You are on page 1of 37

free ebooks ==> www.ebook777.


Employing the Hydraulic Model

in Assessing Emergency Movement 59
Steven M.V. Gwynne and Eric R. Rosenbaum

Introduction RSET (i.e., the time taken to reach safety) and

therefore complete this component of a
This chapter provides the engineer with a model performance-based assessment. This model,
to quantify egress performance. This model is albeit imperfect, quantifies the egress perfor-
formed from a set of numerical tools that vary mance of a design and, importantly, enables
in their scope and sophistication. Guidance is comparisons to be made between different
provided on the capabilities of these tools and design variants to be made.
on when they should be employed, making refer- The hydraulic model is presented as a means
ence to the data on which these tools are based. of quantifying egress performance that can sup-
Detailed examples are presented to clarify the port an engineering approach and expert analy-
application of these tools, along with a descrip- sis. Hydraulic models are based on a
tion of how the use of these tools fits in with other simplification of egress behavior where the
fire engineering calculations. This chapter will, evacuating population is described by a set of
therefore, allow the engineer to assess egress equations. This population moves from egress
performance in a responsible and informed component to egress component (e.g., from a
manner. corridor to a stairwell), with the speed of their
Prediction of evacuee movement is an essen- movement dictated by the equations that form the
tial component of performance-based fire safety model. Guidance is provided on how best to
analysis. Safe egress from fire is assumed to be employ these equations, on the scenarios to
achieved if the required safe egress time (RSET) which this model can be applied, and on the
is sufficiently shorter than the available safe limitations of the approach.
egress time (ASET), where ASET is defined as The inherent structure of the hydraulic model
the time until fire-induced conditions within a described in this chapter tends to an optimistic
building become untenable. Methods to evaluate estimate of evacuation time. It assumes that the
the development of fire-induced conditions and exit paths will be continually used at maximum
tenability criteria are addressed elsewhere in this capacity from the moment of alarm to total evac-
handbook. uation. The model should be considered as a
The model discussed in this chapter baseline calculation to be extended as appropri-
provides the engineer with a means to establish ate to account for delays caused by human
decisions, notifications, and other factors (see
S.M.V. Gwynne (*) Chaps. 58 and 64).
National Research Council Canada For each evacuee the RSET can be subdivided
E.R. Rosenbaum into a number of discrete time intervals, the sum
Jensen Hughes, Inc. of which constitute the total RSET:

M.J. Hurley (ed.), SFPE Handbook of Fire Protection Engineering, 2115

DOI 10.1007/978-1-4939-2565-0_59, # Society of Fire Protection Engineers 2016
free ebooks ==>
2116 S.M.V. Gwynne and E.R. Rosenbaum

RSET ¼ td þ tn þ t pe þ te ð59:1Þ assumed that for an individual evacuee, tpe

and te are basically sequential. There is a period
where before the individual has determined that an
td ¼ Time from fire ignition to detection; that is, evacuation response is required (through the per-
the detection phase ception of sufficient risk levels and the
tn ¼ Time from detection to notification of subsequent completion of preparatory actions)
occupants of a fire emergency; that is, the and a period where this response is conducted
notification phase (i.e. where protective actions are taken). That is
tpe ¼ Time from notification (or cue reception) not to say that the individual need be static in
until evacuation commences; that is, the either time period or that the performance of
pre-evacuation phase actions are not iterative or cyclical—only that
te ¼ Time from the start of purposive evacuation at a certain point in time, the individual decides
movement until safety is reached; that is, the that the situation requires them to take protective
evacuation phase actions and their subsequent actions broadly
The components described are considered the reflect an attempt to disengage from the current
core elements of egress analysis (see Chap. 64 actions and take protective actions.
for further discussion), although it is recognized However, across a population, tpe and te are
that other components will certainly contribute to neither independent of each other nor mutually
evacuee performance (e.g. the Pre-Warning exclusive [1]. There may be significant overlap
delay incurred through staff actions and their between these components given the varying
decision-making which may prolong notification conditions evident at different locations within
of the general population). the structure, the different levels of information
The RSET elements td and tn typically involve available, and the differences in the abilities of
a technical solution and human interaction, the population [2].
including fire detection devices and fire alarm RSET can be reduced into two sets of
equipment, and also human intervention, such components: the phase prior to evacuee involve-
as the discovery of a fire by a staff member and ment, made up of td and tn, and the escape phase
notification of the population. The theory and (tesc) where
design of detection systems are covered else-
where in this handbook (see Chap. 40). tesc ¼ t pe þ te ð59:2Þ
The element tpe relates to the individual and
collective responses of the occupants; that is, the It should be noted that, in reality, the evacuation
time between them being notified of the incident phase can be interrupted through behavioral
and the time to commence their evacuation. This actions and developments in the incident
can be prolonged by a number of complex scenario [3].
activities (see Chaps. 58 and 64). These include This chapter describes the basic hydraulic
receiving a cue; interpreting the cue; validating model enabling te to be calculated. It also
the cue; performing pre-evacuation activities; describes the extension of the hydraulic model
and determining an appropriate response (see to also include tpe in the calculation and there-
Chap. 58). All of these contribute to the time fore allow an estimation of tesc to be produced. A
spent in the pre-evacuation phase, prior to com- methodology is presented to enable the engineer
mencing purposive evacuation movement to a to determine the RSET value as part of a
place of safety. performance-based assessment. It provides suffi-
The element te is the time from when an cient information for the engineer to calculate
individual initiates evacuation movement up to RSET under a number of different incident
the point that he or she reaches safety. For an scenarios, while also making the engineer aware
individual evacuee, tpe and te are basically of the limitations and assumptions of the hydrau-
sequential. Crudely speaking, it is typically lic model (Fig. 59.1).
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2117

Approaches to modeling
emergency movement

Model capabilities and


First order—Focus on critical component

Components of the basic
Second order—Perform more complete
hydraulic model

Example applications of
the hydraulic model

Examining scenarios using the Basic approach—Calculate te

extended hydraulic model Extended approach—Calculate tesc

Understanding and employing safety

factors with the hydraulic model

Modeling the impact of

tenability criteria

Fig. 59.1 Structure of this chapter

competitive; and it would involve immediate and

Establishing Egress Performance direct movement once the incident was discov-
ered. Although some of these assumptions are
Over the last few decades an increasing effort has contradictory, they have had a direct impact on
been made into investigating human behavior in the engineering calculations made for a long
response to fire. This research has provided a period of time and continue to exert some degree
clearer understanding of egress behavior and of influence on egress design decisions to this
the factors that influence egress performance. day [3].
As a consequence, human behavior can be In recent times, a more detailed and compre-
taken into consideration when designing emer- hensive understanding of human behavior in fire
gency procedures and modeling human perfor- has been established (see Chaps. 58 and 64). This
mance. Prior to this time, human behavior was understanding has been derived from the exami-
disregarded altogether, seen as immeasurable, nation of actual incidents, the collection of
and/or drastically simplified according to a few empirical evidence, and the development of
basic assumptions. The former understanding of behavioral theories. All of this has, to a large
human behavior was based on a number of degree, refuted the assumptions that had previ-
assumptions: people’s behavior would likely be ously dominated. This realization has allowed
panic based [4]; it would likely be selfish and engineers (as well as behavioral researchers,
free ebooks ==>
2118 S.M.V. Gwynne and E.R. Rosenbaum

Procedural Environmental / Individual

Active fire protection Scenario-Based Cognitive abilities
Emergency signage Presence of fire effluent Language/culture
Notification system Background pollution Exposure to cues
(noise/visual) Behavioral Responses
Emergency training Location
Lighting levels Investigate/search
Emergency literature Fatigue
Debris Wayfind
Performance of fire drills General health
Presence of fatalities Evacuate
False/inaccurate alarms Sensory/cognitive impairment
Structural damage “Panic”
Staff/fire warden Size
Loss of routes Re-enter
Organizational Information levels Remain/delay
Architectural/Structural Fight fire
Safety culture Familiarity
Building type Collect items
Normal use of structure Role
Physical dimensions Secure item
Security procedures Responsibility
Geometry of enclosure Communicate
Communication system Age
Number and arrangement Process information
Existence of social hierarchy Gender
of egress routes Provide assistance
Distribution and size Activity
Complexity of space Exhibit nonevacuation
of population Social affiliation
Visual separation behavior
Nature of population Engagement
Lighting and Seek refuge
(non-emergency) signage Commitment
Defend in place
Extent of passive fire Physical abilities/limitations
protection Proximity to incident

Fig. 59.2 Factors that can influence egress performance

procedural designers, and evacuation modelers) an engineer prior to the application of such
to take human behavior into account, albeit methods.
imperfectly, when trying to establish evacuation
It is now felt that the evacuation process is not Models
simply a matter of initiating an evacuation and
then controlling the ensuing hysterical crowd Several approaches are available to the engineer
response; instead, it is now viewed as a more to establish egress performance; that is, estimate
multifaceted event in which people’s responses tesc. Each of these approaches requires the appli-
are sensitive to the incident scenario, the infor- cation of a model: a simplified version of reality
mation available, and the local conditions used as an indicator of actual egress perfor-
(among other things). The problem of under- mance. All of these approaches are limited. One
standing human behavior in fire is not the simple or more of the following four model approaches
process previously assumed. Instead, it relies on are usually applied:
a number of factors that can interact and can • Model Approach A: The application of pre-
influence the outcome in different ways scriptive codes. The expertise embedded
(Fig. 59.2). These factors also influence the engi- within the regulations is assumed to satisfac-
neering methods required to assess performance. torily represent (or at least account for) the
Ideally these factors should be considered in any performance of the evacuating population.
assessment of egress performance; however, the Generally, these codes focus on the physical
methods employed in this assessment are limited constraints imposed by the structure and
and, to different degrees, exclude many of the exclude behavioral and procedural factors.
key factors influencing egress performance. It is • Model Approach B: The performance of an
critical that these limitations are understood by egress trial. An (un)announced trial is
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2119

Factors that actually influence Factors that can be Factors that can be Factors that can be
occupant performance during > empirically or theoretically > represented by sophisticated > represented within the
an evacuation supported computer models hydraulic approach

Fig. 59.3 Difference between the actual and modeled evacuation

conducted in order to assess the outcome of a the performance of the population when tra-
simulated incident using the population of versing the component in question. These
interest. This approach has a number of results are then pieced together to form a
limitations: it is expensive; there are ethical network representation of a structure to
issues in achieving realism in such an event; describe the performance of the population
the trial produces only a single data point; and when traversing an egress route.
the structure has to be in place already [5]. In reality, none of these approaches include
• Model Approach C: The application of a all of the factors that influence the outcome of
(computer-based) simulation model. An an evacuation; that is, they represent only a
attempt is made to incorporate our under- subset of those factors mentioned in Fig. 59.2.
standing of human behavior in fire within a Indeed, given the relatively immature state of
computer-based model. This is then applied to the study of human behavior in fire, it would be
a representative set of scenarios in order to not be possible for the models to include all of
establish egress performance. The quality of the factors affecting egress. It is vital to under-
the results is highly dependent on a number of stand the limitations of these models in order to
factors: the sophistication and validity of the more reliably interpret and assess the results
model used, the expertise of the user, and the produced. There is a difference between the
scenarios examined and their number of factors that actually affect an evacu-
appropriateness [6]. ation and the number that can be modeled. The
• Model Approach D: The application of an gap between this prediction and reality is
engineering calculation. Here, empirical data outlined in Fig. 59.3. The hydraulic model
are distilled into a representative set of discussed here is an engineering calculation,
equations. These equations are deemed to rep- that is, model approach D.
resent a simplified version of evacuation
movement (instead of behavior), where the
results are largely determined by the physical Model Limitations
attributes of the components involved; for
example, the people, the structure, and so Many factors influence the outcome of an evacu-
on. As such, they largely overlook many of ation; models have the potential to incorporate a
the complexities apparent in the human subset of these factors. This potential influence is
response to fire. These calculations can be based on the assumption that (1) sufficient theo-
applied at the level of the structure (see retical support exists (i.e., that the factors have
Chap. 64) or the level of the structural com- been identified and formalized); (2) there are data
ponent. For instance, details of the structure that can be incorporated into the model (i.e., that
can be included in an equation that generates the factors can be quantified in some way); and
an overall egress time; for example, the num- (3) there are no limitations in the technology
ber of floors, the population size, and the used to apply the model (e.g., hardware or
egress width available. software).
Alternatively, the attributes of a particular The hydraulic model is limited in the factors it
structural component (e.g., staircase, section can represent. Several aspects of the model
of corridor, etc.) can be used. These describe should be noted:
free ebooks ==>
2120 S.M.V. Gwynne and E.R. Rosenbaum

• Behaviors that detract from movement are not engineer may want to apply several simulta-
explicitly considered. neously (e.g., the hydraulic model and a com-
• The numbers of people in a structural compo- puter simulation model) in order to have a
nent are considered rather than their identity stronger basis for the results [1]. Care should be
and their individual attributes. shown in the application of the hydraulic model
• Movement between egress components is and the presentation of the results produced.
considered (e.g., from room to room), rather With responsible use, it is able to produce rea-
than within them. sonable results in many situations. In some
• The results are deterministic and will there- situations a more sophisticated model should,
fore remain the same unless changes are made ideally, be employed; for example, where com-
to the scenario or the assumptions employed. plicated procedures are in place, where complex
The expert user can, to some degree, compen- flows are expected, and where the population is
sate for these limitations, but these limitations heterogeneous (see Chap. 60).
are inherent in the hydraulic model. Therefore, In the next sections the use of the basic
given the nature of the hydraulic model, it is able hydraulic model to estimate te (the evacuation
to represent only a small subset of behavioral time) is discussed. The empirical evidence
factors (primarily related to those that influence supporting the hydraulic model is outlined, and
movement). the calculations involved are described. Two dif-
Some time has been spent outlining the ferent versions of the basic hydraulic model are
limitations of the hydraulic model along with described: a simplified approach (first order) and
the other modeling approaches available. As the full approach (second order). Both act at the
with any model, it is critical that the engineer is level of the structural component but do so to
aware of these limitations prior to its use. When different degrees of computational rigor. The
the hydraulic (or any other) model is employed, a engineer must select one of these versions
brief description of these limitations should be based on the project, his or her expertise, and
presented along with the results produced and the the time available. Several examples are
conclusions drawn. However, despite these provided demonstrating how the hydraulic
limitations, it is possible to employ the hydraulic model can be applied. Finally, guidance is
model to establish egress performance in a con- provided on how these calculations can be
sistent and informative manner. A hydraulic employed and, by extending the model, the
model can assess te by quantifying egress perfor- types of scenarios that should be examined. An
mance, and therefore provide insight into the engineer should consider all of these issues when
effectiveness of a design. In a similar manner, determining te (and then eventually tesc).
the extended version of the model, described
later in this chapter, is able to estimate tesc.
In many cases, the hydraulic model is an Estimating te Using the Basic
acceptable method to model egress. Examples Hydraulic Model
include where only a general estimate of egress
time is required and where the hydraulic model is This section describes the fundamental
the most sophisticated model available to the components of egress movement that form the
engineer: the prescriptive codes may be too basic elements of the hydraulic model; that is, the
restrictive and not allow dedicated data to be equations used in calculating the te component in
incorporated; the resources available may not Equation 59.1.
extend to the use of a complex simulation Research-based engineering calculations for
model; and egress trials may be precluded as predicting emergency population flow have
the structure may not yet have been built. emerged over the past few decades. The major
Where these other models are available, the contributors include Predtechenskii and
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2121

Milinskii [7], Fruin [8], and Pauls [9, 10]. A RSET ¼ td þ tn þ t pe þ ðttrav þ tflow þ tne Þ
number of other contributions have been made
to the field in recent years; [11–25] however, the
methods presented here (originally developed by where te is broken down in three constituent
Nelson and MacLennan [26]) were based primar- parts: ttrav is the time spent moving toward a
ily on the major contributors highlighted above. place of safety, tflow is the time spent in
A more complete list of the research performed congestion controlled by flow characteristics,
relating to human behavior and movement is and tne is the time spent in nonevacuation
presented in Chaps. 58 and 64, as well as else- activities that do not directly contribute to the
where [4, 27–37]. population moving to a place of safety. Even
It should be noted that at the time of writing this equation is a simplification, although it
there is some discussion regarding the validity of does demonstrate that in the evacuation phase
several of the data-sets on which these models there is likely to be an amount of time spent in
are based. This has led to data-sets being with- activities other than moving directly to an exit.
drawn from the SFPE handbook and from other Given the assumptions associated with hydraulic
publications (see Chap. 64). However, their con- models, the RSET calculation using the basic
tribution to the model described here is not hydraulic model produces the following
removed given that the data-sets are broadly equation:
comparable, that they form a core component to
the original model derived by Nelson et al. [26], RSET ¼ td þ tn þ t pe þ ðttrav þ tflow Þ ð59:4Þ
that they have not yet been proven invalid (in an
The behaviors that do not directly contribute to
available, peer-reviewed publication), and that,
the evacuation are not modeled (tne).
perhaps more importantly, there is a lack of other
Calculations based on these assumptions require
equivalently comprehensive data-sets available.
compensatory actions in order to account for the
As mentioned, the sources included here are,
factors not included. These actions are discussed
in most cases, compatible and supportive of each
later in this chapter.
other. All are based on the relationship between
the speed of movement and population density of
the evacuating population stream. The equations
Fundamental Movement Calculations
derived from these sources are based on the
following assumptions:
The modeled evacuation time (i.e., the time
1. All persons start to evacuate at the same time.
predicted by the hydraulic model) utilizes a
2. Occupant flow does not involve interruptions
series of expressions that relate data acquired
caused by evacuee decisions.
from tests and observations to a hydraulic
3. The evacuees are free of impairments/
model of human flow. These primarily relate to
disabilities that impede their movement.
the following considerations: effective width,
Given the discussion presented in Chaps. 58
population density, speed, flow characteristics,
and 64 and from Fig. 59.2, these assumptions
time for passage through a component, and
exclude a number of factors and behaviors that
transitions between components. Each of these
might detract from egress performance. These
considerations is discussed in detail by Proulx
assumptions also have the effect of separating
[38]. By taking these considerations into account,
the egress components (presented in Equa-
egress movement can be quantified using the
tion 59.1) into distinct activities that are then
hydraulic model. Figure 59.4, shows a typical
treated separately during the calculations; in real-
relationship between the source data and the
ity, these components would be coupled.
derived equation. Although the expressions indi-
When representing an actual event, Equa-
cate absolute relationships, there is considerable
tion 59.1 can be rewritten as
free ebooks ==>
2122 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.4 Relation

between speed and density Spot measurement
on stairs in uncontrolled 1.0

Evacuees’ speed down stairs

total evacuations (Dashed Case average
line from Fruin [8])

(meters per second)


S = 1.08 – 0.29 d

0 1 2 3 4 5
Density (persons per square meter)

variability in the data. The engineer may wish to boundary layer widths. The effective width of
take this into account during the calculation any portion of an exit route is the clear width of
process. that portion of an exit route less the sum of the
The equations and relationships presented in boundary layers. Clear width is measured
the following paragraphs can be used indepen- 1. From wall to wall in corridors or hallways
dently or collected together to solve more com- 2. As the width of the treads in stairways
plex egress problems. Several examples 3. As the actual passage width of a door in its
outlining the use of these equations are presented open position
later in this chapter. 4. As the space between the seats along the aisles
of assembly arrangement
Effective width, We The effective width is the 5. As the space between the most intruding
usable width of the component, or We. Persons portions of the seats (when unoccupied) in a
moving through the exit routes of a building row of seats in an assembly arrangement
maintain a boundary layer clearance (i.e., main- The intrusion of handrails is considered by
tain a distance between themselves and the object comparing the effective width without the
in question) from walls and other stationary handrails and the effective width using a clear
obstacles they pass (see Fruin [8], Pauls [9, 10], width from the centerline of the handrail. The
and Habicht and Braaksma [39]). This clearance smaller of the two effective widths then applies.
is needed to accommodate lateral body sway and Using the values in Table 59.1, only handrails
assure balance. Personal preference dictates that that protrude more than 2.5 in. need be consid-
people attempt to maintain space around them- ered; that is, if the handrail protrudes less than
selves assuming that the population density is 2.5 in. into the stair width, then the overall calcu-
sufficiently low. lated width will still be less than the 6 in. reduc-
Discussion of this crowd movement phenom- tion produced by the stairwell. Minor midbody
enon is found in the works of Pauls [9, 10], Fruin height or lower intrusions such as panic hardware
[8], and Habicht and Braaksma [39]. The useful are treated in the same manner as handrails.
(effective) width of an exit path is the clear width
of the path less the width of the boundary layers. Population Density, D Population density, D, is
Figures 59.5 and 59.6 depict effective width and the measurement of the degree of crowdedness in
boundary layers. Table 59.1 is a listing of an evacuation route. The calculations in this
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2123

Fig. 59.5 Measurements a

of effective width of stairs Wall Open
in relation to walls, side
handrails, and seating Nominal
stair width
3.5 in. 3.5 in.
(8.9 cm) (8.9 cm)

6 in. 6 in.
(15.2 cm) (15.2 cm)

Area of tread use

Stair tread

Nominal aisle
stair width
3.5 in.
(8.9 cm)
Half aisle Half aisle

Effective Effective
width width

Bench handrail

Aisle stair tread

(48 in. preferred)
(122 cm)

chapter are based on population density Unless specifically stated, the entire popula-
expressed in persons per square foot (or persons tion of the first egress component (i.e., the com-
per square meter). It should be noted that ponent from which the egress movement starts)
researchers employ several different units when is included in any flow calculation. This will
describing population density. These units demonstrate the capacity limits of the route ele-
include the number of people per unit of space, ment. If the evacuating population is widely dis-
the space available per person, and the propor- persed within an egress component (i.e., it would
tion of floor space occupied [7]. In reality, the take them significantly different times to reach
population density will be dependent on the size connected egress components), the calculation is
of the individuals present. These sizes may vary based on an appropriate time step that reflects the
greatly. Here, the sizes are assumed to be time of their arrival. At each time increment, the
uniform or averaged across the population. population density of the exit route is based on
free ebooks ==>
2124 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.6 Public corridor

effective width
Boundary queue

design width

Viewing Railing
Departure lounge

Suspended TV display
Effective flight information




Effective Wall
design width

Table 59.1 Boundary layer widths The population density factors in subsequent
portions of the egress system are determined by
Boundary layer
calculation. The calculation methods involved
Exit route element (in.) (cm)
are contained in the section of this chapter titled
Stairways—wall or side of tread 6 15
Railings, handrailsa 3.5 9
Theater chairs, stadium benches 0 0
Corridor, ramp walls 8 20 Speed, S Speed is defined as the movement
Obstacles 4 10 velocity of exiting individuals, or S.
Wide concourses, passageways <18 46 Observations and experiments have shown that
Door, archways 6 15 the speed of a group or an individual in a group is
Where handrails are present, use the value if it results in a a function of the population density. The
lesser effective width relationships presented in this chapter have
been derived from the work of Fruin [8], Pauls
those that have entered the route minus those that [9, 10], and Predtechenskii and Milinskii [7]. If
have passed from it. In such situations the engi- the population density is less than approximately
neer may wish to separate the components into 0.05 persons/ft2 (0.54 persons/m2) of exit route,
subcomponents; for example, a corridor could be individuals will move at their own pace, inde-
broken into several components reflecting the pendent of the speed of others. If the population
different performances of the population. In this density exceeds about 0.35 persons/ft2 (3.8
case, the density calculations would then be persons/m2), it is assumed that no movement
based on the entire population of each of the will take place until enough of the crowd has
new components. passed from the crowded area to reduce the
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2125

population density. Between the population den- Figure 59.7 is a graphic representation of the
sity limits of 0.05 and 0.35 persons/ft2 (0.54 and relationship between speed and population den-
3.8 persons/m2), the relationship between speed sity. The speeds determined from Equation 59.5
and population density is assumed to be are along the line of movement; that is, for stairs
represented by a linear function. The equation the speeds are along the line of the treads.
of this function is Table 59.3 provides convenient multipliers for
converting vertical rise of a stairway to a distance
S ¼ k  akD ð59:5Þ along the line of movement. The travel on
where landings must be added to the values derived
S ¼ Speed along the line of travel from Table 59.3. To be conservative, it should
D ¼ Population density in persons per unit area be assumed that the population does not increase
k ¼ Constant, as shown in Table 59.2 velocity when traversing a landing between stairs
¼ k1; and a ¼ 2.86 for speed in ft/min and but continues on at the same reduced rate
density in persons/ft2 associated with stair movement.
¼ k2; and a ¼ 0.266 for speed in m/s and Although, in reality, population densities of
density in persons/m2 greater than 0.175 persons/ft2 (1.9 persons/m2)
can be achieved, it is suggested that these
densities should not be assumed in an engineer-
Table 59.2 Constants for Equation 59.5, evacuation ing design [3]. This density produces the maxi-
speed mum achievable flow rate; beyond this density,
Exit route element k1 k2 the flow rate falls rapidly. If the population den-
Corridor, aisle, ramp, doorway 275 1.40 sity increases significantly beyond 0.37 persons/
Stairs ft2 (4 persons/m2), then crush conditions might
Riser (in.) Tread (in.) develop [9, 10, 40]. This suggested maximum
7.5 10 196 1.00 compares to the occupancy levels suggested in
7.0 11 212 1.08 NFPA 101®, Life Safety Code®, of 0.142
6.5 12 229 1.16 persons/ft2 (1.54 persons/m2) to 0.003 persons/
6.5 13 242 1.23 ft2 (0.022 persons/m2) depending on the type of
1 in. ¼ 25.4 mm occupancy [41]. The suggested maximum

Fig. 59.7 Evacuation Density (persons/m2)

speed as a function of
0 0.5 1 1.5 2 2.5 3 3.5 4
density. S ¼ k  akD, 300
where D ¼ density in 1.5
persons/ft2 and k is given in
Table 59.2. Note that speed 250 1.25
Corridor, ramp,
is along line of travel
Movement speed (ft/min)

aisle, doorway
Movement speed (m/s)

200 1

150 0.75

100 0.5

Various stairs
50 0.25
per Table 3-13.4

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Density (persons/ft2)
free ebooks ==>
2126 S.M.V. Gwynne and E.R. Rosenbaum

Table 59.3 Conversion factors for relating line of travel Table 59.4 Maximum (unimpeded) exit flow speeds
distance to vertical travel for various stair configurations
Speed (along line of travel)
Stairs riser (in.) Tread (in.) Conversion factor Exit route element (ft/min) (m/s)
7.5 10.0 1.66 Corridor, aisle, ramp, 235 1.19
7.0 11.0 1.85 doorway
6.5 12.0 2.08 Stairs
6.5 13.0 2.22 Riser Tread (in.)
1 in. ¼ 25.4 mm 7.5 10 167 0.85
7.0 11 187 0.95
6.5 12 196 1.00
density is therefore beyond the heaviest occupant
6.5 13 207 1.05
load suggestion in the Life Safety Code and
should therefore be adequate for all occupancy 1 in. ¼ 25.4 mm
A conservative approach is therefore adopted
regarding the population densities that might be
Fs ¼ Specific flow
achieved during the movement of the population.
D ¼ Population density
As can be seen in Fig. 59.7, a maximum popula-
S ¼ Speed of movement
tion density of 3.76 persons/m2 is assumed. This
The flow rate unit is often referred to in
limit constrains the movement of the population.
persons/ft/minute or persons/m/second. This
The relationship between speed/flow and density
change in units will have no impact on the
is similarly affected by this constraint, with
achievable population densities below those that
Fs is in persons/min/ft when density is in
might be expected in reality and curtailed earlier
persons/ft2 and speed in ft/min; Fs is in persons/
than might be expected.
s/m when density is in persons/m2 and speed in
The maximum speed is possible, but not inev-
itable, when the density is less than 0.05 persons/
Combining Equations 59.5 and 59.6 produces
ft2 (0.54 persons/m2). These maximum speeds
are listed in Table 59.4. Fs ¼ ð1  aDÞkD ð59:7Þ
Within the range of dimensions listed in
Tables 59.2, 59.3, and 59.4, the evacuation where k is as listed in Table 59.2.
speed on stairs varies approximately as the The relationship of specific flow to population
square root of the ratio of tread width to tread density is shown in Fig. 59.8. In each case the
height. There is not sufficient data to appraise the maximum specific flow occurs when the density
likelihood that this relationship holds outside this is 0.175 persons/ft2 (1.9 persons/m2) of exit route
range. space. It is possible to establish Fs from Equa-
tion 59.7 and solve for D. There is a maximum
Specific flow, Fs Specific flow, Fs, is the flow of specific flow associated with each type of exit
evacuating persons past a point in the exit route route element; these are listed in Table 59.5.
per unit of time per unit of effective width, We, of
the route involved. Specific flow is expressed in Special Consideration for Door Mechanism In
persons/min/ft of effective width (if the value of Table 59.5 and Fig. 59.8 the maximum achiev-
k ¼ k1 from Table 59.2), or persons/s/m of effec- able specific flow rates for corridors and
tive width (if the value of k ¼ k2 from doorways are considered equivalent. This is
Table 59.2). The equation for specific flow is based on the original calculations made by Nel-
son and MacLennan [26]. However, this is based
Fs ¼ SD ð59:6Þ on the assumption that the entire effective width
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2127

Fig. 59.8 Specific flow as Density (persons/m2)

a function of population
0 0.5 1 1.5 2 2.5 3 3.5 4
density 25

(persons/min/ft effective width)

20 aisle,

(persons/s/m effective width)


Specific flow

Specific flow
15 0.80


Various stairs 0.40

per Table 3-13.5

0 0.00
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Density (persons/ft2)

Table 59.5 Maximum specific flow, Fsm These factors may act to limit the flow
Maximum specific flow through the doorway. When doors on an egress
Exit route Persons/min/ft Persons/s/m of route are not mechanically held open, these
element of effective width effective width factors should be considered. In such
Corridor, aisle, 24.0 1.3 circumstances, where the interaction with the
ramp, doorway door leaf influences performance, it may be
Stairs more conservative to assume a maximum achiev-
Riser Tread able flow rate based on the number of door leaves
(in.) (in.) available rather than the actual width (effective
7.5 10 17.1 0.94
or otherwise) of each door leaf; that is, increasing
7.0 11 18.5 1.01
the door leaf size may not produce a linear
6.5 12 20.0 1.09
increase in the achievable flow, given the need
6.5 13 21.2 1.16
to hold the leaf open and the reduction in the
available door width due to the position of the
of the doorway is available and that the passage closing leaf. A maximum flow rate of 50 persons/
of the population through the doorway is not min/door leaf is suggested for doors that are not
influenced by the door mechanism itself. If the mechanically held open [3, 8]. Fruin originally
door leaf is not mechanically held open, then the noted flow rates between 40 and 60 persons/min
traversing population may be forced to hold it through exits with door leaves; however, the
open, delaying their passage. These actions have lower flow rate of 40 persons/min was produced
the potential for slowing the evacuees’ move- during observations involving slow-moving
ment through the opening, producing a reduced occupants and so is discounted [3].
flow rate. It may also reduce the width available The data used to support the flow rate through
to the width of a single person (i.e., the width doors are several decades old and may not accu-
required for the person holding the door and rately reflect the movement and shape
passing through the exit), rather than the full characteristics of current populations; for exam-
available width of the exit leaf. In this case, the ple, the impact that the increasing levels of obe-
exit width available may be dynamic and reduced sity in some parts of the world might have on the
even from the calculated effective width. capacity of egress routes and on movement rates
free ebooks ==>
2128 S.M.V. Gwynne and E.R. Rosenbaum

[42]. In addition, the data from which these require a different approach and have an impact
relationships were derived were collected from on the overall calculation produced. The
nonemergency pedestrian movement and/or transitions can be categorized into merging or
egress drills. The functions described should not branching flows.
be assumed to necessarily provide conservative
predictions; the engineer should therefore factor Transitions Transitions are any points in the
this into the design recommendations made. exit system where the character or dimension of
a route changes or where routes merge or branch.
Calculated flow, Fc The calculated flow, Fc, is Typical examples of points of transition include
the predicted flow rate of persons passing a par- the following:
ticular point in an exit route. The equation for 1. Any point where an exit route becomes wider
calculated flow is or narrower. For example, a corridor may be
narrowed for a short distance by a structural
Fc ¼ Fs W e ð59:8Þ
change, an intruding service counter, or a
where similar element. The calculated density, D,
Fc ¼ Calculated flow and specific flow, Fs, differ before reaching,
Fs ¼ Specific flow while passing, and after passing the intrusion.
We ¼ Effective width of the component being 2. A point where the terrain changes; that is, the
traversed point where a corridor enters a stairway.
Equation 59.8 is based on the assumption that There are actually two transitions: one occurs
the achievable flow rate through a component is as the egress flow passes through the doorway,
directly proportional to its width. the other as the flow leaves the doorway and
Combining Equations 59.7 and 59.8 produces proceeds onto the stairs.
3. The point where two or more exit flows
Fc ¼ ð1  aDÞkDW e ð59:9Þ merge; for example, the meeting of the flow
from a cross aisle into a main aisle that serves
Fc is in persons/min when k ¼ k1 (from
other sources of exiting population. It is also
Table 59.2), D is in persons/ft2, and We is in
the point of entrance into a stairway serving
ft. Fc is in persons/s when k ¼ k2 (from
other floors.
Table 59.2), D is persons/m2, and We is m.
4. Where a flow branches into several other
flows. A decision has to be made regarding
Time for passage, tp The time for passage, tp, is
the proportion of the incoming flow that uses
the time for a group of persons to pass a point in
each of the outgoing flows, that is, into several
an exit route and is expressed as
other egress components. The proportion of
t p ¼ P=Fc ð59:10Þ the flows will be influenced by a number of
different behavioral and procedural issues
where tp is time for passage (tp is in minutes where (refer to Chaps. 58 and 64, and also to
Fc is in persons/min; tp is in seconds where Fc is Predtechenskii and Milinskii [7]). The propor-
persons/s). P is the population size in persons. tion of flow using each of the egress
Combining Equations 59.9 and 59.10 yields components may be apportioned evenly,
according to the capacity of the components,
t p ¼ P=½ð1  aDÞkDW e  ð59:11Þ
or according to behavioral/procedural issues,
There are several transition configurations that such as familiarity. Once this apportionment
may arise during an engineering calculation that has been established, then each of the flow
involve the interaction between flows of people. calculations proceed as before and can be
These transitions need to be identified, as they conducted independently of each other.
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2129

The following rules apply when determining Fs(in) ¼ Specific flow arriving at transi-
the densities and flow rates following the passage tion point
of a transition point: We(in) ¼ Effective width prior to transi-
1. The flow after a transition point is a function, tion point
within limits, of the flow(s) entering the tran- We(out) ¼ Effective width after passing
sition point. transition point
2. During the transition between two (b) For cases involving two incoming flows
components, it will be necessary to establish and one outflow from a transition point,
the density in the new component; that is, it such as that which occurs with the
is assumed that sufficient information is merger of a flow down a stair and the
available on the previous component to entering flow at a floor,
enable this calculation to be made. The den-
Fsðin1Þ W eðin1Þ þ Fsðin2Þ W eðin2Þ
sity in the new component will be calculated FsðoutÞ ¼
by solving for D in Equation 59.9; this will W eðoutÞ
produce a quadratic in D. In order to do this, ð59:13Þ
the flow rate into the component will need to
be known. Unless the maximum value is where the subscripts (in-1) and (in-2)
achieved, there will normally be two indicate the values for the two incoming
solutions of D produced: one above and one flows.
below Dmax (where Dmax is the density value (c) For cases involving other geometry
that produces the maximum flow). Nelson formations merging together, the follow-
and MacLennan [26], and Predtechenskii ing general relationship applies:
and Milinskii [7], and Milke [40] state that    
Fsðin1Þ W eðin1Þ þ    þ FsðinnÞ W eðinnÞ
the smaller of the D values should be    
employed; that is, less than or equal to ¼ Fsðout1Þ W eðout1Þ þ    þ FsðoutnÞ W eðoutnÞ
Dmax. If the larger D value (greater than ð59:14Þ
Dmax) is used, it implies that the flow rate
between the two components both rises and where the letter n in the subscripts (in-n)
falls during a single transition. This is not and (out-n) is a number equal to the total
considered to be reasonable. number of routes entering (in-n) or leav-
3. The calculated flow, Fc, following a transition ing (out-n) the transition point.
point cannot exceed the maximum specific 5. Where the calculated specific flow, Fs, for the
flow, Fsm, for the route element involved route(s) leaving a transition point, as derived
multiplied by the effective width, We, of that from the equations in rule 4, exceeds the max-
element. imum specific flow, Fsm, a queue will form at
4. Within the limits of rule 2, the specific flow, the incoming side of the transition point. The
Fs, of the route departing from a transition number of persons in the queue will grow at a
point is determined by the following rate equal to the calculated flow, Fc, in the
equations: arriving route minus the calculated flow leav-
(a) For cases involving one flow into and ing the route through the transition point.
one flow out of a transition point, 6. Where the calculated outgoing specific flow,
Fs(out), is less than the maximum specific flow,
FsðinÞ W eðinÞ Fsm, for that route(s), there is no way to pre-
FsðoutÞ ¼ ð59:12Þ
W eðoutÞ determine how the incoming routes will
merge. The routes may share access through
the transition point equally, or there may be
Fs(out) ¼ Specific flow departing from
total dominance of one route over the other.
transition point
For conservative calculations, assume that the
free ebooks ==>
2130 S.M.V. Gwynne and E.R. Rosenbaum

route of interest is dominated by the other 50 people starting at the top of the stairs with
route(s). an initial density of 1.5 p/m2 (p ¼ persons) to exit
A simple example is presented in order to from the door at the end of the corridor?
clarify the required calculations [40]. This simple
example is followed by a more comprehensive Solution The velocity can be calculated
example illustrating the two different hydraulic according to Equation 59.5:
models: first- and second-order hydraulic
models. S ¼ k  akD

For 7/11 stair, k ¼ 1.08:

Example 1 A 1.8 m (approximately 6 ft) wide
(descending) 7/11 stair has 10 risers and leads to 1:08  ð0:266Þð1:08Þð1:5Þ
a 10 m long (approximately 33 ft), 1.8 m wide ∴S ¼ 0:65 m=sð128 ft=minÞ
corridor (approximately 6 ft). At the end of the
The time to traverse the stair can then be calcu-
corridor, there is a 1.3 m (approximately 4 ft
lated. The distance to be covered is
3 in.) wide door (Fig. 59.9). This door is mechan-
ically held open. How long does it take for

10 risers ¼ 70 in: ¼ 1:78 m; 10 treads ¼ 110 in: ¼ 2:79 m

diagonal length ¼ 3:31 mðapproximately 10 ft 10 in:Þ

Therefore, the time to cover the stairs is 3.31/ The time delay for the last person to start on
0.65 ¼ 5.1 s. the stair can be calculated as follows. With a flow
The width of the stair can be calculated using rate of 1.46 p/s on the stair (and thus of the queue
Table 59.1 (1.8–0.3 m). Given that the density entering the stair), the time for the queue to
(1.5 p/m2) and the velocity (0.65 m/s) are known, dissipate is calculated using Equation 59.10:
the flow rate on the stair using Equation 59.9 can
t p ¼ P=Fc
now be determined:
Fc ¼ Fs W e ¼ ðk  akDÞDW e The time for the population at the top of the stairs
to enter the stairs is then
Fc ¼ 1:46 p=s
50=1:46 ¼ 34:2 s
This value produces a specific flow rate less than
the maximum value and so it can therefore be The time for the entire population to enter the
used during the calculation. staircase and the time to traverse the staircase are
now known. Given this, the time taken for the
Fig. 59.9 Geometry used last person to enter and traverse the staircase can
in Example [1] be determined. The time to traverse the corridor
now needs to be calculated.
The flow rate into the corridor is 1.46 p/s; that
is, that produced on the staircase. Equation 59.9
states that

Fc ¼ Fs W e ¼ ðk  akDÞDW e

where Fc can be set to 1.46; this equation can

now be solved for D.
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2131

Fig. 59.10 Movement of A B C D E F

people through the

10 m

3.3 m

0 10 20 30 40 50 60
Time (s)

For a corridor, k is set to 1.40. Given the for the population to flow through the doorway is
corridor is 1.8 m in width, the effective width is then calculated using Equation 59.10:
We ¼ 1.4 m (4 ft 7 in).
t p ¼ N=Fs W e ¼ 50=ð1:3=1:0Þ ¼ 38:5 s
Therefore, solving for D

0:266D2  D þ ð1:46=1:4  1:4Þ ¼ 0 The final solution is not simply formed from
0:266D2  D þ 0:744 ¼ 0 adding these values together, as some of them
occur simultaneously. The final result is better
Solving for D, it is therefore either 1.03 or 2.72 explained by referring to Fig. 59.10.
p/m2. The dashed line indicates the movement of the
  first person. This person is not influenced by
∴ D ¼ 1:03 or 2:72 p=m2 0:1 or 0:25 p=ft2 queuing at any point and is therefore constrained
only by the velocity values derived from the
From the previous discussion regarding selecting
densities calculated on the different components.
D values, a value of 1.03 p/m2 (0.1 p/ft2) should
This person therefore spends 5.1 s traversing the
be chosen.
stairs and 9.8 s traversing the corridor, reaching
Given that the density is known and the rela-
the final exit after 14.9 s (marked A and B in
tionship between density and velocity is
Fig. 59.10). The entire population entered the
expressed in Equation 59.5:
staircase after 34.2 s (marked C in Fig. 59.10)
S ¼ k  akD and has reached the end of the stairs after
34.2 + 5.1 ¼ 39.3 s (marked D in Fig. 59.10).
the velocity, S, can be calculated as being 1.02 m/ The last person from this group will have reached
s (201 ft/min). The time to traverse the corridor is the exit at 49.1 s, assuming that person did not
therefore 10/1.02 ¼ 9.8 s. The flow through the encounter any congestion approaching the door
doorway at the end of the series of components (marked E in Fig. 59.10). Given that the first
can now be determined. Given the narrowing at person has reached the exit after 14.9 s and that
the door, the formation of the queue should be the congestion at the final exit lasted for 38.5 s,
examined. The specific flow rate at the door is this congestion is not clear until 53.4 s (marked F
calculated using Equation 59.8: in Fig. 59.10, indicating the end of the solid
Fs ¼ Fc =W e ¼ 1:46=1:0 curve). Therefore, the last person to arrive
interacts with the congestion at some point prior
Fs ¼ 1.46 p/s/m (27 p/min/ft) > Fsm, where the to reaching the door; that is, the congestion still
value of Fsm is 1.3 p/s/m (24 p/min/ft). The time exists when that person arrives. The evacuation
free ebooks ==>
2132 S.M.V. Gwynne and E.R. Rosenbaum

time is then determined by the time taken for the components (i.e., between areas where the phys-
congestion to clear at the final exit (e.g., 53.4 s). ical constraints affecting egress performance
change) is calculated. This is more labor inten-
sive than the first-order approach, requiring a
First- and Second-Order Hydraulic larger number of calculations to be made; how-
Models ever, it does require fewer assumptions and
provides information on the movement between
The various calculations discussed in the previ- each of the structural components in the egress
ous section can be combined in order to assess route rather than a subset of them. A second-
the movement component of the evacuation pro- order analysis is by no means a trivial task and
cess; that is, to calculate te. By applying these requires judgment based on the structure exam-
calculations, the necessary movement ined and the incident scenario. This process is
components (e.g., flow rates, velocities, popula- outlined in Fig. 59.11b.
tion densities, and travel speeds) can be
established enabling the overall movement time
to be found. Example Applications

The first- and second-order hydraulic models can

First-Order Hydraulic Model be better understood through the description of
two example applications. (More examples can
There are several ways in which these movement be found elsewhere [43].) A relatively simple
calculations can be used; two are described here. example is presented, although even in this case
The first-order hydraulic model represents a the difference between the effort required in
simplified approach: instead of calculating the applying the two versions of the model is
flow of people between individual components, apparent.
this method focuses on the component that places
the most severe constraint on the flow of people Example 2 Consider an office building
around the structure and then uses this constraint (Fig. 59.12) with the following features:
to determine the movement time. The engineer is 1. There are nine floors, 300 ft by 80 ft (91 m by
required to establish the time to reach the 24 m).
controlling component; the time for the popula- 2. Floor-to-floor height is 12 ft (3.7 m).
tion to traverse this component; the time for the 3. Two stairways are located at the ends of the
last person to leave the controlling component; building (there are no dead ends).
and the time for the last person to reach safety 4. Each stair is 44 in. (1.12 m) wide (tread
from the controlling component [43]. This pro- width) with handrails protruding 2.5 in.
cess is outlined in Fig. 59.11a. This approach (0.063 m).
makes greater use of the maximum flow rates 5. Stair risers are 7 in. (0.178 m) wide and
and densities allowed, given the reduced level treads are 11 in. (0.279 m) high.
of calculation required. The controlling compo- 6. There are two 4 ft by 8 ft (1.2 m by 2.4 m)
nent will depend on the nature of the structure; landings per floor of stairway travel.
for example, the controlling component could be 7. There is one 36-in. (0.91-m) clear width door
a stair, an exit from the stair, an exit from a room, at each stairway entrance and exit. These are
and so on. assumed not to be mechanically held open.
8. The first floor does not exit through
Second-Order Hydraulic Model 9. Each floor has a single 8-ft (2.4-m) wide
corridor extending the full length of each
The second-order hydraulic model requires that floor. Corridors terminate at stairway
the flow of people between each of the structural entrance doors.
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2133

Fig. 59.11 (a) First-order a b

hydraulic model; (b)
second-order hydraulic Define network of Define network of
model components that form components that form
structure being structure being
assessed. assessed

Establish egress Establish egress

routes used during routes used during
scenario being the scenario being
examined. examined.

Identify constraining Establish flow characteristics

components; that is, for each component; start
those that limit from component farthest
movement. from safety.

Establish Ensure that flow

flow characteristics. characteristics do not
Identify single exceed maximum
controlling factor C. values.

Calculate time to Assess impact of

reach C and time transitions. Recalculate
to traverse C. flow characteristics, if

Establish time for

last person to flow
overall egress
through C and
reach safety.

Office space
150 occupants

80 ft

Office space
150 occupants

300 ft.

Fig. 59.12 Floor plan for example

free ebooks ==>
2134 S.M.V. Gwynne and E.R. Rosenbaum

10. There is a population of 300 persons/floor. moving with the flow is 38.2/105 ¼ 0.36 min/
Solution A First-Order Approximation. 5. Estimate building evacuation time.
1. Assumptions. If all of the occupants in the building start
The prime controlling factor will be either evacuation at the same time, each stairway
the stairways or the door discharging from can discharge 48 persons/min. The population
them. Queuing will occur; therefore, the spe- of 2400 persons above the first floor will
cific flow, Fs, will be set to the maximum require approximately 25 min to pass through
specific flow, Fsm. All occupants start both exits. An additional 0.36-min travel time
evacuating at the same time. The population is required for the movement from the second
will use all facilities in the optimum balance. floor to the exit. (A more conservative esti-
2. Estimate flow capability of a stairway. mate of the travel time might also include the
From Table 59.1, the effective width, We, time for the first person to move from within
of each stairway is 44  12 ¼ 32 in. (2.66 ft) the second floor to the stair.) The total mini-
(813 mm [0.81 m]). The maximum specific mum evacuation time for the 2400 persons
flow, Fsm, for the stairway (from Table 59.5) located on floors 2 through 9 is estimated at
is 18.5 persons/min/ft (1.01 persons/s/m) of 25.4 min.
effective width. Specific flow, Fs, equals max-
imum specific flow, Fsm. Therefore, using Solution B Second-Order Approximation.
Equation 59.6, the flow from each stairway 1. Assumptions.
is limited to 18.5  2.66 ¼ 49.2 persons/min. The population will use all exit facilities
3. Estimate flow capacity through a door. optimally; all occupants start egress at the
Again from Table 59.5, the maximum spe- same time. All persons are assumed to start
cific flow through a 36-in. (0.9 m) door is to evacuate at time zero.
24 persons/min/ft (1.31 persons/s/m) of effec- 2. Estimate flow density (D), speed (S), specific
tive width. Also, the effective width, We, of flow (Fs), effective width (We), and initial
each door is 36  12 ¼ 24 in. (2 ft) (609 mm calculated flow (Fc) typical for each floor.
[0.61 m]). Therefore, using Equation 59.8, the 3. Divide each floor in half to produce two exit
flow through the door is limited to calculation zones, each 150 ft (45.7 m) long.
24  2 ¼ 48 persons/min. This is less than To determine the density, D, and speed, S, if
the maximum flow rate through an exit that all occupants try to move through the corridor
is not mechanically held open (50 p/min). at the same time, that is, 150 persons moving
Because the flow capacity of the doors is less through 150 ft of an 8-ft (2.4-m) wide
than the flow capacity of the stairway served, corridor:
the flow is controlled by the stairway exit
doors (48 persons/stairway exit door/min). D ¼ 150 persons=1200 ft2 corridor area
4. Estimate the speed of movement for estimated ¼ 0:125 persons=ft2
stairway flow. From Equation 59.5, S ¼ k  akD.
From Equation 59.5 the speed of movement From Table 59.2, k ¼ 275.
down the stairs is 212  ð2:86  212  0:175Þ
¼ 105 ft=minð0:53 m=sÞ. The travel distance S ¼ 275  ð2:86  275  0:125Þ
between floors (using the conversion factor ¼ 177 ft=min ð54 m=minÞ
from Table 59.3) is 12  1.85 ¼ 22.2 ft
From Equation 59.7, Fs ¼ ð1  aDÞkD.
(6.8 m) on the stair slope plus 8 ft (2.4 m) travel
Fs ¼ ½1  ð2:86  0:125Þ  275  0:125 ¼ 22
on each of the two landings, for a total floor-to-
floor travel distance of 22:2 þ ð2  8Þ ¼ persons=min=ftð1:2 persons=s=mÞ effective
38:2 ftð11:6 mÞ. The travel time for a person width
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2135

From Table 59.5, Fs is less than the maxi- 154  48 ¼ 106 persons=min
mum specific flow, Fsm; therefore, Fs is used
for the calculation of calculated flow. 5. Estimate impact of stairway on exit flow.
From Table 59.1, the effective width of the From Table 59.1, effective width, We, of the
corridor is stairway is

8  ð2  0:5Þ ¼ 7 ftð2:13 mÞ 44  12 ¼ 32 in:ð2:66 ftÞ ð0:81 mÞ

From Equation 59.9, calculated flow, Fc ¼ From Table 59.5, the maximum specific flow,
(1  aD)kDWe. Fsm, is 18.5 persons/min/ft (1.01 persons/s/m)
effective width. From Equation 59.12, the
Fc ¼ ½1  ð2:86  0:125Þ  275  0:125  7 specific flow for the stairway,
¼ 154 persons=min FsðstairwayÞ ¼ 24  2=2:66 ¼ 18:0 persons=min=ft
At this stage in the calculation, calculated ð0:98 persons=s=mÞeffective width
flow, Fc, is termed initial calculated flow for
In this case, Fs is less than Fsm and Fs is used.
the exit route element (i.e., corridors) being
The value of 18.0 p/min/ft for Fs applies
evaluated. This term is used because the cal-
until the flow down the stairway merges with
culated flow rate can be sustained only if the
the flow entering from another floor.
discharge (transition point) from the route can
Using Fig. 59.8 or Equation 59.7 and
also accommodate the indicated flow rate.
Table 59.2, the density of the initial stairway
4. Estimate impact of stairway entry doors on
flow is approximately 0.146 persons/ft2 (1.6
exit flow.
person/m2) of stairway exit route. From Equa-
Each door has a 36-in. (0.91-m) clear width.
tion 59.5 the speed of movement during the
From Table 59.1, effective width is
initial stairway travel is
W e ¼ 36  12 ¼ 24 in: ð2 ftÞð0:61 mÞ
212  ð2:86  212  0:146Þ
From Table 59.5, the maximum specific flow, ¼ 123 ft=minð0:628 m=sÞ
Fsm, is 24 persons/min/ft effective width.
From Equation 59.12, This value differs from that produced in the
  first-order, which is based on the maximum
FsðdoorÞ ¼ FsðcorridorÞ W sðcorridorÞ =W eðdoorÞ achievable density rather than a calculated
¼ ð22  7Þ=2 ¼ 77 persons =min =ft
From the first-order solution, the floor-to-
ð4:2 persons=s=mÞ effective width floor travel distance is 38.2 ft (11.6 m). The
time required for the flow to travel one floor
Since Fsm is less than the calculated Fs, the
level is
value of Fsm is used. Therefore, the effective
value for specific flow is 24 p/min/ft. 38:2=123 ¼ 0:31 minð19 sÞ
From Equation 59.8, the initial calculated
flow, Using Equation 59.8, the calculated flow is

Fc ¼ Fs W e ¼ 24  2 ¼ 48 persons=min Fc ¼ 18:0  2:66 ¼ 48 persons=min

through a 36-in. (0.91-m) door. Since Fc for the After 0.31 min, 15 (i.e., 48  0.31) persons
corridor is 154 p/min while Fc for the single exit will be in the stairway from each floor feeding
door is 48 p/min, queuing is expected. The to it. If floors 2 through 9 exit all at once, there
calculated rate of queue buildup will be will be
free ebooks ==>
2136 S.M.V. Gwynne and E.R. Rosenbaum

15  8 ¼ 120 persons in the stairway 6. Estimate impact of merging of stairway flow

and stairway entry flow on exit flow.
After this time the merging of flows between From Equation 59.13,
the flow in the stairway and the incoming flows
at stairway entrances will control the rate of

Fsðout‐stairwayÞ ¼ FsðdoorÞ  W eðdoorÞ þ Fsðin‐stairwayÞ  W eðin‐stairwayÞ =W eðin‐stairwayÞ
¼ ½ð24  2Þ þ ð18  2:66Þ=2:66
¼ 36 persons=min=ftð1:97 prsons=s=mÞ
effective width

From Table 59.5, Fsm for the stairway is 18.5 Therefore, referring to Equations 59.8, 59.10,
persons/min/ft (1.01 persons/m/s) effective and 59.11, at
width. Since Fsm is less than the calculated
ð135=48Þ  60 þ 49 ¼ 218 sð3:6 minÞ
Fs, the value of Fsm is used.
7. Track egress flow. all persons have evacuated the ninth floor.
Assume all persons start to evacuate at At
time zero. Initial flow speed is 177 ft/min
(0.9 m/s). Assume that congested flow will ½ð135=48Þ  60 þ 49 þ 19 ¼ 237 s ð4:0 minÞ
reach the stairway in approximately 0.5 min.
the end of the flow reaches the eighth floor.
This conservative assumption is based on the
population having to travel a distance of
between 50 ft and 150 ft (15.2 m and 237 þ f½ð135=ð2:66  18:5Þ  60g
45.7 m) to the exit traveling at 177 ft/min ¼ 401 sð6:7 minÞ
(0.9 m/s); that is, the derived travel speed in
the corridor. At 0.5 min, flow starts through all persons have evacuated the eighth floor.
stairway doors. Fc through doors is At
48 persons/min for the next 19 s (0.31 min).
ð401 þ 19Þ ¼ 420 s ð7:0 minÞ
At 49 s, 120 persons are in each stairway and
135 are waiting in a queue at each stairway the end of the flow reaches the seventh floor.
entrance. At
How the evacuation progresses from this
point on depends on which of the floors take 420 þ f½135=ð2:66  18:5Þ  60g
precedence in entering the stairways. Any ¼ 584 sð9:7 minÞ
sequence of entry may occur [1]. To set a
boundary, this example estimates the result all persons have evacuated the seventh floor.
of a situation where dominance proceeds At 603 s The end of the flow reaches the 6th
from the highest to the lowest floor. (10.1 min) floor
The remaining 135 persons waiting at each At 767 s All persons have evacuated the 6th
stairway entrance on the ninth floor enter (12.8 min) floor
through the door at the rate of 48 persons/ At 786 s The end of the flow reaches the 5th
(13.1 min) floor
min. The rate of flow through the stairway is
At 950 s All persons have evacuated the 5th
regulated by the 48 persons/min rate of flow
(15.8 min) floor
of the discharge exit doors. The descent rate At 969 s The end of the flow reaches the 4th
of the flow is 19 s/floor. (16.2 min) floor
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2137

At 1133 s All persons have evacuated the 4th scenarios examined and more reliably provide
(18.9 min) floor an estimate of the longest RSET value to be
At 1152 s The end of the flow reaches the 3rd expected, as compared with the assessment of
(19.2 min) floor a single scenario.
At 1316 s All persons have evacuated the 3rd
(21.9 min) floor
At 1335 s The end of the flow reaches the 2nd
(22.3 min) floor
At 1499 s All persons have evacuated the 2nd
Employing Extended Hydraulic Model
(25.0 min) floor to Calculate tesc
At 1518 s All persons have evacuated the
(25.3 min) building Factors Influencing an Evacuation

From this example it is clear that in some The extended hydraulic model provides a foun-
situations little difference exists in the results dation to evaluate evacuation performance, that
produced by the use of the two basic hydraulic is, escape time (tesc) as opposed to evacuation
models (first and second order). This may be movement time (te). Figure 59.3 indicates that
expected in simple geometries and simple there are more factors that actually influence an
movement scenarios. However, as the evacuation than can currently be modeled. The
scenarios and geometries become more com- factors that can be included are dependent on our
plex, the results produced by the two versions understanding of real-life phenomena, on the
of the model may differ significantly, espe- data available, and on the limitations of the
cially if there are difficulties in establishing model adopted. As has already been stated, the
the controlling element in the first-order hydraulic model outlined here can be employed
approximation. in the examination of different egress scenarios.
The second-order hydraulic model This is critical in generating a robust and repre-
produces a larger set of information (e.g., the sentative solution.
time to clear components, the time to clear A number of behaviors can influence the per-
floors, the movement conditions between all formance of the population. It is possible to
structural components, etc.). However, it is implicitly represent some of these behaviors
sensitive to which of the components have (i.e., the consequences of these behaviors) by
precedence (e.g., in merging flows) and in manipulating parameters associated with the
the proportion of the population using partic- hydraulic model and then examining a range of
ular routes. This may require several scenarios.
calculations to establish the most conservative In order to increase the information obtained
result. The first-order hydraulic model in any egress analysis and the confidence in this
produces only the overall evacuation time information, a representative set of egress
and the results relating to the constraining scenarios could be examined. Producing one
component. It should also be noted that in “definitive” result is insufficient, given the
complex geometries identifying the many scenarios that can actually develop and
constraining component is not a trivial task, also given the limitations of the modeling
and it can be extremely time consuming. approach. Presenting a single answer may pro-
Given that an assessment of a structure is duce overconfidence in the accuracy and validity
necessary, it is important to identify the of the result.
scenarios involved in this assessment. It may
not be possible to definitively establish the
worst-case scenario prior to the calculation. Basic Variables
It is therefore essential to examine several
representative scenarios that form a set of Given the scope of the extended hydraulic
predictions. This will offer insight into the approach, the scenarios that can be examined
free ebooks ==>
2138 S.M.V. Gwynne and E.R. Rosenbaum

are limited and require the manipulation of a few of individual components, rather than the entire
basic variables: egress route.
• The routes available during the evacuation
• The evacuee’s use of the routes available dur- Evacuee’s Use of the Routes Available During
ing the evacuation the Evacuation In reality, egress routes are
• The movement attributes of the evacuating rarely used according to their design capacity
population and the presence of impairments (i.e., efficiently). Instead, routes are used
and disabilities according to occupant familiarity, visual access,
• The time taken to respond to the call to evac- the procedure in place, and the evolution of the
uate—the pre-evacuation time (tpe) incident itself. [4, 45] This reality can be
When using the extended hydraulic model, represented within the hydraulic model by
these variables can be manipulated to produce a modifying the proportion of the population
number of different scenarios and therefore an using a particular route. Although it might be
envelope of results. By manipulating these difficult to gather accurate data on the use of
variables, at least a small subset of the behaviors the routes available, engineering judgment
that might detract from egress performance can could still be used to assess the potential impact
be accounted for, albeit in an implicit way. The of the unbalanced use of the egress routes avail-
manner in which this is performed will be depen- able. For instance, in the example shown in
dent on the nature of the occupancy being exam- Fig. 59.12 it was assumed that the population
ined and the scenarios that are considered as used the egress routes in an optimal manner;
realistic. The scenarios examined (and omitted) that is, that they split evenly between the
by the engineer should be justified through a staircases. It may be the case that one route is
detailed explanation accompanying the reported more familiar to the population than another
results. In addition, the scope of the results can be (through normal use, proximity to elevators, con-
extended to represent tesc rather than just te. nectivity to nearby car park, etc.). This inefficient
use can then be reflected in the calculations by
Routes Available During the Evacuation In more people using one of the routes available. If
some scenarios it is possible that routes will be it is assumed that 75 % of the population make
lost due to the nature of the incident or given the use of one of the staircases, then applying the
use of the space available; for example, a route first-order model produces an overall egress time
might not be protected. This loss can be of approximately 37.9 min (1800/48 + 0.36). As
considered. a consequence, the other staircase would clear
Some existing codes require such forms of more quickly.
analysis. For instance, in the British Standard
BS5588 [44], the largest exit is discounted, Movement Attributes of the Evacuating
assumed blocked by the incident. In NFPA 101 Population The makeup of the population can
one scenario requires the evaluation of the vary over time. Variations in the population’s
impact of a fire located in the primary means of capabilities can be represented through the mod-
egress [41]. ification of the maximum velocities that can be
The hydraulic model can be manipulated to attained. The range of the population’s
represent the loss of available egress routes. In capabilities can be extended to include the pres-
the example shown in Fig. 59.12, an entire stair- ence of the mobility impaired. Although the
case could be presumed lost due to the nature of reduction of velocities represents only one aspect
the incident. This would have a profound impact of impairment (e.g., it does not represent behav-
on the results produced. Instead of an overall ioral issues, pre-evacuation issues, the impact
evacuation time of 25.4 min, it instead required that the presence of the mobility impaired might
50.4 min, when applying the first-order approxi- have on population densities, etc.), it is still an
mation. The calculation may also involve the loss important consideration. This may be of
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2139

particular importance where the egress perfor- trying to account for these factors. Where it is not
mance is not dominated by congestion and flow relevant or possible, then it should be clearly
constraints, but by travel distances. Looking at stated allowing the exclusion of the factors
the example presented above, in the first-order from the assessment to be judged.
model, the presence of the impaired may increase
the 0.36 min to travel between the floors on
stairs. In the second-order model, it may have a Developing Escape Scenarios
more complex effect. It could also be taken into
consideration when determining the tpe compo- As mentioned previously, when assessing the
nent, with a population with impairments egress performance of a structure, it is vital to
extending the preparation required. produce results that cover a range of scenarios.
This process adds to the robustness of the results
Time Taken to Respond Part of the RSET cal- and the credibility; that is, it is not credible that
culation (see Equation 59.1) requires the assess- the use of a hydraulic model (or any other model)
ment of a pre-evacuation phase, tpe; this phase would produce a definitive, single result. This is
is the time between notification and the time for discussed in more detail in Chap. 57 and has been
the population to evacuate. It can be varied given discussed and developed elsewhere (e.g. in PD
the scenario being represented, the notification 7975-6:2004) [46]. Ideally, a number of
system in place, the procedures employed, and so scenarios should be examined. A viable set of
on. This time may then allow limited scenarios can be produced by varying factors
comparisons to be made between different pro- within the hydraulic model (including those
cedural measures, notification systems, and the identified in the previous section) in a logical
like. Pre-evacuation time may be particularly manner. Given that several factors are
important where scenarios are not dominated by represented (e.g., speed, flow, route availability/
flow and the egress route capacities [1]. However, usage, pre-evacuation times), each of these
it should be recognized that the potential benefits factors can be modified to implicitly represent
of distributing the response of the population different scenarios. Using this approach, a range
cannot be represented in the hydraulic approach of viable scenarios can be examined that produce
given its fundamental assumptions. different RSET values. Once complete, the lon-
In reality, the relationship between the gest RSET value generated would then be used
pre-evacuation phase and the evacuation phase for comparison against the ASET value
is complex. It is not simply a case of adding the produced.
times of the two phases together. Given the Purser identified two base scenarios that can
scenario, the extent of the pre-evacuation time then be modified through the manipulation of
distribution may increase or reduce the level of model variables in order to produce sets of
congestion produced. This complex relation- scenarios for analysis (Fig. 59.13) [1]. This is
ship is difficult to represent unless the evacuees just one suggested approach at gaining a broader
are simulated on an individual basis (see insight into the evacuation performance of a
Chap. 60). structure and establishing RSET using an engi-
Data are required in order to include these neering calculation; however, it is indicative of
factors in the calculation. These data, particularly the different scenarios that can be examined.
regarding the pre-evacuation phase, are scarce Purser identified that in sparsely populated
and not always reliable (see Chap. 64). Although spaces, the overall egress time produced (tesc) is
this limitation should be acknowledged, it does more sensitive to the time taken to traverse the
not preclude further engineering analysis. Even distance to a place of safety and the time taken to
where engineering judgment is required, it is still respond, than to the time for congestion to evap-
critical to assess the robustness of the results by orate. In such situations, it is unlikely that egress
free ebooks ==>
2140 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.13 Generation of

scenarios by manipulating Scenario 1: Scenario 2:
parameters High-density conditions Low-density conditions

Parameter A: Parameter A:
Egress routes Egress routes
available = A1 available = A2

Parameter B: Parameter B:
Use of egress Use of egress
routes = B1 routes = B2

Parameter C: Parameter C:
population population
attributes = C1 attributes = C2
(tp-e , tflow, ttrav) (tp-e , ttrav)

components will be overloaded and generate evacuation time, by examining scenarios where
queues that dominate the egress performance. the results are determined by different factors;
In densely populated spaces, Purser identified for example, flow and/or travel and response. For
that it is more likely that congestion will be a single structure the impact of these different
produced. In such situations, the response of scenarios can be assessed to establish which of
individuals may be influenced by observing the them produces the most prolonged egress time.
activities of other evacuees [45, 47]. This is When using the Purser approach to establish
likely to reduce the distribution of the evacuation time from a particular structure, it
pre-evacuation times produced. This implies is assumed that the engineer cannot be sure
that the population will arrive at structural which of these scenarios will produce the longer
components within a smaller range of times. escape time, prior to the calculation being
Therefore, the time to reach a point of safety is conducted. Both scenarios are therefore exam-
likely to be highly sensitive to the clearance of ined. Each of the scenarios includes an assess-
congestion along the egress routes. ment of the population movement (i.e., te) and
When applying a sophisticated simulation the pre-evacuation phase (i.e., tpe). The com-
model, the factors that determine the outcome bined result of these two phases is termed tesc
of a scenario (e.g., whether it is determined by (see Equation 59.2). Where other new terms are
flow, travel, etc.) will be a result of the analysis used below, they are described.
(see Chap. 60). Given the limitations of the Purser identified two situations (labeled
hydraulic model, the engineer has to impose Scenarios 1 and 2 in Fig. 59.13). In Scenario
these conditions prior to the calculation being 1 it is assumed that congestion dominates the
conducted (e.g., whether congestion or travel results produced (tflow). In such situations the
distance determines the time to reach safety) time required for the evacuation of an enclosure
and then assess their impact; the critical behav- depends on the pre-evacuation time and unre-
ioral/movement components are determined stricted walking time of the first few occupants
prior to the calculations being made. The Purser to start to leave; these determine the time for
approach can be applied in order to establish congestion to develop. Once queues have formed
what underlying factors determine the overall at the “constraining” component, the time to
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2141

clear the building becomes a function of the [1, 47]. Given the limited data currently available
number of occupants and the evacuation flow it appears that the 99th percentile can be
rate capacity of these components. The evacua- reasonably approximated by a value of four to
tion time for an enclosure is estimated to be five times the mean pre-evacuation time [25, 30,
tesc ¼ t pe1 þ ttrav þ tflow ð59:15Þ
Again the example shown in Fig. 59.12 can be
where tpe1 is the pre-evacuation time associated used to illustrate the two different scenarios. Let
with the first people to respond. Purser uses the us assume that the structure represents an office
first percentile of a representative pre-evacuation space as indicated. In the first analysis, it is
distribution to estimate this value [1]. If this is assumed that the evacuation is dependent on
not available, then the average pre-evacuation flow characteristics (i.e., Purser’s Scenario 1)
time associated with the occupancy (which is and will therefore make use of the flow
more likely to be available; see Chap. 64) should calculations already made. In addition,
be used as a conservative estimate. tflow is the pre-evacuation times will be extracted from the
time of total occupant population to flow through work conducted by Fahy and Proulx to support
the most restrictive components. ttrav is the time these calculations [25]. Given that it is an office
taken to traverse the average distance to a place space, pre-evacuation times ranging from 1 to
of safety. (The maximum distance should be 6 min will be assumed. These times are
employed in this calculation for a more conser- employed as conservative estimates of the 1st
vative approach.) and 99th percentiles.
It is apparent that either the first- or second- Given the results already produced, the evac-
order hydraulic model can be used to generate uation time for Scenario 1 can be estimated as
tflow and ttrav; indeed the combination of these being
two parameters approximates the te term previ- tesc ¼ 1 þ 25:3 ¼ 26:3 min
ously described (see Equation 59.4).
In Scenario 2 it is assumed that congestion This time is based on the assumption that ttrav and
does not dominate the results; the results are tflow is approximated by the results produced in
primarily influenced by the time taken to reach the second-order model.
safety (ttrav) and the time to respond (tpe). This If instead it is assumed that, for some reason,
scenario is not necessarily based on the assump- this scenario was not determined by flow (i.e.,
tion that congestion does not develop; only that Purser’s Scenario 2), then the following calcula-
the impact of the congestion is dominated by the tion can be made:
extensive pre-evacuation phase and the distances
that need to be traversed, and is not a factor in the tesc ¼ 1 þ 6 þ ð0:5Þ þ ð8  38:2Þ=187
calculation. The egress time for an enclosure is ¼ 9:2 min
given by Here, tpe1 is again assumed to be 1 min, while
tesc ¼ t pe1 þ t pe99 þ ttrav ð59:16Þ tpe99 is assumed to be 6 min, with both values
being derived from Fahy and Proulx [25]. The
where tpe99 is the pre-evacuation time for the distance calculations generated by the first-order
last few occupants to respond (i.e. the 99th per- model are used here. In this case Scenario
centile). Purser uses the 99th percentile of a 1 produces the most prolonged evacuation times
representative pre-evacuation distribution to esti- and would therefore be used in the estimation of
mate this value [1]. If this is not available, then a the RSET value.
multiple of the mean pre-evacuation time should If it is now assumed that the space is instead a
be employed; it is not appropriate to use the mid-rise apartment building, rather than an office
average pre-evacuation time. The pre-evacuation space, then different pre-evacuation times are
times usually form a log-normal distribution suggested by the data [25]. In this case,
free ebooks ==>
2142 S.M.V. Gwynne and E.R. Rosenbaum

pre-evacuation values of 1–24 min are used the factors that might inhibit evacuation perfor-
(again derived from Fahy and Proulx [25]). In mance (e.g., tne) and also to limitations in the
this case the calculations become data available (see Chap. 60). The difference
between modeled evacuation movement time
Scenario 1 : tesc ¼ 1 þ 25:3 ¼ 26:3 min
and actual evacuation movement time can be
and expressed in the following terms:

Scenario 2 : tesc ¼ 1 þ 24 þ ð0:5Þ þ ð8  38:2Þ=187 teact ¼ tmod

e e ð59:17Þ
¼ 27:1 min
Once the set of scenarios have been examined, teact ¼ Actual time from when purposive evacua-
the longest tesc value produced should then be tion movement commenced to when safety
employed to generate the RSET calculation. In was reached
this case Scenario 2 produces the most extended temod ¼ Modeled estimate from when purposive
evacuation time and would therefore be used in evacuation movement commenced to when
any RSET calculations. safety was reached
This is certainly not the only approach for e ¼ Modeling error
producing a range of scenarios (or at understand- It is assumed here that the relationship is
ing what factors should be taken into consider- multiplicative; however, the relationship could
ation [51]). However, it does demonstrate how also be additive.
several different scenarios can be considered The modeling error, e, is a function of
using the hydraulic approach allowing elements that interfere with the model prediction.
comparisons to be made between the results pro- In the case of a hydraulic model, this includes
duced. This approach can be taken further by • Delays caused by the egress management
incorporating the other parameters deemed to activities of wardens or others directing the
be amenable to the hydraulic approach; for evacuation
example, manipulating the routes available, • Time delays involved in the stopping and
familiarity, mobility impairments, and so on. restarting of flows at merging points and
The results will always be limited by the conflicting flows
sophistication and fidelity of the model • Evacuee behaviors that detract from their
employed. However, the examination of differ- movement to safety
ent scenarios is critical in providing a reasonable Similar inaccuracies exist in the modeled
understanding of the conditions that might arise. pre-evacuation phase.
Inevitably, there may be cases where data are not From Fig. 59.2 there are many factors that can
available to support these calculations. These interfere with an evacuation, but that cannot be
situations will require engineering judgment. explicitly represented by a hydraulic model. It
These cases should be documented and based should be noted that many of these factors are
on the most reliable and appropriate information also beyond the most sophisticated simulation
available. models currently available (see Chap. 60).
All of these factors can increase the discrep-
ancy between the modeled and actual results.
The first step in appraising emergency movement
Addressing Modeling Error
is usually to calculate the modeled evacuation
time, temod. The use of model calculations
The hydraulic model, whether it is the standard
provides a reproducible base of reference in
or extended version, produces only modeled
appraising the impact of overall systems, individ-
predictions. The actual egress time will exceed
ual components, or changes in systems. If, how-
the modeled time by an unknown amount. This is
ever, the results of the modeled evacuation time
due to the exclusion from the model of many of
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2143

are to represent a realistic evacuation time (or are The most conservative approach requires the
to be compared against expected fire develop- application of the safety factor to the entire
ment), then the engineer should understand that RSET calculation:
the modeled movement time is seldom achieved   
in reality; that is, that e is greater than 1 in Equa- Safety margin ¼ ASET  e0 td þ ta þ t pe þ te mod
tion 59.17. A conservative estimate of the move- ð59:20Þ
ment time requires the modeled time and an
appraisal of modeling error (see Equation 59.17). This is the approach adopted by Tubbs and
This will allow teact to be approximated (or at Meacham [29]. If the same values are assumed
least surpassed) by the modeled time, which is throughout, this approach will generate the larg-
achieved through the application of a safety fac- est RSET value of the three methods (shown in
tor. The employment of a safety factor is a rec- Equations 59.18 through 59.20). This approach
ognition that the hydraulic model omits some does assume that the errors that exist are compa-
factors that may prolong the time to reach safety rable between the behavioral and technical
and/or represents other factors in a simplistic components. Alternatively, separate error factors
manner. might be applied to these components, such that
h 0 0  i
For the design of a structure to be acceptable, Safety margin ¼ ASET  e1 ðtd þ ta Þ þ e2 t pe þ te mod
a sufficient margin of safety is required between
ASET and RSET. In order for the engineer to ð59:21Þ
have confidence in the RSET calculations, a
where e10 may be based on information provided
safety factor, e0 , is employed that approximates
by the manufacturer of the technology involved;
e (i.e., the discrepancy between the modeled and
and e20 is based on the research literature avail-
actual movement time):
able. Given the method adopted, the safety mar-
   gin needs to be acceptable, even after the RSET
Safety margin ¼ ASET  td þ ta þ t pe þ e0 te mod
value has had a safety factor applied. Guidance
on the values to employ in order to estimate the
The application of the safety factor described in modeling error (particularly relating to the
Equation 59.18 is based on the assumption that behavioral components) can be established (see
the inaccuracies are found in the evacuation Chap. 64). The basis for the safety factors
movement component and that these employed should be clearly stated and supported.
inaccuracies need to be addressed (see Chap. 64
and the SFPE Task Group document [43]). In this
case, the engineer would need to be confident in Using the Hydraulic Model
the accuracy of the other components in the cal- in Conjunction with Other Models
culation. A more conservative estimate would be
to apply the safety factor, e0 , to all of the behav- The hydraulic model can be used in a number of
ioral components (i.e., both the pre-evacuation different ways, depending on the resources and
and evacuation phases): expertise available. Currently the expertise in the
  use of hydraulic models far outweighs the exper-
Safety margin ¼ ASET  td þ ta þ e t pe þ te mod
tise in applying simulation models, that is,
ð59:19Þ computer-based models that attempt to represent
the evacuation by simulating the activities of
Although this is more conservative, it does individual agents. However, it is anticipated
require the assumption that the error levels in that this will change in the coming years espe-
the pre-evacuation and evacuation movement cially as larger and more complex spaces are
components are comparable and can be examined. Indeed, the hydraulic model is often
addressed by the same safety factor. calculated using a computer. It should also be
free ebooks ==>
2144 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.14 The three uses a b c

of the hydraulic and
simulation models (a) Used H or S H and S H and S
on their own; (b) used in
parallel; (c) used in
conjunction with each
other [1]
Single set of Two sets of Two sets of results
results produced results produced produced in an iterative
independently of process. Results from S
each other used to examine critical
components, examine
factors beyond scope
of H, confirm results
from H, and suggest
modifications, if required.

recognized that several current computer models project. It is suggested that the hydraulic model
are little more than the hydraulic model coded can be employed in three distinct ways
into computer software (see Chap. 60). There- (Fig. 59.14) [1]. The manner in which it is
fore, as engineers become more familiar with the employed will be dependent on a number of
computer models available and the results that factors including the expertise and the resources
they can produce, the following points will available.
become increasingly apparent: The hydraulic model (identified as H in
• Engineers will gain expertise in a number of Fig. 59.14a) is commonly used on its own to
modeling approaches. determine the evacuation time. Computer simu-
• Engineers will become more familiar with the lation models (indicated as S in Fig. 59.14a) are
capabilities of a number of different modeling also now routinely employed on their own. Alter-
approaches. natively, more than one model can be employed
• Several models will be applied to the same (Fig. 59.14b). The hydraulic and simulation
problem. models may be employed independently of each
These points will have an impact not only on other and then the final results compared. This
how hydraulic models are used but also on the would allow comparisons to be made between
nature of the hydraulic analysis and the the results, the strengths of the different models
expectations of the results produced. Any dis- to be exploited, and the level of confidence in the
crepancy between the results produced by simu- findings to be increased.
lation and hydraulic models (in their format and The application of more than one model
their content) will therefore become more appar- provides benefit but also results in additional
ent to the engineer. Therefore, when the hydrau- efforts since results have to be calculated more
lic model is applied: than once requiring additional time, expertise
• A number of scenarios should be examined. and analysis. However, using multiple models
• The assumptions and limitations should be may provide some engineering benefits (see
identified. Table 59.6). A typical analysis may include
• A detailed set of results should be presented. both a simple, computationally inexpensive
The difference between the typical results model (e.g., hydraulic approach) and a more
produced when employing simulation and refined representation of evacuee response (e.g.,
hydraulic models becomes all the more evident simulation tool). This then allows some addi-
if these models are used together on the same tional confidence in the overall results produced.
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2145

Table 59.6 Engineering benefits of the use of multiple models

Benefit Description
Triangulation Given that there is no absolute confidence in any one model being employed,
the results of several models may be compared to determine whether the
conclusions reached are consistent between different approaches.
Refinement The scenario may require examination of elements of the evacuation process
not represented in the underlying model employed.
Scope The project may be of such a scale (e.g. WTC) that the most refined models
cannot be employed to the whole task. In such projects it may pay for the engineer
to employ the most refined models in critical areas, which have the greatest
influence over the conclusions drawn. These would then be used in conjunction
with the underlying model to assess performance at key spatial or temporal locations.

This would allow comparisons to be made Some simulation tools are currently under
between the results, the strengths of the different development that would allow such coupling of
models to be exploited, and the level of confi- simple and more complex tools within the same
dence in the findings to be increased based on the environment. In this instance, the computational
assumption that consistent model results improve resources available could be targeted at the areas
confidence. deemed to be most critical within the same
Finally, models can be run in an iterative computational environment [52]. Currently,
manner (or even in a coupled manner), with the users need to define in advance which areas of
results of one influencing the scenarios examined the geometry are represented in a refined manner,
by another. For instance, in a project involving while others are represented in a cruder manner.
extremely large structures, the hydraulic model This distinction requires an understanding of the
can be employed to provide an overview of the importance of particular routes, the computa-
results produced, possibly suggesting areas for tional impact of design decisions and the impact
further analysis (Fig. 59.14c). A more “sophisti- that these decisions might have on the results
cated” simulation model could then be used to produced. In the future, these decisions may be
confirm the key assumptions and findings pro- conducted dynamically, where the models allo-
duced by the hydraulic model (e.g., areas of cate locations to one or other of the various levels
congestion) and suggest remedies. The simula- of representation, precluding the need for the
tion tool can confirm the results by examining user to make this judgment in advance.
sections of the structure or events of particular
interest to provide detailed analysis. By using the
simulation model in a more focused way, fewer Impact of Tenability on ASET and RSET
computational resources will be required and the
results produced by the hydraulic model, espe- As part of a performance-based assessment, a
cially in critical locations, can be validated decision has to be made regarding tenability;
(Fig. 59.15). A hybrid approach of hydraulic that is, the point at which the conditions preclude
and simulation analysis may allow for detailed the evacuation to “safely” continue [3]. The ten-
analysis to be conducted, where previously the ability limits will then be used to determine the
cost of a full-scale computational analysis was ASET value: calculations are made to determine
prohibitive. This hybrid approach may also be when the environmental conditions reach the
useful when resources are scarce or the scale of tenability criteria stated. The ASET value pro-
the project is beyond the capabilities of a sophis- duced will then be the benchmark against which
ticated model; for instance, where an evacuation the RSET results will be compared.
involves a business district or complex, rather In reality, environmental conditions can have
than a single building. a behavioral and a physical (physiological)
free ebooks ==>
2146 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.15 Example

application of both
hydraulic and simulation
models Whole structure
represented in
hydraulic model

Critical re-examination of
components whole structure
identified given simulated

Critical components
modeled within
simulation model.
Validate hydraulic
model and perform
more detailed

impact on the performance of the evacuating severe environmental conditions will then have a
population (see Chaps. 58, 61, and 63). Ideally, direct impact on the modeled performance of the
the impact of the tenability criteria should be evacuees.
reflected in the model scenarios employed; that Suggested values are available for tenability
is, if the environmental conditions reach a point limits (refer to Chaps. 61 and 63). These values,
at which they are expected to influence physical or similarly derived empirical values, should be
or behavioral performance, then this should be used to inform the selection of reasonable and
reflected in the model employed. The assumed informed tenability criteria. This critical compo-
occupant performance will have an impact on the nent in the performance-based assessment should
validity of tenability criteria selected and on the be clearly supported in any results reported;
credibility of the results produced. effectively these values determine the amount
There may be a temptation for engineers to of time available to complete the evacuation.
select tenability criteria that artificially prolong Given that these tenability criteria are
the ASET time; that is, that the environmental established, their impact can be reflected within
conditions are allowed to develop to a relatively the hydraulic model. Some data (albeit, in some
severe level before the tenability limits are instances, supported by engineering judgment)
reached, allowing a longer RSET calculation to are available to reflect the impact that a
be acceptable. This situation might include deteriorating environment can have upon egress
severely reduced visibility, elevated temperatures, performance (see Chaps. 58, 61, and 63). Data
and smoke layers descending close to the floor. By relating to physical performance are provided by
coupling the assumed tenability limits with the Jin, who indicates the possible effect that
movement calculations made, there will at least deteriorating visibility has on travel speeds (see
be some counterbalance to these assumptions Chap. 61), specifically in relation to smoke.
(Fig. 59.16), possibly encouraging a more conser- Other physical and behavioral data are also avail-
vative approach to be adopted throughout. The able (see Chaps. 58 and 63). Data on the impact
adoption of tenability criteria that represent of smoke are particularly important given that in
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2147

Fig. 59.16 Coupling

tenability criteria to egress Tenability criteria Smoke levels,
performance employed by temperatures,
engineer radiative flux, etc.

Fire calculations are Conditions produced by

made (e.g. CFD, zone assumed tenability
model, engineering criteria influence
calculations, etc.) evacuation performance


many instances evacuees will be more likely to of smoke can be included into the hydraulic
interact with smoke than with the fire itself. model is shown in Fig. 59.17.
Although there may be many ways in which In Fig. 59.17a, route selection is not
the environment influences evacuee perfor- influenced by the presence of smoke. Within the
mance, it can be simplified especially given that hydraulic model the maximum travel speed
the evacuating population is typically moving to within this component can be reduced to simulate
minimize their exposure. This will produce sev- the impact on the evacuees traveling through this
eral scenarios to be considered (in addition to component when filled with smoke (e.g. at or
those mentioned in the previous sections) when below the tenability threshold). In Fig. 59.17b,
the hydraulic model is employed. These it is assumed that the smoke-filled corridor is not
scenarios relate to the attainable travel speeds used; this is reflected in the hydraulic model by
and the routes available. assuming that a larger population uses the East
The conditions produced (e.g., specific smoke corridor. The potential impact of the two differ-
visibility levels are reached, see Chap. 63) might ent behavioral scenarios can then be compared.
block off certain routes. For instance, it could be Attempting to match the development of the
assumed that when smoke reaches a certain level, fire with the progress of the population in the
a proportion of the population might not proceed hydraulic approach is cumbersome and would
through the smoke. The tenability criteria require numerous additional assumptions. To
reached (e.g., smoke level) could also be deemed assess the maximum impact, a scenario could
to influence attainable travel speeds in the be examined where the entire population refused
affected areas. This can be modeled by reducing to pass through the smoke; similarly a scenario
the maximum achievable travel speed; for could be examined where the entire population is
instance, in Chap. 61 data are provided relating assumed to pass through the affected area at
smoke level to visibility and travel speed. A reduced speeds.
conservative approach might be to assume the It may be impractical to employ these factors
impact at the level of the tenability criteria in within all egress calculations. There is certainly a
the affected spaces (i.e., where the environment lack of supporting data. However, even if these
is deteriorating) throughout the entire evacua- additional scenarios are not considered, the
tion; that is, that whenever tenability criteria are impact that the deteriorating environment can
reached the related impact on travel speed is have on egress performance should at least be
assumed to be present throughout the evacuation. acknowledged and explained in the presentation
An example of the way in which the two effects of any results. Otherwise, it is assumed that the
free ebooks ==>
2148 S.M.V. Gwynne and E.R. Rosenbaum

Fig. 59.17 Example of North

the impact of smoke on the
hydraulic model

Group’s travel
reduced due to
interaction with



These routes
used by more
people as smoke
prevents access
to other route


environmental conditions have no impact upon therefore in conducting performance-based

the performance of the evacuating population analyses. When employing the model, it should
prior to tenability conditions being reached. be remembered that the results produced will be
This may be credible; however, it should be optimistic, and therefore remedial measures
acknowledged. should be employed to compensate for this.
In most designs, there would ideally be no The limitations associated with the model do
physical interaction at all between the evacuating not prevent its being used to examine different
population and the deteriorating environmental egress scenarios; neither do they excuse the over-
conditions. That does not preclude the population simplistic use of this model or presentation of the
seeing the developing conditions, which might results. Those employing this model need to pro-
influence their behavior. Even here the untenable vide sufficient information on the approach
conditions could be modeled in the hydraulic adopted, the assumptions made, the scenarios
model through the loss of available egress routes. examined, and the results produced.
When using the hydraulic model, it is still
possible to examine a number of evacuation
Summary scenarios and incorporate the effect of various
factors on the performance achieved. Given the
This chapter described the application of the potential for hydraulic and simulation models to
hydraulic model and its capabilities in assessing be employed together, it becomes even more
emergency movement. The model is able to pro- important to provide comparable levels of detail
vide a reliable means of assessing RSET and and confidence in the results produced.
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2149

Acknowledgment The authors acknowledge that this teact Actual time from when purposive evac-
work includes substantial sections from the original chap- uation movement commenced to when
ter written by Harold “Bud” Nelson in partnership with
Hamish MacLennan and then Fred Mowrer. safety was reached
temod Modeled estimate from when purposive
evacuation movement to when safety
was reached
Nomenclature e Modeling error
e0 Approximation of e employed within
ASET Available safe egress time calculation

RSET Required safe egress time

td Time from fire ignition to detection
tn Time from detection to notification of 1. D.A. Purser and S.M.V. Gwynne, “Identifying Criti-
occupants of a fire emergency cal Evacuation Factors and the Application of Egress
tpe Time from notification (or receipt of Models,” Interflam 2007, Interscience
cues) until evacuation commences Communications, London, UK (2007).
2. D. Boswell and S.M.V. Gwynne, “Air, Fire and ICE:
te Time from start of purposive evacua- Fire & Security Challenges Unique to Airports,” Fire
tion movement until safety is reached and Security Today (Aug. 2007).
tesc Escape phase, being the sum of the 3. H. Nelson, personal communication (2007).
pre-evacuation (tp-e) and evacuation 4. J.D. Sime, “Escape from Building Fires: Panic or
Affiliation?” Ph.D. Dissertation, University of Surrey,
(te) phases UK (1984).
ttrav Time spent moving toward a place of 5. E.R. Galea, Validation of Evacuation Models, CMS
safety Press Paper No. 97/IM/22, CMS Press, London, UK
tflow Time spent in congestion controlled by (1997).
6. E.D. Kuligowski and S.M.V. Gwynne, “What a User
flow characteristics Should Know About Selecting an Evacuation Model,”
tne Time spent in nonevacuation activities Fire Protection Engineering, Human Behaviour in
that do not directly contribute to the Fire Issue (Fall 2005).
population moving to a place of safety 7. V.M. Predtechenskii and A.I. Milinskii, Planning for
Foot Traffic in Buildings (translated from the
We Effective width Russian), Stroizdat Publishers, Moscow (1969).
D Population density English translation published for the National Bureau
S Travel speed of Standards and the National Science Foundation,
k Constant used to calculate travel speed Amerind Publishing Co., New Delhi, India (1978).
8. J.J. Fruin, Pedestrian Planning Design, Metropolitan
a Constant used to calculate travel speed Association of Urban Designers and Environmental
Fs Specific flow Planners, Inc., New York (1971).
Fsm Maximum specific flow 9. J.L. Pauls, “Effective-Width Model for Evacuation
Fc Calculation flow through a component Flow in Buildings,” in Proceedings, Engineering
Applications Workshop, Society of Fire Protection
tp Time for a group of persons to pass a Engineers, Boston (1980).
point in an exit route 10. J.L. Pauls, “Calculating Evacuation Time for Tall
P Population size in persons Buildings,” in SFPE Symposium: Quantitative
Fs() Specific flow and associated direction Methods for Life Safety Analysis, Society of Fire Pro-
tection Engineers, Boston (1986).
of movement 11. R.J.C. Stanton and G.K. Wanless, “Pedestrian Move-
We() Effective width of a particular compo- ment, Engineering for Crowd Safety,” in Engineering
nent given its location for Crowd Safety (R.A. Smith and J.F. Dickie, eds.),
tpe1 Pre-evacuation time of the first people pp. 71–79, Elsevier, Amsterdam (1993).
12. K. Ando, H. Ota, and T. Oki, “Forecasting the Flow of
to respond People,” Railway Research Review, 45, pp. 8–14
tpe99 Pre-evacuation time of the last people (1988).
to respond
free ebooks ==>
2150 S.M.V. Gwynne and E.R. Rosenbaum

13. B.D. Hankin and R.A. Wright, “Passenger Flow in 29. J.S. Tubbs and B.J. Meacham, Egress Design
Subways,” Operational Research Quarterly, 9, pp. Solutions: A Guide to Evacuation and Crowd Man-
81–88 (1958). agement Planning, John Wiley and Sons, New York
14. G. Proulx, “Lessons Learned on Occupants’ Move- (2007).
ment Times and Behaviour During Evacuation 30. S.M.V. Gwynne, Optimizing Fire Alarm Notification
Drills,” Interflam 96, Interscience Communications, for High Risk Groups, Report Prepared for the Fire
London, UK, pp. 1007–1011 (1996). Protection Research Foundation, National Fire Pro-
15. J.-P. Stapelfeldt, Angst- Und Paknikstande Aus Der tection Association, Quincy, MA (June 2007).
Sicht Des Brandschutzes (Conditions of Anxiety and 31. D. Canter, Fires and Human Behaviour, 2nd ed.,
Panic from the Viewpoint of Fire Protection), vfdb- Fulton, London, UK (1990).
Zeitschrift 2, 41 (1986). 32. T.J. Shields, K.E. Dunlop, and G.W.H. Silcock,
16. S.J. Melinek and S. Booth, “An Analysis of Evacua- “Escape of Disabled People from Fire; A Measure-
tion Times and Movement of Crowds in Buildings,” ment and Classification of Capability for Assessing
BRE Current Paper CP 96/75, Borehamwood, UK Escape Risk,” BRE Report 301 (1996).
(1975). 33. H. Muir, C. Marrison, and A. Evans, “Aircraft
17. R.A. Smith, “Volume Flow Rates of Densely Packed Evacuation: The Effect of Passenger Motivation and
Crowds,” in Engineering for Crowd Safety Cabin Configuration Adjacent to the Exit,” CAA
(R.A. Smith and J.F. Dickie, eds.), pp. 313–319, Paper 89019, Civil Aviation Authority, London, UK
Elsevier, Amsterdam (1993). (1989).
18. A. Polus, J.L. Schofer, and A. Ushpiz, “Pedestrian 34. P.G. Wood, “The Behavior of People in Fires,” Fire
Flow and Level of Service,” Journal of Transporta- Research Note 953, Building Research Establishment,
tion Engineering, Proceedings ASCE, 109, pp. 46–57 Fire Research Station, Borehamwood, UK (1972).
(1983). 35. J.L. Bryan, Smoke as a Determinant of Human Behav-
19. S.J. Older, Pedestrians, Dept. Scientific and Industrial ior in Fire Situations, University of Maryland, Col-
Research, Road Research Laboratory, Ln 275/SJO, lege Park (1977).
Crowthorne, England (1964). 36. J.P. Keating and E.F. Loftus, “Post Fire Interviews:
20. I.A.S.Z. Peschl, “Passage Capacity of Door Openings Development and Field Validation of the Behavioral
in Panic Situations,” BAUN, 26, pp. 62–67 (1971). Sequence Interview Technique,” Report
21. L.F. Henderson, “The Statistics of Crowd Fluid,” GCR-84–477, National Bureau of Standards,
Nature, 229, pp. 381–383 (1971). Gaithersburg, MD (1984).
22. S.P. Hoogendoorn and W. Daamen, “Pedestrian 37. B. Latane and J.M. Darley, “Group Inhibition of
Behavior at Bottlenecks,” Transportation Science, Bystander Intervention in Emergencies,” Journal of
39, 2, pp. 147–159 (2005). Personality Psychology, 10, 3, pp. 215–221 (1968).
23. A. Seyfried, T. Rupprecht, A. Winkens, O. Passon, 38. G. Proulx, “Movement of People: The Evacuation
B. Steffen, W. Klingsch, and M. Boltes, “Capacity Timing,” in SFPE Handbook of Fire Protection Engi-
Estimation for Emergency Exits and Bottlenecks,” in neering, 3rd ed. (P.J. DiNenno et al., eds.), National
Interflam 2007, Interscience Communications, Fire Protection Association, Quincy, MA,
London, UK (2007). pp. 3-342–3-366 (2002).
24. J.D. Averill and W. Song, “Accounting for Emer- 39. A.T. Habicht and J.P. Braaksma, “Effective Width of
gency Response in Building Evacuation: Modeling Pedestrian Corridors,” Journal of Transportation
Differential Egress Capacity Solutions,” NISTIR Engineering, 110, 1 (1984).
7425, National Institute for Standards and Technol- 40. J. Milke, personal communication (2005).
ogy, Gaithersburg, MD (2005). 41. NFPA 101®, Life Safety Code®, National Fire Pro-
25. R.F. Fahy and G. Proulx, “Toward Creating a Data- tection Association, Quincy, MA (2006).
base on Delay Times to Start Evacuation and Walking 42. J. Fruin and J. Pauls, Human Factors of Means of
Speeds for Use in Evacuation Modelling,” in 2nd Egress History, Current Problems and Implications
International Symposium on Human Behaviour in for the Future, World Safety Conference and Exposi-
Fire, Interscience Communications, London, UK, tion, Session SU47, National Fire Protection Associa-
pp. 175–183 (2001). tion, Boston (2007).
26. H.E. Nelson and H.A. MacLennan, “Emergency 43. SFPE Task Group on Human Behavior, Engineering
Movement,” in The SFPE Handbook of Fire Protec- Guide to Human Behavior in Fire (June 2003)
tion Engineering, 2nd ed., (P.J. DiNenno et al., eds.), 44. BS5588, Fire Precautions in the Design, Construction
National Fire Protection Association, Quincy, MA, and Use of Buildings, BSI British Standards, London,
pp. 3-286–3-295 (1996). UK (2004).
27. J. Bryan, “A Selected Historical View of Human 45. J.D. Sime, “Visual Access Configurations: Spatial
Behavior in Fire,” Fire Protection Engineering, Analysis and Occupant Response Inputs to Architec-
pp. 4–16 (Fall 2002). tural Design and Fire Engineering,” in IAPS Confer-
28. S. Gwynne, E.R. Galea, M. Owen, and P.J. Lawrence, ence, Eindhoven, The Netherlands (Aug. 1997).
Escape As a Social Response, Society of Fire Protec- 46. BSi PD-7974-6:2004, The Application of Fire
tion Engineers, Bethesda, MD (1999). Safety Engineering Principles of Fire Safety
free ebooks ==>
59 Employing the Hydraulic Model in Assessing Emergency Movement 2151

Design of Buildings, Part 6: Human factors: Life safety Publishing Corp., Newport, Australia, pp. 561–570
strategies—Occupant evacuation, behaviour and con- (1986).
dition (Sub-system 6), British Standards, (2004) 52. N. Chooramun, P.J. Lawrence, E.R. Galea, An agent
47. D.A. Purser, personal communication (2007). based evacuation model utilising hybrid space
48. D.A. Purser, “People and Fire,” Inaugural discretisation, Safety Science, Vol 50, pp
Lecture Series, University of Greenwich, London, 1685–1694, 2012.
UK (1999).
49. S. Gwynne, E.R. Galea, J. Parke, and J. Hickson, “The
Collection and Analysis of Pre-Evacuation Times Steven M.V. Gwynne, PhD is a senior research officer at
from Evacuation Trials and Their Application to the National Research Council Canada. He specializes in
Evacuation Modelling, Fire Technology, 39, 2, pp. pedestrian and evacuation dynamics. His work includes
173–195 (2003). the collection of data, the development of behavioral
50. J. Parke, S. Gwynne, E.R. Galea, and P. Lawrence, theories, and the application of models to assess people
“Validating the building EXODUS Evacuation Model movement under emergency and nonemergency
Using Data from an Unannounced Trial Evacuation,” scenarios.
in Proceedings of the 2nd International Conference
on Pedestrian and Evacuation Dynamics (PED 2003), Eric R. Rosenbaum, P.E., is Vice-President of Jensen
CMS Press, University of Greenwich, London, UK, Hughes, Inc. a fire protection engineering, code consult-
pp. 295–306 (2003). ing, and research and development firm. He has been
51. J.D. Sime, “Perceived Time Available: The Margin of actively involved in the evaluation of egress facilities for
Safety in Fires,” in Fire Safety Science—Proceedings structures throughout the world.
of the First International Symposium, Hemisphere