〈http://www.who.int/mediacentre/factsheets/fs311/ en/〉. [38] Cheryl D. Fryar, Margaret D. Carroll, and Cynthia L. Ogden, Prevalence of Obesity Among Children and Adolescents: United States, trends 1963–1965 through 2009–2010, NCHS Health E-stat, National Center for Health Statistics, Hyattsville, MD. accessed at: 〈http://www.cdc.gov/nchs/data/hestat/obesity_ child_09_10/obesity_child_09_10.htm〉. [39] J. Levy, L.M. Segal, K. Thomas, R. Laurent St, A. Lang, J. Rayburn, F as in Fat: How Obesity Threatens America's Future 2013, Trust for America’s Health/ Robert Wood Johnson Foundation, USA, 2013. [40] M. Spearpoint, H.A. MacLennan, The effect of an ageing and less fit population on the ability of people to egress buildings, Saf. Sci. 50 (2012) 1675–1684. [41] M. Hulens, G. Vansant, A.L. Claessens, R. Lysens, E. Muls, Predictors of 6-minute walk test results in lean, obese and morbidly obese women, Scand. J. Med. Sci. Sports 13 (2003) 98–105. [42] X.Z. He, D.W. Baker, Body mass index, physical activity, and the risk of decline in overall health and physical functioning in late middle age, Am. J. Public Health 94 (9) (2004) 1567–1573. [43] E.D. Kuligowski, R.D. Peacock, A Review of Building Evacuation Models, NIST Technical Note 1471, US Government Printing Office, Washington (2005), p. 156. [44] S. Gwynne, E.R. Galea, A Review of Methodologies and Critical Appraisal of Computer Models Used in the Simulation of Evacuation from the Built Environment, Society of Fire Protection Engineers, Bethesda, Maryland, USA, 2004. [45] F. Stahl, Final Report on the 'BFIRES/Version 1' Computer Simulation of Emergency Egress Behaviour During Fires: Calibration and Analysis, NBSIR 791713, US Department of Commerce, NBS, Center for Building Technology, Washington DC, March 1979. [46] T.M. Kisko, R.L. Francis, EVACNET þ: a computer program to determine optimal building evacuation plans, Fire Saf. J. 9 (2) (1985) 211–220. [47] R.F. Fahy, EXIT 89: An evacuation model for high-rise buildings, in: Proceedings of the 3rd International Symposium of Fire Safety Science, University of Edinburgh, July 8–12 , Elsevier Science Publishers, London. 1991, pp. 815– 823. [48] N. Ketchell, S. Cole, “The EGRESS code for human movement and behaviour in emergency evacuations”, in: R.A. Smith, J. Dickie (Eds.), Engineering for Crowd Safety, Elsevier, Amsterdam, 1993, ISBN 0 444 899200. [49] E.R. Galea, J.M. Perez Galparsoro, J. Pearce, A brief description of the EXODUS evacuation model, in: Proceedings of the18th International Conference on Fire Safety, San Francisco, USA, vol.18, 1993, pp.149–162. [50] Mott MacDonald, STEPS Simulation of Transient Evacuation and Pedestrian Movements User Manual, Mott McDonald, 2003 (unpublished work). [51] H. Klupfel, T. Meyer-Konig, Characteristics of the PedGo Software for Crowd Movement and Egress Simulation, in: Proceedings of the 2nd International Conference in Pedestrian and Evacuation Dynamics (PED), University of Greenwich London, UK, 2003, pp. 331–340. [52] P.A. Thompson, E.W. Marchant, A computer model for the evacuation of large building populations, Fire Saf. J. 24 (1985) 131–148. [54] Legion International: accessed at: 〈http://www.legion.com/legion-software〉. (on-line). [55] P.A. Thompson, Developing New Techniques for Modelling Crowd Movement, Ph.D., Department of Building and Environmental Engineering, University of Edinburgh, Scotland, 1995. [56] J. Cappuccio, Pathfinder: a computer-based timed egress simulation, Fire Prot. Eng. 8 (2000) 11–12. [57] TraffGo “Product Information - PedGo Editor” Pamphlet, 2005. [58] AEC Magazine, “Oasys MassMotion”, 5 July 2011, accessed at: 〈http://aecmag. com/software-mainmenu-32/445-oasys-massmotion〉. [59] T. Korhonen, S. Hostikka, Fire Dynamics Simulator with Evacuation: FDSþ Evac Technical Reference and User's Guide, VTT, Finland, 2009, ISBN 978-951-387180-2. [60] K. Togawa, Study of Fire Escapes Basing on the Observation of Multitude Currents, Report no. 14, Building Research Institute, Tokyo, 1955. [61] J.L. Pauls, in: D. Canter (Ed.), Fires and Human Behaviour, John Wiley & Sons Ltd, London, 1980, pp. 227–275 , Chapters 13 and 14. [62] P. Thompson, H. Lindstrom, P.A. Olsson, S. Thompson, “Simulex: analysis and changes for IMO compliance”, in: Proceedings of the Second International Conference in Pedestrian and Evacuation Dynamics (PED), University of Greenwich, London, 2003. [63] D.A. Purser, Toxicity assessment of combustion products, in: P.J. DiNenno, et al., (Eds.), SFPE Handbook of Fire Protection Engineering, 3rd ed.,National Fire Protection Association, Quincy, MA, 2002, pp. 2-83–2-171.