Engineering Data

Abstract

This chapter is an updated version of the previous chapter “Evacuation Timing” that appeared in the fourth edition of the SFPE Handbook. This new version of the chapter represents a significant change to previous versions, moving from a narrative description of important case studies that include data to a tabular representation of a broader range of data-sets. It is hoped that this approach provides a useful reference resource for readers [1, 2].

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

This chapter is an updated version of the previous chapter “Evacuation Timing” that appeared in the fourth edition of the SFPE Handbook. This new version of the chapter represents a significant change to previous versions, moving from a narrative description of important case studies that include data to a tabular representation of a broader range of data-sets. It is hoped that this approach provides a useful reference resource for readers [1, 2].

Evacuation and human behavior in fires is an important consideration in any fire safety engineering design. Research in this area essentially started in the mid-1950s with the work of John Bryan [3] and was continued over the proceeding decades by numerous researchers from within fire safety and those in adjacent areas of research [4]. A renewal of enthusiasm in this work was observed with the arrival of many new researchers into the field in the 1990s and the advent and adoption of performance-based fire safety design (see Chap. 57). The improved understanding that this work generated, and the subsequent modeling capabilities that it both suggested and facilitated, forms the basis of a key element of the fire safety equation; i.e. comparing the time for a population to reach safety with the time for the conditions to become untenable, typically adopted as part of performance-based design (see Chap. 57). This work is enabled and driven by the availability of detailed, comprehensive and appropriate data to support the development of theories and engineering practice alike. This chapter represents an attempt to help the reader locate and select data for use in this work.

It is generally (although not universally) acknowledged that “hard” fire science alone cannot solve the fire problem; knowledge of human behavior is also essential [5]. The vagaries of human performance cannot be eliminated from the design process. There is a degree of skepticism in the field of fire safety engineering as to whether human behavior in fire can be understood and then represented credibly in egress analysis [6]. However, the expansion of this section of the handbook and the increasing reference to human response in regulatory documentation and guides [7, 8] is nothing if not evidence of the broad recognition that, although difficult, it is a ‘necessary evil’. This chapter provides a summary of some of the key data (predominantly produced after 1985) that might support those in the fields of fire safety and fire engineering who are attempting to understand and represent human behavior in fire. This chapter should provide a useful introduction to this data and point to key reference sources for the practitioner to employ in their work.

Work on understanding evacuation and human behavior in fire, developing tools to assess it and then producing designs that account for it, is multi-disciplinary in nature and requires expert input from a variety of sources. This input may take the form of ideas, models, theories and data. The variety of this input produces new ideas in the field, disparate opinions and expertise. It also introduces many inconsistencies in the terminology and techniques employed, the assumptions made and the overall objectives. These inconsistencies may influence model development, configuration, application and then the interpretation of the results produced by the model.

It should also be noted at the outset that human behavior in fire is still a relatively immature area of analysis. This is true of our understanding of basic phenomena (see Chap. 58), but particularly true of the numerical data-sets available to describe egress performance. These data originate from different sources, are provided in different formats, are based on different assumptions and described in different ways [9]. This chapter attempts to take these differences into account allowing the reader to select between the data available in as informed a manner as possible.

Given the variety of data sources available and the data formats employed, only a summary of each selected data set is provided in this chapter, providing insight into the data available and aiding the data selection process. As such this chapter provides a resource to inform the reader’s selection of data, but it is essential that he/she should further investigate the data from the original source as part of the selection process. The limitations of the data available are acknowledged and therefore the informed selection between these often partial and incomplete data-sets becomes all the more important.

Using this Chapter

This chapter is intended to aid the representation of egress performance within engineering models (see Chaps. 59 and 60). There is an enormous variety in the methods adopted and the assumptions made in these models—and therefore in the data used by them. The model may be embedded within regulatory codes (i.e. the behavioral assumptions made in order to develop the regulatory rules), it may be an empirical model (where the data forms the very basis of the model), an engineering model (where data is used to derive relationships that then represent some aspects of the response), or a simulation model (where data allows the more complex computer-based tools to be configured, calibrated and validated, see Chaps. 59 and 60). This variety is reflected in the impact that the selection of different data-sets will have on the results produced by these models.

This chapter will support the use of these models by presenting data that could be used in the quantification of egress performance as part of regulatory assessment (the primary function), and presenting data that enables the examination of design variants, developing procedures, or further research. The chapter is intended to provide an overview of this data. It will provide, in most cases, a tabular summary of a range of data-sets available for the basic behavioral elements required to assess egress performance. These elements, derived from the basic engineering timeline which is typically used to describe expected performance phases [10], will be briefly discussed. The data-sets will be presented in the same format within each of the core behavioral elements identified, thus simplifying comparison between them. These behavioral elements have been deliberately selected as those that are fundamental to the most basic egress analysis (see Chaps. 59 and 60). It is hoped that this approach will allow the reader to compare the appropriateness of the data-sets shown and then follow up on the details of the data-sets once an initial selection has been made. The tables will provide information sufficient for the reader to identify whether the data set is representative of the scenario that they wish to represent. However, it is essential that the reader accesses the original source of the data set (also provided in the table) for further information to enable them to utilize it with confidence. It is acknowledged that this chapter presently excludes some data that might be of value; specifically, data regarding the decision-making process during the evacuation is not included. Although this data is critical, it is not widely employed within the modeling or the performance based design process (irrespective of how much the authors would like it to be). Given space constraints, this is left to be addressed elsewhere (see Chap. 58).

This chapter is not intended to provide definitive guidance on which of the data-sets should be used in particular scenarios of interest. This decision should be taken through examining this chapter in conjunction with several of the other chapters in this handbook. It would currently be impractical to provide definitive data-sets for the myriad scenarios that may be examined and the many engineering models that might be employed [11, 12]. This chapter is, therefore, intended to provide the reader with sufficient information to identify data-sets of interest and then explore them further; i.e., provide an initial reference for the fire safety engineer to focus their search and guidance to aid in this search. In addition, there are a number of online resources that the reader may consult to find contemporary reviews of the data available [13–16]. Even here, the reader is advised to seek out the original source material to ensure suitability.

The data-sets might be used in a number of ways including:

• Configuring hydraulic calculations for use e.g. augmenting the guidance provided in Chap. 59.

• Developing hydraulic calculations independently from Chap. 59.

• Configuring egress tools for use (see Chaps. 59 and 60). For instance, providing model parameter values to be used in representing a particular scenario (see Chap. 57).

• Developing egress tools (see Chaps. 59 and 60).

• Examining the assumptions of the work of others (see Chap. 57).

• Performing expert analysis i.e. helping to support a particular expert opinion regarding evacuee performance.

The approach outlined has been adopted to ensure that it benefits a range of potential readers, although primarily focused at the fire safety engineer. It is expected that engineers, designers, those assessing designs (e.g. Authorities Having Jurisdiction (AHJ)), model developers and researchers may find value in the data-sets provided. These users will have different requirements of the data-sets presented and needs that might only be completely satisfied through a full and lengthy description of each data-set.

Using the Data Provided

The data described in this chapter should not be employed without an understanding of the subject matter involved; i.e. human behavior in fire. The data-sets provided in this chapter are intended to be used in conjunction with the methods and data provided in a number of the companion chapters in this handbook. It is therefore highly recommended that the reader examines all of the chapters in this section prior to selecting any of the data-sets for further use. This should then at least provide a basic understanding of the key factors and behavioral elements to be taken into consideration during the selection process. When taken as a whole, this set of chapters should provide the reader with sufficient information to understand expected human response, quantify this response, represent this response as part of engineering analysis (within a design scenario) and then interpret and present the results in an informed manner. A schematic of the assumed relationship between the subject areas covered in these chapters is presented in Fig. 64.1, although it is recognized that the exact relationship between these chapters will be somewhat dependent upon the nature of the application.

Fig. 64.1

Relationship between subject areas in this section

As noted above, the user will typically require a basic understanding of the subject matter involved. The currently available theories which explain human response in fire are provided in Chap. 58. This chapter should form the basis of any analysis as it describes the subject matter involved and enables the reader to develop a qualitative understanding of the evacuation process with the evacuee as the focus before attempting to quantify performance. The nature of the engineering design process and the emergency procedures that might appear within that process are described in Chaps. 56 and 57. These chapters formulate the way in which the behavioral theories might be applied; i.e. what aspects of the underlying subject matter should be represented and what impact might they have in a real-world environment. Chapters 61 and 63 provide data to quantify this design process; they provide a numerical underpinning to the qualitative and scenario-based descriptions outlined earlier. These chapters are vital in translating the real-world elements into the simulated environment. Finally, Chaps. 59 and 60 describes the models that might be applied to utilize this data in order to assess performance within the simulated environment.

Quantifying Egress

This section discusses the key factors that influence the quantification of egress. In section “Subject Matter: Human Behavior in Fire”, the primary subject matter of egress analysis, i.e. human behavior in fire, is briefly discussed. An understanding of the subject matter directly influences how the models are constructed, the data that is deemed to be required and then use of the data within the quantification process.

Section “Engineering Timeline” describes the engineering timeline. This is important as it defines, at a high-level, the core components that need to be represented during the engineering analysis and therefore that need to be supported by data. As has been noted, in reality, individuals operate at a more refined level and in a more complex manner. It is the discrepancy between the individual and engineering levels that often leads to misinterpretation of the data available. It is anticipated that the following sections will help the reader address some of the issues highlighted.

Subject Matter: Human Behavior in Fire

In contrast to the panic model previously assumed, the response of an individual during an incident can be better characterized as involving a decision‐making process [17, 18]. This process is influenced by internal and external elements that can interact and change as the incident progresses. If this response is over-simplified or ignored, then the engineer may neglect key elements that they need to represent using the data available. It is therefore important to understand the key subject matter elements that may be represented or ignored in specific data-sets in order to assess their appropriateness for the application in question. These elements include:

• The procedures in place within the structure as dictated by the resident organization,

• The environmental conditions to which the individual is exposed at any point in time,

• The actions of those around the individual given the relationships that exist between them,

• Τhe physical, cognitive, sensory and experimential attributes of the individual,

• The information available and the sub-set of this information that is noted by the individual (given their current alertness, attention levels and actions),

• The individual’s perception of this information, their ability to understand it and to assess the situation and the threat posed to them or significant others,

• The individual’s role and position within the current group/organizational/social structure, and the associated norms and responsibilities with this role,

• The viable options available to the individual in any situation,

• The individual’s ability to assess and select a response option, given the temporal, physical and social constraints present, and

• The ability of the individual to enact this option [19].

The decision making process is a highly coupled, iterative process where each of the elements are often interchanged and modified in real-time, depending on the scenario faced. Understanding this process and the factors that influence response is critical for model development and model application; i.e. critical to the engineering process in general [20, 21]. Although this process is complex, it is comprehensible and is in contrast to both the panic model (where an almost complete absence of process is assumed) and stimulus-response approach (where the process is reduced to individual responses initiated only by external triggers).

The decision making process is then an important influence upon performance and therefore upon the data that might be collected. Any data collected or used will need to account for the key elements of this process (see Chap. 58). However, these elements may be combined, simplified or deliberately omitted in the data representation (along with other limitations with our understanding of the subject matter). It is therefore important that the manipulation of these elements in the original data analysis needs to be clearly understood and justified this can only be achieved through an understanding of the decision-making process itself.

This description of the decision-making process is very much from the individual’s perspective. The combined actions and interactions of individuals produce emergent conditions by which a design might then be judged during an engineering analysis (see section “Engineering Timeline” discussing the engineering timeline). Within the engineering analysis, these emergent conditions may be predicted by the model employed or imposed by the user [22]. The data will need to reflect some aspect of the engineering timeline (see Fig. 64.2). In reality the partitions present in the timeline may not be sufficient to reflect the complexity and variety of the individual decision-making processes or the associated data. It is therefore important for the engineer to understand the compromises being made by selecting certain data-sets.

Fig. 64.2

Engineering timeline

The individual decision-making process is described in more detail in Chap. 58. This process can also be accounted for within the scenario design; i.e., the outcome of the decision-making process can be reflected in the scenarios to which the egress models are employed. This is described in Chap. 57. In addition, the manner in which the decision-making process and associated behavioral actions are represented within the egress models applied is described in Chaps. 59 and 60. As is apparent, the assumption that the individual goes through a decision-making process during an evacuation is fundamental to many of the chapters in this section. In this chapter, the decision-making process influences the categorization of the data-sets and the manner in which the data is presented.

Engineering Timeline

Typically, engineering and computational models are employed to establish an engineering timeline, or some aspect of it. This timeline represents the total evacuation process, which is broken down into several distinct components. This timeline typically operates at the level of the population, rather than the individual and is therefore a simplification of numerous individual actions and responses that take place, the complexity of which has been alluded to above and in Chap. 58. The components within the timeline are therefore aggregates of lower level performance, based on technical or human resources. Different versions of this timeline are available in numerous regulatory and guidance documents [7, 22, 23]. A version of this timeline is shown in Fig. 64.2. This timeline also represents subject areas discussed in other chapters in this Handbook.

The engineering timeline shown in Fig. 64.2 includes the following components:

• Detection time (t _det )—The interval between fire ignition and the first detection of the fire by a device or an individual.

• Warning time (t _warn )—The interval between detection of the fire and the time at which an alarm signal is activated or notification of occupants takes place.

• Pre-evacuation time (t _pre )—The interval between the time at which a general alarm signal or warning is given and the time at which the first deliberate evacuation movement is made. This consists of two components: recognition time and response time.

• Recognition time (t _rec )—The interval between the time at which the alarm signal is perceived and the time at which the occupant interprets this signal as indicating a fire/emergency event. This time includes investigation and milling, for example, to determine the situation.

• Response time (t _res )—The interval between recognition time and the time at which the first move is made to evacuate the building. This time includes activities such as fire-fighting, warning others, gathering family members and pets, dressing, retrieving personal belongings, calling the fire department, and so on.

• Travel time (t _trav )—The time needed, once movement toward an exit has begun, for all occupants to reach a place of safety.

• Evacuation time (t _evac )—The time from the alarm signal to the time at which the occupants reach a place of safety. This is the sum of the pre-evacuation time (t _pre ) and the travel time (t _trav ).

• Required Safe Escape Time (t _RSET )—The calculated time necessary between ignition of a fire and the time at which all occupants can reach an area of safety. This is the sum of the detection time (t _det ), the warning time (t _warn ) and the evacuation time (t _evac ). This is equivalent to the escape time (t _esc ).

• Available Safe Egress Time (t _ASET )—The calculated time available between ignition of a fire and the time at which tenability criteria are exceeded in the means of egress. t _ASET should be longer than t _RSET by an acceptable margin of safety.

The terms highlighted in bold are the focus of this chapter. It should be noted that the terms used to describe t_RSET can vary. The term Required Safe Escape Time [7] is adopted here, but Required Safe Egress Time [23] is also often used. The exact terminology employed to describe the other components also differs based on its origin and intended use. This has been described in detail by Gwynne [9], where a range of different terms is explored. A summary of the various terms employed for two of the key behavioral components, t _pre and t _trav are provided in Table 64.1.

Table 64.1 Range of terminology used to describe behavioral components

The engineering timeline extends from the time of ignition to when untenable conditions are reached. The time of ignition is the starting point of the fire event. Following ignition, a time should be calculated for detection to take place. Detection could take anywhere from a few seconds to a few hours depending on the type of fire and the detection devices in place. Detection could be accomplished by staff/occupants who discover the fire or perceive cues of the incident (e.g., smell of smoke) or through the activation of a detection system.

An elapsed time might need to be calculated between detection and alarm activation. In some cases, these two events are almost simultaneous; for instance, smoke detectors are often linked to a general alarm which issues an immediate warning signal. There could, however, be a delay between detection and alarm, for example, if occupants discover a fire and have to manually activate the fire alarm signal at a pull-station, or where a procedure is in place that requires confirmation of the incident by a member of staff and staff communication to allow the public alarm to be provided [11, 12].

The first two components of the RSET calculation i.e., t _det and t_warn, can be formed from a combination of technological and/or human resources and the balance of these resources will depend on the scenario and on the procedural measures in place to respond to it. Where automatic detection and alarm is employed, typically, t _det and t _warn will be determined with reference to manufacturer’s guidance on sensor response and often do not take staff activities into account. However, this is not always the case and in some instances human resources play a more prominent role and therefore need to be accounted for within the engineering timeline.

The selection of the procedural resources employed to detect and alert will be influenced by a number of factors; e.g., the cost, the infrastructure, the nature of the occupancy and the expected incident scenarios. However, it is certainly possible that staff (or other occupants) will be involved in the detection of the original incident and the raising of the alarm and it is important to recognize that this involvement might delay the point at which the pre-evacuation phase begins for the general population; i.e., there is a potential delay as staff interpret the cues available to them and then respond.

Any delay in detection and warning (also previously referred to as Pre-Warning time [11, 12, 24, 25] may be procedural and/or cognitive (i.e. part of the formal procedure) and/or dependent on the decision-making activities of the staff in question. The need for and extent of human involvement in issuing a warning can vary depending on the emergency response procedures of the affected building and the detection and warning systems used.

Typically, the delay in detection and warning is either not accounted for in engineering analysis or is based on the time for the technical system in place to detect the incident, rather than the staff response to it. Just as with the pre-evacuation time for an evacuee, these times i.e. t _det and t _warn can amount to a significant time period depending on the nature of the event, the technological and human resources available, the nature of the space, the staff involved, the staff hierarchy present and the procedures in place.

There are often legitimate reasons as to why staff involvement prior to the raising of the alarm is deemed necessary in some buildings, especially major public venues: (a) an evacuation can lead to significant disruption to the normal operation of the building; (b) starting an evacuation automatically regardless of the circumstance is not only commercially costly, but also unacceptable to the public, and may prejudice later evacuee interpretation of emergency alarms; (c) evacuation is not without risk and can produce additional (and more significant) hazards than the original incident; (d) the evacuation may need to be carefully managed, requiring staff to be briefed and in position before commencement or (e) depending upon the circumstances a full scale evacuation of a building may not be the most appropriate course of action. Many of these potential procedural influences can be recognized during the design phase and, as such, the engineer may be able to factor these into the design process through the use of an extended warning/delay time.

Often, detection and notification systems have ‘failsafe’ measures to ensure that once an incident has been detected it is not entirely ignored. For instance, should a detection signal not be acknowledged or acted upon within a certain period of time e.g., 2 min then a general alarm will automatically be signalled. If enabled, these may provide an upper limit as to the warning time in such a procedure. However, where this measure is deliberately overridden, disabled or absent, then the warning delay may not be limited or predicted by the ‘failsafe’ setting associated with the system. Again, the potential for this may be acknowledged within the design process, albeit that it is difficult to quantify.

The engineer might therefore choose to include a detection and warning delay component to account for the staff decision-making and response activities, especially if they form part of the formal procedure and can therefore be expected to take place. In effect, the warning period could be better represented by including an estimate of the accumulation of individual staff decision-making processes and subsequent activities that go to delay the provision of the warning.

When a cue, a warning by a notification system or an individual has been perceived, then the population has the opportunity to commence evacuation. However, typically there is a delay between initial awareness of the incident and purposive movement towards safety, the pre-evacuation time (t _pre ) which is due to the population’s interpretation of the information available and/or actions that they might need to perform prior to evacuating. Purser (and others) identified that pre-evacuation time comprises of two subcomponents; i.e. recognition and response [26]. In the recognition phase, the occupant will perceive information such as the fire alarm signal, the sight of smoke, or warning by others. Interpretation of this information may take some time. As information on the unfolding event is perceived, the occupant will start responding by taking different actions such as investigating the situation, attempting to fight the fire, calling for help, and gathering belongings and family members i.e. actions that do not necessary move them closer to a place of safety. Eventually the decision to evacuate the building, to search for a refuge or to protect-in-place will be taken, which will complete the pre-evacuation phase (assuming that the entire population decides to eventually evacuate). In this chapter, the data is typically presented in terms of the overall pre-evacuation time. Where the data is broken down to a lower level (e.g. recognition and response), this is also noted.

The travel time only starts when the occupant has initiated movement towards a place of safety, whatever that might be. In this context, travel relates to the movement performed as part of the evacuation process. It is acknowledged that an individual may also move during the pre-evacuation phase, but it is assumed here that this is addressed in the pre-evacuation delay experienced. In its most basic form, travel time should be calculated from the traversal time to move along the specified egress route and the flow time through various elements of the egress system (as discussed in Chap. 60). In reality, an evacuee may engage in numerous activities during this phase that are not directly involved in moving towards a place of safety (see Chap. 58). However, these are not typically (or explicitly) represented in an egress analysis, except where they involve an evacuee responding to the evacuation movement of others; e.g. being delaying in congestion.

All the times previously discussed together comprise The Required Safe Escape Time (t_RSET) is calculated from the time of ignition until the last occupant to evacuate has reached a location of safety. This time should be less than the time for untenable conditions to occur in the egress path (t _ASET ). The time between the population reaching safety and the conditions becoming untenable should be sufficient to provide an acceptable margin of safety (t _marg in Fig. 64.2). The t _RSET value can therefore be formulated as follows

$$ {t}_{RSET}={t}_{det}+{t}_{warn}+{t}_{pre}+{t}_{trav} $$

Depending on the scenario, each of the t _RSET phases can have a significant impact on the time for the target population to reach safety. Although there is some general (primarily qualitative) discussion of the detection time, t _det and the warning time t _warn in section “Detection and Warning Phases: Human Aspect”, of this chapter, the focus here is on the primary behavioral components: t _pre and t _trav . The data-sets described will relate specifically to the quantitative assessment of these components and the factors that might contribute to this assessment.

In section “Model Approaches and Data Requirements”, the approaches adopted within egress models are discussed, specifically relating to the requirements that these approaches place on the data available.

Model Approaches and Data Requirements

Establishing the time for a population to reach safety is relatively complex, and it is considerably more difficult than estimating movement time alone (see Chap. 58 and previous section). The various attempts at representing this process as part of egress analysis are described in Chaps. 59 and 60.

There are a range of different models that are used to explicitly or implicitly establish egress performance; i.e. models that produce a quantitative assessment or assume a performance level. These models require a range of data either in their development or their application in support of the behavioral model in situ [22]. The various types of models employ different techniques, cover different areas of the evacuation process and operate at different levels of sophistication and refinement. As a consequence, models require data in different formats and to address different subject matter. Models can be broadly categorized into six different types, each with their own data needs:

• 1. Prescriptive Codes: Pre-defined rules based on experience (i.e., expertise and lessons learned from real incidents), that are then codified into a guidance/regulatory framework. Data are used to support the development of these rules, which form an implicit behavioral model. This data is not used to subsequently apply the model (as might the case in other types of models) but rather to support the development of the rules.

• 2. Full-scale evacuation demonstration: The use of a representative population and scenario(s), e.g. the evacuation of an actual office block to gain insight into performance of a structure under specific conditions. Data may be used to help inform expectations regarding performance and then organize the management and data-collection activities. In this context, the mock evacuation is taken as an indication of how the population may perform during an actual incident; i.e., is a model of reality.

• 3. Theoretical Model/Expert analysis: Data-sets are used to develop a theory describing some performance component. A set of theories are then employed as part of expert analysis/engineering judgment to assess some issue. This assessment may well use further data to support the analysis and make it more specific. This process is highly dependent on the availability and use of data from the development of the theories to their application.

• 4. Engineering calculation at the level of the Structure, based on empirical correlation/analysis: Empirical data relating to an entire structure are analyzed to produce high-level functions to predict performance assuming similarities (i.e., at the structural level). Data-sets are used directly in the production of the model and in the application to examine the performance of other similar structures. Data-sets may also be used in verification and validation of the model developed.

• 5. Engineering calculation at the level of the Component (Hydraulic models): Data-sets are collected from the evacuation of structures and then analyzed to produce low-level component-based functions to predict performance at the component level (e.g., doorway, corridor, stair, etc.). These are then chained together to represent the egress performance along paths out of the structure based on the flow, speed and density relationships assumed and the structural and occupancy assumptions made. Data-sets are then used directly in the production of the model and in the application. Data-sets may also be used in verification and validation of the model developed.

• 6. Computational Egress models: These tools include the coding of the previous three bullets. These provide different levels of sophistication and have different data requirements depending on the nature of the model itself. These may require data both to develop and configure the models for use. Data-sets may also be used in verification and validation of the model developed.

Egress models will employ different methods to produce a performance assessment [22]. Each model type also represents the key components (e.g., the structure, the population, behavior, procedural activities, environmental conditions, etc.) in some form using different techniques and to a different degree of refinement (see Chap. 60). More refined models represent individual agents and their movement within a detailed representation of the structural space; i.e. where the internal structure, obstacles, etc., are represented. In this space, the simulated agents may be subject to environment, behavioral and procedural influences present, i.e. they attempt to represent some of the influences and processes described previously in a simplified manner. Conversely, other models employ a less refined representation with the population flowing between more crudely defined architectural spaces. Other models fall somewhere in between these two extremes [22, 27]. The variety of this representation (and therefore the relationship with the data available) is further complicated through the approach employed to represent evacuee behavior, i.e. whether it focuses on movement alone, has a behavioral model, or somewhere in between (see also Chaps. 59 and 60). In addition to each model needing data in their development, each has their own data needs in their application; however, they will all need data in order to be applied at all.

Depending on the sophistication of the model, they could be employed in a number of different application modes, each of which places a different onus on the user and on the data required. For instance, the level of configuration required of the user will be dependent on the sophistication of the model, the predictive capabilities of the model and the application. This will influence the scope and detail required of the data-sets sought out. This is described in more detail in Chap. 60.

Human Behavior Data

The evolution of the field has led to an incomplete, disorganized and disparate understanding of the subject of human behavior in fire. This has been due in part to the diverse background of those contributing to the field (e.g., engineers, social scientists, field researchers and model developers) and the changing opinions in the field as to what constitutes necessary information.

Human behavior in fire is not adequately supported by the data available [9] and this should be understood before interrogating the data-sets presented. This suggests that there may well not be an ideal data ‘fit’ for the scenario at hand, i.e. compromises are often required. The data-sets that are available are therefore often not sufficient, i.e. not sufficiently comprehensive or detailed for the range of intended applications. This limitation should influence model development, selection and application. This constraint underlies the selection of the approach adopted during this chapter which is to provide a synopsis of a wide variety of data-sets rather than a detailed narrative of a limited number of seminal research or well publicized incidents.

Given the increased use of these egress calculation techniques (through greater adoption of performance based design to accommodate innovation, ethical issues, financial issues), it is essential that the underlying methods employed are appropriate to their respective, often wide ranging applications. These methods require the provision of accessible, detailed and unambiguous data in order for them to be appropriately configured and validated. Empirical data-sets addressing human performance are difficult to locate, not sufficiently detailed, dispersed, and are often employed without sufficient understanding of the context in which they were collected. The application of this data could be improved with better knowledge of:

• The conditions of the event from which the data originated,

• The methods used before and after the event to collect and analyze the data, and

• A detailed representation of the data itself.

Without this context data can be misunderstood and misapplied. The intention here is to provide sufficient information for the user to identify the underlying conditions influencing the data and/or identify where omissions exist regarding our understanding of these conditions. The importance of context is discussed in the following sections.

Data Collection: Context

Data can be obtained using a variety of research methods and data collection techniques. It is important to understand how the data was obtained, since the choice of research method and data collection techniques can influence the validity (internal and external) and reliability of the results. Validity refers to the correctness of the study findings; in other words, the extent to which it measures what it is supposed to measure. Internal validity is the extent to which cause and effect relationships can be accurately identified within the study, whereas external validity refers to whether the findings of the study are generalizable to a real-life setting. It is often useful to consider the external validity of the research relative to the objective of the study. For example, if a study aims to determine the effectiveness of a particular type of alarm system on pre-evacuation time, then the external validity of an announced evacuation would be lower than that of an unannounced evacuation since it is expected that in the latter case participants’ behavior would be more similar to a real emergency than in the former. This may influence the value which is placed on the data that is subsequently produced. Reliability, on the other hand, refers to the repeatability of a study; i.e. whether the study is properly documented such that the conditions can in fact be characterized and described in detail and that the data collection techniques used are sufficiently precise. Although not explicitly considered here in relation to the data sets presented, the validity of any data-set should be established by the reader when examining the appropriateness of the data-sets for use. It is therefore important when considering any data set to determine how representative, detailed, comprehensive, consistent and robust the data is and whether it is appropriate for this data to be used in the application at hand.

Broadly speaking, research methods can be categorized as follows [9].

• Experimental trials:

  • Hypothetical scenarios (e.g. response to video playback, simulated environment, tabletop exercises, large-scale pre-planned field exercises, etc.). Typically, those taking part are aware that they involved in the event given the artificial nature of the surroundings and the obvious preparations required to organize and conduct the scenario. For instance, getting participants to control an avatar within a simulated environment, or asking commanders to respond to simulated feedback within a command room.

  • Controlled experiments (e.g. laboratory trials). Again, typically participants would be aware of the nature of the trials being conducted. These might focus on, for example, a particular act (time to ascend stairs, open a door), a particular influence of a factor on performance (the impact of smoke on decision-making, etc.), or behavior in a specific situation (e.g. being alone, in a group, etc.), The user of data produced by small-scale component tests should be aware of the primary performance factor being examined and how this might be integrated into the analysis being performed. This will allow them to establish the credibility of the claims made—especially regarding other secondary or peripheral factors derived from the data-set.

  • Field investigations (drills, etc.). These may be announced or unannounced. Depending on the nature of the experiment, the population involved may have some forewarning of the event, may not be exposed to deteriorating environmental conditions, and/or may become aware that the event is not real. All of these may influence the external validity of the results produced.

• Case Studies / Formal incident investigations. These are performed in order to understand the events during an actual incident and what factors contributed to it [9]. Such investigations tend to focus on establishing the chain of events and factors that contributed to the outcome, rather than producing quantitative estimates of times or capabilities relative to the time line. The user should therefore be cautious if presented with such from this type of study.

It should be noted that not all of these research methods are represented in the various tables that present the data-sets examined here. However, this does not preclude the reader from encountering them in data-sets not presented here or in the presentation of future data-sets.

In reality, the individual is part of a temporal-spatial environment formed within the structure in which the incident has occurred, the history of the structure’s use, and the procedural framework associated with the incident—the physical, historical and procedural environment [22]. This is briefly described below in order to further elaborate on the importance of context when understanding the data available, its limitations, its applicability to the target scenario and then how it might eventually be employed.

A building can be seen as a people movement system that operates in three phases: ingress (people enter the building), circulation (people use the building) and egress (people leave the building). Therefore, for the building to function people have to arrive and enter the building; during its use, people circulate around the building and, eventually, people leave the building (see Fig. 64.3). Data may relate to any of these phases—it may have been collected when people entered, used or left a building under different situations. This is not reflected in the engineering timeline which (deliberately) partitions the events into distinct separate sections. This is obviously a simplification. In this chapter the description of the data-sets makes reference to these phases where appropriate and where known.

Fig. 64.3

ICE—the three phases of people movement [27]

The acronym ICE (ingress, circulation and egress) is used to describe the system of people movement [27–29]. The phases shown in Fig. 64.3 are highly coupled—they co-exist and interact; for instance, people may enter a building while others use and leave the building. Furthermore, the way that a building is entered may influence the way that it is left, which is important during any egress analysis. These phases may also occur at any point during the engineering timeline, or vice versa. This has implications for the understanding and use of movement data; i.e., the phase of movement during which a data-set was produced may not necessarily be the same phase to which it is required to be applied. For instance, escalator data is typically collected during circulation movement, but might conceivably be applied in future analyses where escalators are employed as a component used in an emergency. This does not necessarily preclude its use or prevent there being a significant overlap between the underlying conditions and the expected behavior; however, it is important for the data user to be aware of this in order for them to use the data in an informed manner.

In this chapter data-sets representing general circulation, emergency egress and non-emergency egress are presented. In the description of each data-set, the nature of the event being represented is made clear using abbreviated terms (real incident [I], circulation movement [C], an announced evacuation drill [AE], an unannounced evacuation drill [UE] or an experimental trial [ET]), reflecting the variety of emergency/non-emergency circulation and egress being represented.

Data Collection: Techniques

Data collection techniques are the “measuring instruments” of the research study; i.e., these can be employed during the data collection activities of the various research methods highlighted previously. These instruments can be used individually, but often studies combine them to improve the quality of the research. Triangulation of data using different instruments can be used to corroborate findings and lead to improved validity and reliability of results i.e. data. Data collection techniques commonly used in human behavior in fire research can be broadly categorized as surveys (including interviews and questionnaires), observations, reviews and simulations.

• Surveys—getting information from a sample. Data collected through questionnaires or interviews designed to determine the characteristics, actions, opinions etc., of a particular sample. The user should be aware of the nature of the survey (e.g., whether it was open or closed), the content (e.g., what was actually being asked), the mode (telephone, online, face-face), the sampling approach adopted, and the size and nature of the sample obtained

• Observations—getting information on the sample. Movement/actions are directly observed in some form, without necessarily relying on the verbal communication with those involved. The technique employed to collect the data will influence the format in which the data is collected, the focus of data collection resources, the nature of the data collected, whether the data relates to a time, a space, a person, a condition, etc., and the overall precision and accuracy of the data. There are a number of methods available to collect data:

  • Stationary video camera—providing a fixed stream of information for a particular location/situation, using existing CCTV/security cameras or specially located cameras. This data source has the potential to record behavior in both emergency and non-emergency situations; therefore the user of the data should be aware that there may be discrepancies between the scenario recorded and the scenario of interest,

  • Roving video camera—allows for capturing of progressive conditions experienced by selected individuals or groups,

  • Still photograph—allows a snapshot of a situation to be established at a point(s) in time,

  • Human observer—allows individuals to make manual numerical and/or descriptive observations,

  • Human participant—first person accounts of researcher participation in a situation (either contrived or accidental).

  • Electronic Sensor/Automated Measurement/RFID technology—devices deliberately inserted (often requiring the participants to carry a device) into a location to monitor movement/behavior.

  • Scanner—non-visual methods (e.g. laser techniques) to establish the movement of populations. Often this does not involve identifying those involved.

• Simulated data—generating information on the sample. The use of computational tools to explore egress performance under conditions that could not be explored directly, such as catastrophic fire conditions. The strength of this type of data is highly dependent upon the nature and sophistication of the model used. It may be one of the few options available to investigate performance under extreme conditions; however, the limitations of the model, the process and the data produced should be clearly understood. Simulated or compiled data (e.g. engineering curves) are often confused with direct observations and so great care should be used to identify the source appropriately.

Although not based on new data, the review of existing material/secondary resources such as academic literature, journalistic sources, anecdotal evidence, and material from adjacent fields of research [20, 30–32] i.e. compiling information on/from previous samples, can also provide additional evidence. The value of the source will be dependent on the appropriateness and credibility of the secondary source given the intended application. The ability to assess this appropriateness will largely be influenced by the background information provided by the original authors.

Third-party data users need to be aware of the data collection methods employed to better assess the underlying research, the validity and reliability of the data, and hence the suitability of this data for their needs.

Data Collection: Process

Data does not exist independently of the collection process, i.e. data are not collected in a vacuum. The data collection process requires decisions to be made at a number of stages, and these directly influence the scope and refinement of the data, and the applicability of this data. Hence, it is important to understand the process by which data are produced in order to apply it appropriately and responsibly.

Initially, a decision has to be made by someone to acquire or seek out data—an omission or weakness in the data available has to be noted and a need established. The data acquisition process is therefore selective. The data collection methods selected may be based on their appropriateness, but also based on less rational reasons, e.g., available expertise, cost, convenience and these choices directly influence the credibility and applicability of the data.

Once the research is conducted, the data are extracted and analyzed and then the data and the derived understanding are described and presented, i.e., the data is distilled into a representative state from a raw form and summarized along with the background information that is available (according to the information collected) and deemed worthy of inclusion by the researchers. When researchers produce a document and present the data, they decide what is relevant and specifically what is relevant to the project at hand. This means that sometimes insufficient information is presented on the data collection methods, context or the results to enable selection to be made in an informed manner by third parties.

The data-sets are then shared with an audience of interested third parties (in academic journals, conferences or readers of this handbook). Having had access to the data, these third parties (engineers, modelers, other researchers) attempt to understand the data according to the presented format and the associated background information and, on this basis, the data are then applied. However, it is important to note that not only is data acquisition selective, but data use is also selective and not necessarily based on the appropriateness of the data itself, i.e. third party users of data may not necessarily source or utilize the most appropriate data for their application, but instead make judgments based on those sources with which they are most familiar or to which they have ready access. This choice is compounded and made difficult by the limited background information associated with the data and the data being provided in a summarized form. In such circumstances the likelihood of data being inappropriately employed for a particular application is increased.

There are a number of opportunities within this process for the data to be misrepresented, misunderstood and misapplied. In most instances, only a sub-set of the data collected is shared, i.e., it is shared in a reduced/distilled format rather than in a complete format. Potentially, and more importantly, in the vast majority of cases only a limited amount of information is provided on the background conditions evident during the original event, i.e., the scenario that produced the data. The reduced data-set and limited context then requires a greater degree of interpretation by the third party. This increases the potential for the underlying causal factors being misunderstood, the results being misinterpreted, and the data-set being inappropriately applied. Although it may be impossible for the third party viewer to understand the content of the information omitted, it may be possible for them to ascertain the type of information omitted and then draw their own conclusions—determining whether or not this omission is critical.

When viewing past data collection, it is important to understand the assumptions made by the data collectors. The collector’s ‘theoretical’ assumptions would have influenced the methods employed, the data collection techniques used and the data actually collected. The format adopted in this chapter represents an attempt to intervene in this process by providing the reader with sufficient background information for them to establish the appropriateness of a data-set (and the data included and excluded), and also to encourage data users to become aware of the importance of understanding the context in which any data set was developed whenever data-sets are to be employed in the future.

Data Selection and Representation

The intention is that the data-sets described in the following sections are presented in sufficient detail such that they might reasonably be employed as part of the engineering process where deemed applicable. Where there is sufficient data to support it, a table is produced. Where there is insufficient data, the data-sets are described in the text, with a detailed description of each data-set provided.

The following paragraphs describe the sources of the data and the structure of the data presentation. An attempt has been made to ensure consistency between the table designs and content to allow comparison and to improve accessibility.

Sources of Data

Data has been sought from those sources typically considered as credible outlets within the field. These outlets include:

• Journal publications: Journal of Fire Protection Engineering, Fire Safety Journal, Fire Technology, Fire and Materials, Safety Science, International Journal of Performance-Based Fire Codes, Journal of Applied Fire Sciences, Building and Environment, Journal of Transportation Engineering Transportation Research Record, Physica A, and

• Conference proceedings: International Association Fire Safety Science (IAFSS), Interflam, Pedestrian and Evacuation Dynamics (PED), Human Behavior in Fire, Asia-Oceania Association for Fire and Technology, Mobility and Transport for Elderly and Disabled People.

In addition to these publications, other reports were included, where deemed appropriate; for instance, reports of National Bureau of Standards (NBS), National Institute of Science and Technology (NIST), National Fire Protection Association (NFPA), National Research Council Canada (NRCC), Fire Protection Research Foundation (FPRF), Lund Department of Fire Safety Engineering, Building Research Establishment (BRE, UK), VTT (Finland), Fire Research Institute of Japan.. Occasionally, PhD dissertations which are readily accessible (on-line) are also included, although these were identified and used on a more selective basis.

These sources were identified according to the following criteria:

• Publically available—so that the sources can be followed up by interested readers,

• Written in English (or where translations were available on request)—so that the resources can be understood by those reading this handbook,

• Published after 1985—so that the data sources are fairly contemporary. Note that the authors have taken a decision, in a very select number of cases, to include data pre-1985. The decision to include this data was on the basis that the data represented seminal work with respect to that behavioral component, a data set which is commonly used and cited, or where no other more contemporary data was identified.

For each of the core behavioral components, a table is presented which provides not just the data but background information that allows the reader to understand the context in which the data was collected, the scenarios associated with the data-sets and the data collection methods employed. Thus, the user is armed with information which will facilitate an assessment of the potential validity, utility and applicability of the data and enable an informed selection of the data which is most appropriate to his/her application.

Currently, there is some debate regarding the applicability of egress-related data collected during the 1950s–1970s. This includes the work of Pauls [33], Fruin [34], amongst others. In many instances, this data still forms the basis for much of the current egress analysis (using engineering calculations and/or egress models). A number of these authors have requested that their work not be included in current handbook chapters as this may be seen as suggesting that the data should still be used in engineering calculations. This data is therefore not presented in detail here. However, if the reader is interested in this data, detailed reference to it can be found elsewhere [33–35].

Structure of Data Presentation

The data-sets are presented according to each of the key behavioral components that comprise the engineering timeline described previously, i.e., detection and warning phases, pre-evacuation phase and travel phase:

• Detection and Warning Phase: This focuses on the delays in the time for the alarm to be raised in the absence (or override) of automatic detection and alarm system; i.e., when detection and general alarm is reliant on the actions of staff only that can lead to significant delays between detection and notification.

• Pre-Evacuation Phase: This focuses on the pre-evacuation times of occupants who are awake in a range of buildings including Transport, Assembly, Health Care, Educational, Industrial, Mercantile, Residential, and Business premises. It also presents data on important components of pre-evacuation time for those who are asleep (i.e., the time to awaken and the probability of awakening), as well as considering the influence of impairment on time to prepare to evacuate which can be considered another component of pre-evacuation time in some circumstances. The data in this section is therefore sub-divided into the following three categories:

  • Pre-Evacuation—Awake

  • Pre-Evacuation—Asleep,

  • Pre-Evacuation—Impaired

• Travel Phase: This focuses on data related to the movement characteristics (i.e. speed and flow of people). These are provided for each of the most common escape route components namely, horizontal, stairs, exits and escalators. Unimpeded movement speeds of the elderly and others with mobility impairments on the horizontal, ramps and stairs are also presented as is data on both upright and crawling speeds in smoke. The data in this section is therefore sub-divided into the following six categories:

  • Horizontal Movement (flow/speed/unassisted/assisted impaired),

  • Stair Movement (flow/speed/impaired/elderly),

  • Exit (flow/traversal speeds),

  • Escalators (flow, speed),

  • Ramps (for unassisted/assisted mobility impaired)

  • Situational Vulnerabilities (Smoke-upright, Smoke-crawling)

Data-sets are presented for each of the behavioral elements above in tabular form. A simple representation of the engineering timeline is provided at the beginning of each behavioral element to signify the phases to which it relates (see Table 64.2).

Table 64.2 Phase from the engineering timeline within which the behavioral components fall

The tables are presented in as standardized a format as possible—certainly within each of the components presented. The tabular format includes a description of the background conditions under which the data was collected, when and how the data was collected and the key results derived. Other information important to the understanding and interpretation of the key results are also presented. This then places the data in context, allowing the reader to better discriminate between the data-sets and select between them. The importance of many of the attributes of the data collection process and event scenario is apparent—primarily as they apply directly to the selection and application of the data. For instance, the building in which the original data collection took place may be a key criterion in the reader’s selection process. Similarly, the nature of the population and the procedure employed. However, the reader should also be cognizant of the nature of the original incident (i.e. the source of the material) and the manner in which the data were collected. Although, this may only be considered an indirect impact on the relevance of the data it directly affects the external validity of the data collected and the format/content of this data. Understandably, some of the behavioral components have different attributes—given the nature of the component being represented and the detail in which the data-sets are presented. However, in all instances, the following attributes are described (although not necessarily in exactly the same format depending on the different data represented)

• Source (original reference)

• Observational Conditions represented by:

  • L: Location: the country or city in which the event took place (if known)

  • N: Nature—this can relate to the component being examined in the event (real incident [I], circulation movement [C], an announced evacuation drill [AE], an unannounced evacuation drill [UE] or an experimental trial [ET]) and Location (if known)

  • SC: Spatial Configuration

  • P: Participants or Sample (with A: Age, G: Gender and I: Impairment)

  • E: Environmental conditions are presented where the information is available; otherwise, the ‘E:’ is removed for brevity

  • V: Variable—specifically, the variable(s) that influence the data collected (i.e., the independent variables)

• Sample/Data Collection Considerations/Method—data collection techniques are identified as being an [I]nterview, [S]urvey, [V]ideo, [E]xisting or [O]bserver based.

• Results—unless otherwise stated, these are presented in the following format: average [standard deviation, minimum—maximum] and are aligned with the most influential factors presented in the other columns where appropriate. Omissions are denoted with a hyphen (-).

• Additional Information

In order to reduce the space taken up by these tables, several abbreviations are used to ensure that the maximum amount of information is provided. These are outlined next to each of the tables where they are used in order to improve the reading of the data presented along with a description of additional factors described in each type of table (e.g., configuration, performance attributes, etc.). Additionally, in many of the cases, background information is absent, due to limitations in the original description of the data collection process or the event scenario. Where information is missing in the original source material this absence is noted (typically through the presence of a dash ‘–’).

Many of the reference sources provide multiple data-sets that reflect often subtly different conditions. In order to distinguish clearly between these data sets an ‘Observational Conditions’ column in each table provides background information on the data and, in particular, the Variable (V) entry records factors that distinguish the data sets presented. These conditions are wide ranging and may represent: differences in the samples used (e.g. different age groups, abilities), independent variables that the researchers set out to investigate (e.g. type of alarm), the nature of the event (e.g. announced/unannounced evacuation), or spatial configuration of the experimental set up (e.g. up/down direction on stairs/escalator). In each case the variables are noted and abbreviations (derived from those outlined earlier in this section) given. These abbreviations are then used elsewhere in the respective tables to distinguish the values of each variable and to clearly identify (in the results column) which results relate to which combination of variables.

Data-Sets

The data-sets are presented in the following sections. These cover three phases of the engineering timeline: Detection and Warning, Pre-Evacuation and Travel.

In each section, an overview is provided regarding the structure of the associated tables, along with a brief discussion of previous attempts to compile equivalent types of data together, where appropriate. These brief discussions are only intended to indicate that previous reviews exist and indicate the format that some of these reviews have taken. It should not be assumed that these reviews are preferred over the many others that may have been performed.

Detection and Warning Phases: Human Aspect

As noted previously, in an engineering design, both human and/or technological resources may be employed to detect an incident and raise the general alarm to initiate the evacuation procedure; i.e., any delay in detection and warning (also previously referred to as Pre-Warning time [11, 12, 24, 25], may be procedural and/or cognitive. Where automatic detection and alarm is employed, typically, t _det and t _warn will be determined with reference to manufacturer’s guidance on sensor response and often do not (and may not need to) take staff activities into account.

However, it is certainly possible that staff (or other occupants) will be involved in the detection of the original incident and the raising of the alarm and it is therefore important to recognize that this involvement might delay the point at which the pre-evacuation phase begins for the general population.

The human element in establishing warning times therefore arises in any situation in which the behavior of an individual intervenes between the detection of a fire and the raising of the general alarm [11, 12, 24, 25]. The need to determine the times required for persons aware of the incident to initiate warnings to affected occupants is recognized in fire safety engineering standards [11, 12, 24, 25]. However, limited guidance is available on the behavioral parameters involved, how the evacuee behavior should be managed and how they can be quantified in a design context.

In reality, the delays associated with these activities may be difficult for the engineer to estimate precisely. However, the engineers may choose to either insert estimates based from previous incidents to provide some representation of this phenomena where deemed relevant or compile an estimate based on expected procedural activities (discussed in detail elsewhere [11, 12, 24, 25]. Examples of such delays from real incidents (both fire and non-fire emergencies) are shown in Table 64.3.

Table 64.3 Estimated staff-related delays from historical incidents [22, 24, 25]

Staff decision-making may occur in a number of situations, e.g., in discovering a fire, interpreting cues first hand (and recognizing them as indicating a real incident), or being notified of an incident (i.e., via the notification system at the fire panel, by other members of staff, by members of the occupant population, etc.). In these instances, staff members will need to perceive, interpret, determine and then perform the action—in much the same way as other occupants, all of which takes time. Although staff may be better trained and have more experience (both of which will influence the extent of any delay), they would still have to go through this process, delaying their response and, in turn, the response of the general population of the building. The understanding and representation of this component is still relatively immature and, as such, there are relatively few data-sets available to support the representation of this phenomenon within the engineering process.

Pre-evacuation Phase

The time that occupants take to initiate their evacuation movement can be difficult to estimate. In the past, this delay was often not included in an engineering analysis at all [36]. Although, this certainly made the calculation easier (i.e. fewer components needed to be assessed), it also potentially underestimated the expected evacuation times in many of the scenarios examined.

In the calculation of an expected total time to evacuate a building, it is now common practice for engineers to include some time to account for a delay to the start of evacuation, the t_pre. This may be achieved simplistically, by segregating the pre-evacuation phase from the travel phase, or in a more coupled manner where the nature of the pre-evacuation phase can influence the development of the travel phase. The tables in the following sections should help an engineer identify such pre-evacuation times.

There have been a number of previous efforts to explore the pre-evacuation phase and produce a collection of the data-sets available. Key advances in compiling pre-evacuation data-sets were made by Fahy and Proulx [37], who collected together a range of data-sets while presenting contextual information to aid the reader’s assessment (see Table 64.4).

Table 64.4 Delay times (min) derived from actual fires and evacuation exercises reported in the referenced literature [37]

Other pre-evacuation data (referred to as pre-movement) has been compiled by Shi et al. in a similar format, although in this case data was categorized primarily according to the occupancy type [38]. Bruck and colleagues [39–41], Purser and Bensilum [10] and Proulx and McQueen [42] also provide detailed summaries of their own research regarding pre-evacuation responses. Gwynne also examined a range of pre-evacuation data-sets in order to identify how vulnerabilities in the population might be addressed using notification systems and to identify where gaps might exist in our understanding [43, 44].

Peacock et al. performed a detailed analysis of egress trials conducted by NIST [45, 46]. Although, this focused upon stair travel, a number of pre-evacuation data-sets, based on data collected from buildings ranging from 6 to 31 floors in height, were also summarized. These evacuations produced average pre-evacuation times of between 89 and 224 s [45, 46].

Other pre-evacuation times are presented by the type of occupancy involved in the original event and are described in Tables 64.5, 64.6, 64.7, 64.8, 64.9, 64.10, 64.11, and 64.12. In these tables, pre-evacuation times for business, residential, mercantile, industrial, educational, health care, assembly and transport occupancies are provided. Broadly speaking, the same information is provided throughout Tables 64.5, 64.6, 64.7, 64.8, 64.9, 64.10, 64.11, and 64.12: Source Observational Conditions (Location (country where known), Nature, Spatial Configuration, Participants, Environment and Variables), Procedure (Strategy, Staff, Technology), Sample (Collection method and Size), Results (Mean, Standard Deviation, Range in Seconds). Information is only included in the tables if it appeared in the original source or if can readily be derived. If the information is derived by the authors, this is noted in the table. The Additional Information column is used to provide the reader with more detail on other data or analysis provided in the original source.

Table 64.5 Pre-evacuation data—business occupancy

Table 64.6 Pre-evacuation time—residential occupancy

Table 64.7 Pre-evacuation time—mercantile occupancy

Table 64.8 Pre-evacuation time—industrial occupancy

Table 64.9 Pre-evacuation time—educational occupancy

Table 64.10 Pre-evacuation time—health care occupancy

Table 64.11 Pre-evacuation time—assembly occupancy

Table 64.12 Pre-evacuation time—transport

In this examination, pre-evacuation time data have been collected in several ways, primarily from evacuation exercises and experimental/laboratory trials. This data, at least in part, reflects the decision-making process of those involved. Therefore, the credibility (or at least the applicability) of the data collected is dependent upon the information available to the target population during the event since that impacts the external validity of the research. Of critical importance here is whether the participants were aware that it was a drill/experiment rather than a real incident. Prior knowledge may well have influenced the nature of their response, the participants’ ability to prepare for their response and the time they took to respond. It should be noted that in a real incident, the information available to those involved would differ significantly according to the cues available to them and differs from a situation where people definitively know that the incident is not real. A positive understanding of the negative (unambiguous warning of the drill before it commences), is not equivalent to the negative understanding of the positive (evacuee assessment of a smoke cue that may or may not indicate a threat to the individual).

It is important to note that the different data-sets represent markedly different pre-evacuation time distributions. A number of characteristics that could have impacted on the pre-evacuation time have been identified and discussed previously [7, 36]. It is presently understood that the pre-evacuation time will vary according to situational, structural, procedural, organizational, behavioral and environmental factors present. Kuligowski provides a detailed description of the factors that influence the pre-evacuation phase [20], some of which are also discussed in Chap. 58 of this handbook. A brief overview of the factors derived from the most recent work on human behavior is presented below.

It should be noted that not all of the factors discussed in the following paragraphs are represented in the data-sets described, often because this information was not available in the original source. It may not therefore be feasible for engineers to account for these factors within the data selected. However, even if it is not possible or feasible for the engineer to account for these factors, it is still important for them to be aware of their impact during an actual incident. Where the factors can be identified within the data-sets, it is noted in the description below.

Building type, layout and complexity. The building type and layout provides the spatial environment within which the event occurs. It may determine the nature of the occupants, the resources employed, the hazards present, and the social/organizational hierarchy present and may influence the types of actions that might be expected within a particular space. As such, it is a primary influence upon the scenario produced in physical, sociological and psychological terms. The way each floor and the whole building are organized has an impact on the familiarity and use of the space by an occupant both before and during an incident; e.g. through visual access to routes and exit points. Occupants are more likely to spend time obtaining information or devising a plan of action in a complex building or in a building where wayfinding is difficult. The way the building is designed may or may not also provide occupants with visual access to the behavior of others, to the original incident or to procedural attempts at notifying them as the target population. If available, the building type and the population are described in the Observational Conditions column (under Spatial Configuration and Participants). The data-sets are also categorized according to the type of occupancy within which the event occurred. This factor was selected as it is likely to be the first factor that the engineer encounters and is likely to form the base assessment of the scenario represented. Where available, the number of floors involved in the structure is also given.

Procedure—Notification System—the technology employed as part of the procedure to inform the target population that an incident has occurred. This may include a range of different technologies, each of which carries different types of information, suggests different degrees of urgency and which have different degrees of comprehension and intelligibility. This would typically need to be represented as part of the engineering design process. If available, the technological resources that are part of the emergency procedure are described in the Procedure column (under Technology). Typically these are systems that alert that something has happened using sirens, bells, horns [AL]; using a T-3 fire alarm system [T3]; provide information beyond simply an alert using a live voice notification [LV], a pre-recorded voice notification [PV], or a voice notification where the nature of the voice was unclear [-V]).

Procedure—Human Resources—Staff form a key component within the procedural response. It is widely recognized that the presence of well-trained, engaged, authoritative and informed staff presents the most effective means to initiate occupant response [28, 63, 82, 83]. If available the human resources that are part of the emergency procedure are described in the Procedure column (see section “Pre-evacuation Phase”) under Staff. It should be noted that only those staff for whom specific mention of their particular role was made in the original source are included.

Alertness and limitation. Occupants may have situational or innate characteristics that reduce their alertness. Occupants may be asleep, intoxicated, or impaired all of which might reduce the information available to them. The impact of alertness is addressed in section “Pre-evacuation Phase: Asleep” (i.e. that people may be sleeping when cues are provided) while the impact of impairment is addressed in section “Pre-evacuation Phase: Impaired”.

Focus—Commitment and background noise. In situations where occupants focus their attention on a particular point, e.g., at a cinema, attention may be diverted from critical environmental and procedural cues. Similarly, in environments where aural or visual background noise is present, the cues available to the population may be confounded and confused, delaying their response. Occupants may also be committed to their actions, potentially having committed resources to the performance of this action, making them reluctant to interrupt it on the basis of ambiguous cues. The impact of commitment is not addressed per se in Tables 64.5, 64.6, 64.7, 64.8, 64.9, 64.10, 64.11, and 64.12, although the building type may give some indication of the range of activities undertaken by the samples involved. The presence of background noise is noted where known in section “Pre-evacuation Phase: Asleep” in relation to the response of sleeping persons.

Training—Training is a characteristic of the organizational structure within a building, since training should be specifically tailored to each building evacuation procedure present. That is not to say that the populations do not bring more general experience to an incident; however, this is more difficult to predict and quantify within the engineering process. The likelihood and nature of occupant training will depend on the occupancy type; for instance, in public buildings, occupants are unlikely to be trained for that specific building, whereas some form of training is likely to be the norm for office spaces (although the sophistication of this training may vary significantly).

False alarms. The number of false alarms in a building is an important determinant of the efficiency of this system to warn occupants. If the number of false alarms is high, the pre-evacuation time will likely be extended since occupants are unlikely to look for information and will be less receptive to other cues.

Familiarity. Occupants who are familiar with a building, who have participated in evacuation drills, and who are aware of the evacuation procedure are more likely to start evacuation rapidly. What is not well understood is the point at which the performance of drills and training exercises starts to make the occupant population skeptical of the information being provided.

Social Affiliation. The nature of the relationship between the occupant and the surrounding population will influence the manner in which information is perceived and the actions subsequently performed. It will influence the responsibility felt by an occupant for those around them, information exchange, perceived risk and the preparatory actions that might be performed before movement to safety is initiated. Social affiliation is not presented in the tables; however, the building type and sample description may give some indication of the types of relationships present among the sample.

Event Conditions and Proximity to Event. The nature and severity of the incident will influence the type of cues provided to the population. Their proximity to the incident will influence access to these cues, the degree of ambiguity and the perceived sense of risk derived from the cues received.

Surrounding Population and their Actions. The population can be a source of information, with their actions indicating their interpretation of the incident and the options available. However, research has also suggested [84] delayed responses in the presence of others, given their identity and actions. These influences have been referred to as informational and normative social influence [85]. They may also limit viable responses, should routes become overloaded or congestion develops, discouraging the use of certain routes.

The core data-sets relating to the pre-evacuation phase are presented in Tables 64.5, 64.6, 64.7, 64.8, 64.9, 64.10, 64.11, and 64.12. These relate to a range of occupancies and scenarios—cutting across the various factors highlighted above. Further data-sets, exploring the status of the population (i.e. whether they are asleep) and the impairments present in the population (e.g. whether additional preparatory actions are required by the occupants before they initiate egress movement) are presented in more detail in sections “Pre-evacuation Phase: Asleep” and “Pre-evacuation Phase: Impaired” below.

Pre-evacuation Phase: Asleep

As discussed, an extended pre-evacuation time is often the most significant component of the evacuation process. As shown in Table 64.4, these values can range from seconds to hours. Indeed, it is possible that when occupants are asleep the pre-evacuation times could be especially prolonged. Comprehensive reviews of the research literature on arousal from sleep have previously been conducted by Bruck and colleagues [39, 86, 87]. As noted by Bruck [39] this literature falls into two categories: the first comprises investigations into the characteristics of sleep and in particular the arousal thresholds of persons related to the various stages of sleep across the night to different sound stimuli; the second comprises of more focused investigations by those interested in the response of individuals in fire situations who have investigated the likelihood of wakening to a range of specific fire cues, including smoke detector alarms. Since the context of this chapter is in relation to human behavior data for use in fire safety design, the focus here is on the latter. The main findings are summarized in Table 64.13. Table 64.13 includes the summaries already presented in previous literature reviews [39, 86], but has also been updated to include studies conducted post 2005.

Table 64.13 Pre-evacuation sleep data

Table 64.13 presents the results of a series of experimental trials [ET] conducted in controlled environments (in household, laboratory or hospital settings). Table 64.13 contains information on the Source of the data, the Observational Conditions under which the study was conducted (Location (country in which conducted), Nature of the study (since these were all experimental trials [ET], this section provides details of the experimental set up including the nature of the alarm presentation, stage of sleep, and the measure used to determine awakening), Spatial Configuration (own, home, laboratory), Variables, the Experimental Sample (Nature and Size), the Signal Employed (Type, Frequency and Intensity), Results (Mean Auditory Threshold, percentage of Awakening and Time to Awaken as appropriate). The Additional Information column provides other information that may have influenced the outcome [39] regarding the study, e.g., the presence or absence of background noise (if known) and whether subjects were primed to expect a signal or were naïve in this respect.

The main method adopted in the included studies is that of a constant stimulus being presented to the participants. In this case, the frequency of responses is simply measured within the sample to determine the percentage of participants that awoke and the time it took them to awaken. Another common method which is used to determine, not just the response to a particular signal of certain frequency/volume but the waking thresholds (or AAT—Auditory Arousal Thresholds) for the participants, is the method of limits [39]. In this approach, a tone of a standard frequency is presented to a sleeping person at a specific intensity and the intensity increased (usually in five dBA intervals) if there is no response within a certain time period. A response or “no response” to each presentation enables the auditory arousal threshold (AAT) for the person to be obtained. This results in a mean AAT rather than a percentage awakening output (since all eventually awake).

Table 64.13 only provides a summary of some of the most significant data in this area. For a more in-depth treatment of this complex subject area the reader is directed to the comprehensive reviews of Bruck and colleagues [39, 86, 87].

Pre-evacuation Phase: Impaired

Research has shown that during the pre-evacuation phase, building occupants spend time in various activities prior to evacuation. In public spaces, these activities commonly include gathering belongings, turning off computers, packing belongings away, etc. In some evacuation scenarios, particularly in dwellings or other occupancies with sleeping accommodation, these ‘preparation activities’ may also include turning on bedside lights and getting dressed. This ‘preparation time’ can be considered a sub-component of the ‘pre-evacuation’ time and it is recognized that this time could be considerably longer for some individuals, e.g., people with disabilities.

Two studies have directly compared the ‘preparation time’ for disabled and able-bodied persons. An early study by Pearson and Joost [103] was designed specifically to determine the preparation times of subjects exposed to a simulated fire emergency in a residential setting. The first study involved 18 healthy able-bodied male undergraduate students, 11 blind persons (4 female and 7 male aged from 21 to 58) and 9 male wheelchair users. The second study involved 20 healthy young adults (3 male and 17 female aged from 30 to 48), 20 elderly persons in good health (4 male and 16 female aged 59–79) and 20 elderly with arthritic impairments (3 male and 17 female aged 55–80). Each subject was required to complete a range of scenarios that involved the completion of various activities such as searching for a personal item, donning clothes, retrieving personal effects, and activating lights. Pearson and Joost’s [103] study concluded that the mean time for blind subjects to complete these activities was 2.47 times greater than that of young, able-bodied subjects and the mean time for wheelchair users was 2.36 times greater than that of able-bodied subjects.

A later study by Shields et al. [104] also compared the time to prepare to evacuate as part of three evacuation studies of hotel accommodation involving mixed ability populations, including four wheelchair users. Two of these studies were designed to simulate day-time evacuation of hotel accommodation in which participants were initially located in a chair or in their wheelchair and were asked to retrieve a personal item from a drawer before leaving the room. In the third study, designed to simulate a night-time scenario, participants were initially lying on the bed, and were required to turn on a bedside light, get into their wheelchair (if applicable) and retrieve a personal belonging from a drawer before leaving the room. The results from the Shields’ studies suggest that wheelchair users took 1.6 and 1.9 times longer than able-bodied persons to perform the tasks which comprised the day time scenario and 2.4 times longer than able-bodied to perform the tasks which comprised the night time scenario.

These findings are interesting and suggest that, in situations where those with impairments are present, the expected pre-evacuation times might then be increased given the potential difficulties in responding and the need for preparation. However, it is important to note that in both studies the samples were small and the range and severity of disability rather limited. Therefore they should be treated as indicative only and used with care. For a more detailed understanding of the factors described and the base values collected in each case, the reader is referred to the original source material.

Travel Phase

Egress models (engineering or computational) require a number of basic quantitative and qualitative information in order to configure the initial scenario to be examined, i.e. to reflect the core behavioral elements of the engineering timeline. As mentioned these include the pre-evacuation time (addressed in previous sections), quantitative travel elements (travel speed, flow conditions/constraints and their relationship to population density) and qualitative elements (route availability and choice). This section focuses upon the quantitative travel elements only—Chaps. 57 and 58 should be referred to for the more qualitative factors mentioned.

In an engineering context, crowd movement is quantitatively specified using three key characteristics. These are density, speed, and flow. These underpin the engineering hydraulic model presented in Chap. 59 and are also employed or generated in computational egress models. Within the hydraulic model, relationships are assumed between density and speed, and density and flow that then directly influence the results produced. As such, data is often collected relating the three measures in this manner. It is debatable as to the exact causal relationship between flow/speed and density, with density being both an emergent property of crowd movement and a constraining factor. However, in this instance, the relationship is retained given the format and content of the data-sets provided.

Population density is generally expressed as the number of persons in a unit area of measured space, e.g. 2.0 persons/m². Alternatively, this factor can be represented by (i) using the inverse of density, that is, the area per person or pedestrian module (occupancy levels), e.g. 0.5 m² (5.4 ft²) per person, (ii) the distance maintained between occupants, in terms of headway or proximity, or (iii) the perpendicular projected area occupied by the population over a unit area (e.g., m²/m²) [105, 106]. Given that the density (persons/unit area) is still most commonly used, it will be adopted here and a more comprehensive discussion of this issue is left for future editions of this chapter. It is acknowledged that there are limitations with all of these methods of representation, including the population density. These limitations include the averaging of the densities over a large area, the inclusion of persons within the variable that are not directly influencing movement, and the variety of body sizes not represented within the analysis [105].

Speed is the distance covered by a moving person in a unit of time, e.g. 1.0 m/s (3.3 ft/s). The term “flow” is often used in a casual, nontechnical way when the general term “movement” is implied, or when speed is actually being specified. However, flow is specifically the number of people that pass some reference point in a unit of time, e.g. 2.0 persons/s. Flow can be presented in relation to the unit width, in which case it is termed the specific flow and it is presented in units of persons/unit width/time period or it can be presented across a component of a particular width where it is presented in units of persons/time period. In the tables which follow the specific flow is typically employed; however, it will be made clear for each of the data-sets where other formats are employed.

Assuming no other constraint is present, people can move at their desired speed (given their innate capabilities) if there is sufficient space available for them to move freely, i.e. if the surrounding population does not constrain their movement. The desired speed may change with motivation, which can be influenced by time constraints, social considerations and perceived risk, amongst other things, e.g., the nature of the scenario. However, as the space available reduces, so the ability of the individual to maintain their desired speed, and act on their intentions, is reduced. There is a relationship between the ability to maintain a speed and the local population density.

Relationships between speed and density for specific terrains are provided in Chap. 59. These are based on the work of Nelson and Maclennan [107], who derived these relationships from the work of Fruin [34], Pauls [33], and Predtechenskii and Milinskii [106] inter alios. The relationships derived by Nelson and Maclennan [107] represent engineering approximations of several diverse data-sets. As such, these relationships should not be seen as empirical data but as engineering representations.

Another example of this engineering representation is shown in Fig. 64.4. This is derived by Pretorius [108] from the work of Nelson and Maclennan [107] and Kholshevnikov et al. [109]. Kholshevnikov et al. [109] highlighted the different influences that population density might have upon the physical and psychological factors influencing individual movement. Pretorius transposed this onto the assumed relationship between speed and flow produced by Nelson and Maclennan [107] and presented in the SFPE handbook. Pretorius achieved this by using slightly different body sizes (derived from Still [110]) and assuming an engineering rather than empirical approach.

Fig. 64.4

Derived relationship between speed and density [108]

The relationship between flow and density is more complex, given the number of variables involved, i.e., that flow is dependent on the speed of movement and the number of people moving. This relationship tends to be a higher order function than that between speed and density.

The relationship between speed, flow and density becomes more complex as the degrees of freedom in movement increases—the relationship is different on the horizontal from that on the vertical, where movement is more three dimensional in nature. Indeed, some have questioned the direct relationship between speed/flow and population density on stairs, given the impact of the stair configuration upon gait and forward movement [105]. Notwithstanding, the direct relationship is implicitly assumed in much of the data presented, and the nature of this relationship is not investigated here.

A number of reviews of movement data already exist in the literature; they either combine disparate data-sets from a variety of different researchers, or summarize the work of a suite of data collection exercises performed by the same researcher(s). In the following sections, some of these reviews are presented as examples of data-sets collected for a particular behavioral or movement element. However, a number of broader reviews exist that relate to more than one element. These include reviews by Predtechenskii and Milinskii [106], Smith [111], Tubbs and Meacham [112], Thompson [113], Fahy and Proulx [37], Graat et al. [114], Lord et al. [115], Daamen [116], Teknomo [117], Helbing [118], Fruin [34], and Pauls [33]. Many of these broader reviews have been conducted as part of research dissertations and so fall outside of the relatively narrow scope of this chapter.

It is important for the user to understand the nature of the relationship between these variables before applying any of the data presented in this chapter. Some of the models (engineering or egress) assume this relationship [27]. Others can adopt different relationships depending on the data provided to them or the rules embedded, with the relationship occasionally predicted and generated rather than imposed. Either way, it is important for the reader to understand that this relationship is often assumed without due reference to the implications of this assumption.

When employing these quantitative data-sets, a number of qualitative engineering decisions will need to be made. These relate to the use of the space (width) available, the routes available and the routes used, amongst many others. The selection and credible combination of these qualitative decisions are discussed in detail in Chap. 57. The routes available during an incident will be influenced by the environmental conditions present (discussed in Chaps. 61 and 63). The choice and use of the available routes will be influenced by a number of population characteristics (discussed in Chap. 58) and procedural measures (discussed in Chaps. 56 and 58). Guidance should therefore be sought from these other chapters in order to frame the use of the quantitative data-sets presented below.

In the following sections the movement on the horizontal (see section “Travel Phase: Horizontal Movement”), on stairs (see section “Travel Phase: Stair Movement (Up and Down)”), through exits (see section “Travel Phase: Exits and Narrowings”), on escalators (see section “Travel Phase: Escalators (Up and Down)”), in smoke (see sections “Vulnerabilities: Situational: Movement in Smoke” and “Vulnerabilities: Innate: Impaired Movement”) and movement involving those with impairment (see section “Vulnerabilities: Innate: Impaired Movement”) is discussed. Where there is other important review material or material that does not meet the selection criteria highlighted in section “Sources of Data” then this is briefly discussed as are the table formats employed.

Travel Phase: Horizontal Movement

A number of previous reviews have been conducted exploring the relationship between density and speed [37, 38, 109, 113, 116, 119–122]. Irrespective of the material covered, an array of different approaches (i.e. numerical, graphical, descriptive, tabular, etc.), have been adopted to summarize the data and equations have been derived [9]. The approach adopted by a particular reviewer is likely to have been influenced by their theoretical understanding of the factors involved, the purpose of the summary, e.g. intended application area, and the material included in the review itself. Several examples of the different approaches adopted are presented below. The reader is encouraged to examine some of the reviewed material and analysis available from the original source, as a great deal of it falls outside of the remit of this chapter.

Kholshevnikov et al. presented their own work and that of a number of other Russian/Soviet researchers performed over a period of approximately 30 years [109]. This data was gathered in a number of different building types, population types and in a range of environmental conditions. Of particular note in this work is the extreme population densities included in the reviewed material (see Fig. 64.5).

Fig. 64.5

Speed—density relationship for horizontal movement presented by Kholshevnikov et al. [109] (Reproduced from original)

Brocklehurst [120] and Daamen [116] both reviewed a range of material relating to the speeds produced during general circulation activities but presented the material in different ways. Brocklehurst [120] examined the work of Hankin and Wright, Peschel, Henderson, Helbing, Fruin, Older, Polus, Still and presented the average speeds produced given the background factors present during the original events. These factors included the nature of the event, the location, densities, gender/age of the population and the speeds produced. Daamen [116] produced a graphical representation of the data of Fruin, Weidmann, Virkler, Older, Sarkar, and Tanariboon, a functional representation of speed and flow (including the data of O’Flaherty, Lam, Navin, Pauls, Pushkarev in addition to those represented graphically) and mean speed estimates (from data of Daly, Fruin, Hankin and Wright, Henderson, Hoel, Knoflacher, Koushki, Lam, Morrall, Navin, O’Flaherty, Older, Pauls, Roddin, Sarkar, Sleight, Tanariboon, Tregenza, Virkler, and Young). It should be noted that both Brocklehurst [120] and Daamen [116] presented significant reviews of data within their doctoral theses; dissertations such as these are useful sources of data, given the space available for a more detailed review and description of the material available (see Fig. 64.6).

Fig. 64.6

Speed—density relationships summarized by Daamen [116] (Reproduced from the original)

Al-Gadhi [119] also examined a number of pedestrian sources (as well as many collected himself from situations such as the Hajj) that included unidirectional and bidirectional movement, whilst Schadschneider et al. [122] examined several different sources that might be used to support an understanding of evacuation dynamics to facilitate numerical and computational model development. This included a variety of different sources (in terms of the scenario, the geographical origin and the age of the data), including the work of Weidman, Predtechinskii and Milinskii, Older and Helbing [122].

Thompson [113], in his seminal thesis outlining his model development, presented a broad review of core data in order to develop an understanding of the data required to develop a computational egress model and indeed went on to collect his own data in order to develop his model. Thompson [113] summarizes, in some detail, the work of a number of key researchers including Fruin, Predtechenskii and Milinskii, Hankin and Wright, Pauls, Weston and Marshall, Peschl, Ando et al., and Polus, Schofer and Ushpiz.

Finally, in the same vein as the work conducted by Fahy and Proulx [37], Shi et al. [38] attempted to collect together a small number of data-sets in order to develop a database; although in this instance it was conducted to support the development of evacuation modeling tools. Shi et al. [38] produced a table where the data was categorized according to several factors deemed to influence the results produced (see Table 64.14); i.e. free movement or exit movement, density levels (low, optimum, moderate, crush), location type and occupant type. Other comparable reviews have also been conducted [114, 117, 118, 124–126].

Table 64.14 Summary of walking speeds produced by Shi et al. [38] (Reproduced from original)

Data appears (both within fire engineering and from adjacent areas of analysis) in a number of formats and representing an array of different scenarios. This data broadly reflects the difference performance capabilities of pedestrians under different physical and social conditions.

An examination of the review material available was conducted and the data simplified and compiled into a single figure (see Fig. 64.7). Given the lack of contextual information, the different criteria used to include representative data-points and the different results formats used, this figure should, at very best, be seen as only indicative. However, it is does crudely demonstrate that much of the data collected (and the relationships produced) focus on a relatively narrow band of population densities; i.e. between 0 and 6 persons/m². Furthermore, the range of speeds produced within this envelope of densities appears to narrow as the density increases—from a relatively wide range of speeds between 0 and 2 persons/m² that narrow beyond 4 persons/m². This claim is supported by the work presented by Kholshevnikov [67, 109].

Fig. 64.7

Simplistic compilation of reviewed data on horizontal movement speed derived from the work of Hankin, Helbing, Oeding, Daamen (Weidmann, Virkler, Tanariboon, Sarkar, Polus, Lam), Predtechiniski and Milinskii, K+V, VTT, Weston and Marshall, Navin, V+S, Mori, Averill, Ando, Hoskins, Duives [13–15]

A summary of a range of data relating to travel speeds achievable during horizontal movement are presented in Table 64.15, while the subsequent flow rates produced are described in Table 64.16. It should be noted that there is some overlap between the data described above and the data-sets that appear in Tables 64.15 and 64.16. In Table 64.15, the following information is provided: Source, Observational Conditions (Location, Nature, Spatial Configuration, Participants, Environment, Variable), Sample (Collection method and Size), Results (Density, Speed, Relationship between Speed and Density), and Additional Information. In Table 64.16, the following information is provided: Source, Observational Conditions (Location, Nature, Spatial Configuration, Participants, Environment, Variable), Sample (Collection method and Size), Results (Density, Specific Flow, Relationship between Flow and Density), and Additional Information. In both cases, this information is only provided if it appears in the original source or if can readily be derived. In a limited number of cases, the values provided in the tables, although not given explicitly in the original source, have been derived from other information provided in the original source; e.g., a calculation of a mean from the source data. If a value has been derived in this way, then that is noted in the table beside that value.

Table 64.15 Travel data—horizontal travel speeds (m/s)

Table 64.16 Travel data—horizontal flow

Travel Phase: Stair Movement (Up and Down)

In tall structures, especially those where elevators have not been approved for use in fire emergencies, stair movement may form a considerable part of the overall t_RSET calculation. Even in lower rise buildings, stair movement may still be a pivotal, especially for those with difficulties traversing stairs. Indeed, stair movement has determined the outcome of a number of serious incidents [8, 20, 37, 145]—with both the attainable speed, the impact of congestion, fatigue/rest effects, and the direction of required movement all influencing the nature of the outcome.

Movement on stairs presents a different challenge than horizontal movement given the extra degree of freedom available in the movement, the constraints imposed by the stair design and the additional effort required to traverse the stair component. In addition, the relationship between density and speed/flow may be more complex during vertical movement than it is on flat surfaces. For instance, the size and configuration of the treads may have a confounding influence upon this relationship. This issue has recently been examined by Hoskins [105].

Over the last few decades, a number of researchers have examined pedestrian and evacuee performance on stairs. Some [11, 12, 37, 105, 109, 114, 115] have addressed the quantitative aspects of stair movement e.g. speeds and flow rates, while others [33, 146, 147] have focused on the more qualitative aspects of stair movement, e.g. spacing, passing, merging behavior, counter-flow, lane formation, group behavior, gait, falls, handrail use.

Given the importance of stair movement, there are a number of significant reviews available regarding stair data. Some of these, such as the work of Hoskins [105], have attempted to derive representative functions to describe the relationship between density and speed on stairs, and have provided a detailed review of a range of other speed/flow conditions produced and the maximum densities measured by researchers such as Pauls, Khisty, Kagawa, Proulx, Shields, Kratchman, Hostikka, Peacock, Fruin, Daly, Tanaboriboon, Lee, Ye, Galea, Averill, Melinek, Predtechinskii, Templer, Smith, Frantzich, Boyce, Wright, and Fujiyama [105]. It should be noted that Hoskins [105] then goes on to produce a different formulation to generate stair speeds from that typically employed relating stair flow/movement to the physical conditions present.

Shi et al. [38] also examined stair movement data and categorized it according to dimensions, incline and configuration of the stair. Peacock et al. [45] summarized their own studies involving the evacuation of a series of structures involving stair movement, placed the evacuations in a historical context (see Fig. 64.8) and then compared the speed/density relationships with the SFPE curve derived by Nelson and Maclennan (see Fig. 64.9).

Fig. 64.8

Historical comparison of stair speeds produced by Peacock et al. (Reproduced from original [45])

Fig. 64.9

Comparison of speed-density relationships—Peacock et al. and SFPE produced by Peacock et al. (Reproduced from original [45])

Kholshevnikov et al. [73] presented a summary of movement data relating stair descent speeds to population density within multi-purpose, sports buildings, universities, schools, street areas and experimental rigs in the Soviet Union/Russia over a 30 year period. The results are shown in Fig. 64.10. Once again, the elevated population densities present in the Russian studies, relative to the other studies presented above (e.g. those of Peacock et al. [45]) should be noted. Other reviews include those produced by Fahy and Proulx [37] (that categorize performance according to building type and presence of high densities), Graat et al. [114] (where the speeds are categorized according to direction of movement and location type), and Lord et al. [115] (where speeds are categorized according to direction, demographics, and likelihood).

Fig. 64.10

Kholshevnikov et al. review of stair descent data (left) and ascent data (right) (Reproduced from original [109])

A compilation of a range of data relating to travel speeds during movement on stairs is presented in Table 64.17, while the subsequent flow rates produced are presented in Table 64.18. In Table 64.17, the following information is provided: Source (Author and Year), Observational Conditions (Location, Nature, Spatial Configuration, Participants, Environment, Variable), Stair Configuration (Direction of Movement (up/down), Slope, Distance), Sample (Collection method and Size), Results (Density, Incline Speed, Relationship between Speed and Density), and Additional Information. In Table 64.18, the following information is provided: Source, Observational Conditions (Location, Nature, Spatial Configuration, Participants, Environment, Variable), Stair Configuration (Direction of Movement (up/down), Slope, Distance), Sample (Collection method and Size), Results (Density, Specific Flow, Relationship between Flow and Density), and Additional Information.

Table 64.17 Travel data—speed of movement (crowds) on stairs

Table 64.18 Travel data—stair flows

Travel Phase: Exits and Narrowings

As described in Chap. 59, a key element in assessing the movement of sub-populations around a space is the change of widths of the spaces traversed. These typically occur at exit locations, which act as a connector between adjacent spaces. These changes influence the local population densities experienced, along with the potential for operating exit devices (e.g. the ability to get the door open or open to its full extent given the local population density). As such, exits (which include a device for separating two or more spaces that typically also represents a narrowing) and openings (a breach in space separation that typically represents a change in the width of the space) provide an important constraint to the movement of sub-populations within a structure, and certainly one that should be supported by empirical data within any egress analysis.

The representation of exits and openings can be addressed in a number of ways ranging from imposing a flow rate that limits the achievable flow, to allowing the flow conditions to emerge from the predicted conditions produced by the model [9, 22] (see also Chap. 60). Given the variety of attributes that may be required to produce a genuinely predictive model and the manner in which data-sets are typically presented, the following data, in its current form, is likely to be of most value in capping flow rates and validating predictive approaches, This does not preclude its use to configure predictive models; it may, however, mean that the data needs to be manipulated for the specific methods adopted within the predictive model being employed.

Many of the previously reviewed examples of horizontal movement also include reference to flow through exits and openings as a key constraint upon horizontal movement, i.e. a change in the density and flow conditions present. It is also explicitly represented in the original work by Nelson and MacLennan [107] and in Chap. 60. A wide range of previous empirical work in this area is also described in Daamen’s thesis [116] and Daamen and Hoogendoorn [121] have published an array of their own empirical results describing exit flows by exit width, lighting levels, population type and door status (see Fig. 64.11). Daamen [116] also provides estimates for flow capacities associated with a range of different exits in rail rolling stock.

Fig. 64.11

Daamen and Hoogendoorn flow capacity of the exit as a function of population type (Reproduced from original [116, 121])

A number of other works provide specific examples of flow performance through exits and openings (be they narrowings or bottlenecks). These include Seyfried et al. [160, 161], who examined the flow rates generated at exits and narrowings based on their width and the densities evident, and reviewed the estimations of Nelson and MacLennan, Predtechenskii and Milinskii, Weidmann, Hoogendoorn, and the measurements of Kretz, Nagai and Muller (see Fig. 64.12).

Fig. 64.12

Seyfried et al. comparison of flow performance (/s) through exits with work of others (Reproduced from original [160, 161])

Rinne et al. [125] performed a number of evacuations and monitored the performance of 32 exits to establish the flow rates produced. The exits were categorized according to the 22 evacuations from which the data came, the nature of this evacuation, the type of door involved, the door width, the flow performance produced, the people involved and the status of the door [125]. The data from these evacuations are shown in Fig. 64.13.

Fig. 64.13

Rinne et al. data relating to exit performance (Reproduced from original [125])

Several researchers have also looked at the travel speeds that can be achieved at and around exit points. This data can be used where flow rates are predicted/simulated by models, as opposed to being imposed by the user. Kholshevnikov et al. [109] presented a vast amount of data collected over 30 years in Russia from a number of data collection activities and combined it with their own efforts to produce the relationships between speed and density at exits, shown in Fig. 64.14. This is categorized according to the population densities recorded (which ranged from 1 to 13.5 persons/m²) and the occupancies involved (retail, sports structures, train station, and experimental surroundings).

Fig. 64.14

Kholshevnikov et al. relationship between speed and density relating to exit opening (Reproduced from original [109]). Where curve (1) represents different buildings, (2–4) retail, (5) sports structures, (6–9) underground stations, (10–14) experimental conditions

Schadschneider et al. [122] reviewed the experimental work of Kretz, Muir, Muller, Nagai and Seyfried, looking in some detail at the experimental conditions (and the initial distribution of the participants) along with the width of the exit/narrowing and the existence of competition among the participants (see Fig. 64.15).

Fig. 64.15

Speed—density relationships summarized by Schadschneider et al. [122] (Reproduced from the original)

The above paragraphs give a flavor of the range of studies regarding the performance of exits and narrowings. Others which the reader may wish to refer to include Thompson [113], Kendik [162], Smith [111], Gwynne et al. [47, 163], and Gwynne [164].

A summary of other data on exit flow through openings is given in Table 64.19. Although there is some overlap with the review material discussed, it is presented in a more consistent manner, in context with the scenario conditions and is presented in more detail. In order to achieve this, the following information is provided: Source, Observational Conditions (Location, Nature, Spatial Configuration, Participants, Environment, Variable), Door Configuration (Type—Single/Double/Sliding, Direction of Opening Relative to Movement—Against/ With, Width), Sample (Collection method and Size), Density, Results (Component Flow and Specific Flow), and Additional Information.

Table 64.19 Exit flow

Travel Phase: Escalators (Up and Down)

Kinsey [173], in his PhD dissertation, performed a review of data currently available related to movement on escalators, collected and presented more data on escalator usage (speeds and flows) and also considered decision-making in relation to escalator usage.

Tables 64.20 and 64.21, provide a summary of published data on speeds and flows on escalators, respectively. This includes the data reviewed by Kinsey and Kinsey’s own data. The following information is provided in Table 64.20: Source Observational Conditions (Location, Nature, Spatial Configuration, Participants, and Variable), Escalator Configuration (State (Moving/Static), Direction (up/down), Speed, Height, Horizontal Length), Sample (Collection method and Size), Results (Mean and Range) and Additional Information. It should be noted that horizontal speed is utilized in Table 64.20 rather than incline speed presented previously in Table 64.17 relating to stair movement. Horizontal speed is presented here since this is the speed commonly presented in the literature to describe movement on escalators. It should also be noted that in the case of moving escalators that the speed presented is the speed of movement of the individual over and above that of the speed of the moving escalator.

Table 64.20 Travel—unimpeded horizontal walking speed on escalators

Table 64.21 Travel—flow on escalators

In Table 64.21, the following information is provided: Source, Observational Conditions (Location, Nature, Spatial Configuration, Participants, and Variable), Escalator Configuration (State, Direction, Horizontal Speed, and Width), Sample (Collection method and Size), Peak Recorded Flow, and Additional Information. It should be noted that, unless otherwise specified, the width of the escalator is the step width.

Vulnerabilities - Situational: Movement in Smoke

The presence of smoke during an emergency has been shown to have a range of different effects upon evacuee performance. These effects may be psychological (see Chap. 58), physiological and physical (see Chaps. 61 and 63). These impacts may affect the performance in a number of ways in terms of the following:

• Recognition and response to fire (t _pre )—in that the presence of smoke may act as a cue to indicate the existence of an incident and may then influence the time for an individual to respond to the incident,

• Redirection of movement—in that the perception of the presence and severity of the smoke may discourage the use of a particular route,

• Reduction in the efficiency of movement (t _trav )—where the presence of smoke can lead to reduced walking speeds [31] and

• Posture change—in some cases the presence of smoke may lead occupants to abandon walking and adopt crawling behaviors, which in turn can influence the travel speeds that can be attained [31].

It is specifically the last two effects (i.e. the impact of smoke on walking speed and crawling speed), that are discussed in this section. Some discussion regarding the first two impacts listed above (i.e. impact of smoke on recognition and response and impact of smoke on exit choice behaviors), is presented in Chaps. 58, 61, and 63. Data relating to these aspects have also been reviewed in some detail by Xie Hui [182], Ronchi et al. [183], Gwynne [164], and Nilsson [184].

It is now widely acknowledged that some people are prepared to move through smoke in order to reach a position of safety [31]. In some circumstances (e.g. tunnels), it is accepted that evacuees will be exposed to smoke for a short duration [185]. However, in most engineering applications, the interaction between the evacuating population and smoke is assumed not to take place, i.e. the separation of the evacuating population and the deteriorating environment is assumed to be successful. Although engineering designs are likely to be developed on the basis that evacuees are not exposed to deteriorating environmental conditions, the potential for this interaction should still be considered. There may be situations, for example, where the engineer is testing the robustness of his/her design, making extremely conservative assumptions regarding the performance of the evacuating population, or going beyond the standard performance based scenarios considered as part of engineering design. For that reason, information and data regarding movement in smoke is included in this chapter.

It has been estimated that over 60 % of evacuees in small residential buildings move through smoke to evacuate [36]. A study of the behavior of evacuees from the bomb blast in the World Trade Center in 1993 [36] showed that 94 % of occupants of Tower 1 moved through smoke. Given the location of the incident, and the routes available once the evacuation commenced, movement through smoke was unavoidable; however, this did not prevent people commencing or continuing their evacuation. Detailed study of the evacuation of two high-rise residential buildings [36] has also shown that around 96 % of the occupants located above the fire floor moved through smoke.

Studies by Bryan [30] have shown that the proportion of people turning back rather than entering smoke increases with the smoke density. However, the exact proportion of those willing to move through smoke will be scenario-specific. In a space relatively free of smoke where smoke-free options are available, people may turn back rather than attempting to move through dense smoke, while those escaping the enclosure of fire origin or those with no alternative options will continue to move through often dense smoke, particularly if they know it will lead to an exit in a relatively short period of time. The decision to move through smoke may also depend on the person’s motivation, e.g. in dwelling fires, parents are prepared to move through smoke to rescue their children or other precious belongings. In tunnel fires, some have been known to walk for several hundred meters in dense smoke [186].

The examples given in the previous paragraphs have demonstrated that people are willing to move through smoke when required to do so. It is clear, however, that the actual percentage of people that will be willing to move through smoke may be highly dependent upon the scenario faced and the options available. In this section, the actual percentage of people that are willing to move through smoke is not the focus of the discussion; instead, it is the impact that the smoke has on movement performance that is considered in some detail.

Given that people may interact with smoke, it is important to consider the impact that smoke has on walking speed. This section considers the available data sets that describe (1) the correlation between the smoke and occupant speed assuming the individual remains upright, and (2) the travel speeds that can be expected should an individual decide/be forced to crawl. The likelihood of an individual initiating crawling behavior is not addressed.

In section “Upright Movement in Smoke” we present several data-sets related to the travel speeds during the upright movement in smoke on a horizontal place. In section “Crawling Movement in Smoke” we present several data-sets on the travels speeds attained when crawling in smoke. In both cases, only a small number of data-sets are available and, therefore, rather than presenting in tabular form, a more detailed discussion of the data-sets is presented.

In the studies which have considered the impact of smoke on movement, smoke conditions are generally represented through an extinction coefficient (m⁻¹) or optical density (m⁻¹) (see Chaps. 61 and 63). The extinction coefficient unit is adopted here. The smoke has a number of influences upon an individual’s ability to move: (1) highly irritant smoke can cause intense pain to the eyes, affecting the individual’s ability to keep his/her eyes open; (2) the extinction coefficient of the smoke reduces visibility levels. The toxic and irritant agents can also have physiological effects that further reduce the individual’s well-being and then indirectly reduce their walking speed (see Chap. 63).

Before examining or employing the data discussed below, it is important to note that the experimental conditions present in these studies vary considerably in terms of the presence and type of irritant gases, the characteristics of the participants, the structural configuration employed in the trials, and measurement techniques adopted. The variation in experimental conditions is a feature of all of the data discussed in this chapter; however, particular care should be taken with regards to movement in smoke, given the limited data available. This area of investigation is particularly susceptible to ethical constraints which may limit the conditions to which trial participants might be exposed, and which inevitably impacts on the credibility and utility of the results. Again, it is recommended that readers should refer to the original papers before making decisions as to the most relevant data set to use.

The Ph.D. theses of Xie Hui [182], Nilsson [184], Kady and Davis [187], Ronchi [188] include extensive reviews of the data available relating to different aspects of movement in smoke and the reader is referred to these for further information.

Upright Movement in Smoke

Four data-sets which suggest a correlation between extinction coefficient and attainable speed are presented below: Jin [189], Frantzich and Nilsson [190, 191], Wright et al. [192], and Galea et al. [193]. An overview of these data-sets, derived from the work of Xie Hui [182], is provided in Table 64.22.

Table 64.22 Overview of smoke travel speeds for upright individuals (Derived from the work of Xie Hui [182])

The first set of experiments in this area was performed by Jin in 1976 [189, 194–196]. The details of his experiments are discussed at length in Chap. 61 of this handbook; therefore only a brief overview is provided here. Jin tested 38 individual subjects (age range 20–51 years) in a 20-m-long corridor filled with two types of smoke analogous to the early stage of a fire: the first was a highly irritant white smoke produced by burning wood cribs and the second, a less irritant black smoke produced by burning kerosene. From the experimental results Jin derived a correlation between the extinction coefficient and the achievable walking speeds for both the irritant and non-irritant smoke (see Fig. 64.16).

Fig. 64.16

Walking speed in smoke from Jin [189] (Reproduced from original)

In the experiments involving irritant smoke, the visibility fell away sharply as the extinction coefficient approached 0.5 m⁻¹. The occupant speed of movement, which was initially over 1 m/s reduced rapidly to approximately 0.3 m/s as extinction coefficients reached approximately 0.5 m⁻¹. The sharp drop in walking speed is explained by the fact that the subjects could not keep their eyes open in the irritant smoke, causing a zigzag movement or use of the walls for guidance (see Chap. 61).

A more gradual decrease in walking speeds was evident in non-irritant smoke (see Fig. 64.16). Participants once again moved initially at over 1 m/s and slowed down to approximately 0.5 m/s when the extinction coefficient reached 1.0 m⁻¹. In these trials, participants continued to walk at a reduced speed, behaving as if in darkness and feeling their way along the walls, but without the associated discomfort and clenched eyes that were apparent in the irritant smoke. The Jin data-set, although small, provides a direct comparison between the impact of irritant and non-irritant smoke in comparable conditions.

Frantzich and Nilsson’s study [190, 191] (see also Chap. 61) involved a series of experiments conducted in a smoke-filled tunnel. The sample comprised 30 males and 16 females aged 18–29 years (45 valid measurements were made). Participants were asked to walk through a tunnel 37 m long by 5 m wide, with a ceiling height of between 2.5 and 2.7 m in one of two illumination levels (lights on (21 lx) and lights off) and wayguidance systems (flashing lights, rows of flashing lights and floor markings). The route was more complex than that in Jin’s experiments in that participants had to negotiate a number of obstacles, namely pillars which supported the ceiling and six cars placed inside the tunnel. The tunnel was filled with artificial smoke, with acetic acid used to provide irritation. The range of extinction coefficients measured was 1.9–7.4 m⁻¹, which was greater than in Jin’s experiments. The participants were provided with the scenario that they had entered the tunnel in their car, and they were asked to get out and act as they would in such a situation if it was a real incident. The movement of participants was filmed using thermal imaging infra-red cameras. The calculated walking speeds ranged from 0.2 to 0.8 m/s (speeds included time stopping to rest) and results indicate that the speed decreased across the data-set as the extinction coefficient increased. Subjects also continued walking despite the low visibility and also tended to walk using the tunnel walls for guidance, an observation also noted by Jin (see Chap. 61,).

A comparison of Frantzich and Nilsson [191] and Jin’s data [189] is provided in Fig. 64.17.

Fig. 64.17

Comparison of Frantzich and Nilsson and Jin’s data (Reproduced from Ronchi [188])

From Fig. 64.17 it is apparent that there is a wide scattering in the data currently available, which seems to have been due to differences in participants’ characteristics, the levels of irritancy produced, the complexity of the spaces (Jin’s experiments involved a simple corridor whilst Frantzich and Nilsson’s was more complex including obstacles), and potentially the wayguidance systems in place. It should also be noted that the participants were in close proximity to the walls in both the Jin and Frantzich and Nilsson trials; the opportunity for evacuees to use walls as a means of guidance may well be different in an open space which in turn may influence their behavior and performance. It is therefore difficult to determine, with any level of confidence, a representative walking speed for a particular extinction coefficient. Frantzich and Nilsson developed two regression models from their data [191] for situations with no illumination (lights off) and illumination (lights on). The relationship between the smoke extinction coefficient K and walking speed for the illuminated movement (R² = 0.32) is given below. This is the relationship that has been adapted by Korhonen and Hostikka [197] for use in the FDSEvac model.

$$ Walking\ Speed\left(m/s\right)=0.706-0.057K $$

Frantzich and Nilsson [191] also conducted regression analysis to consider the influence of being in proximity to walls and produced the following (R² = 0.454):

$$ Walking\ Speed\ \left(m/s\right)=0.692-0.073K+0.139\left( proportion\ wall\ contact\right) $$

Wright et al. [192] also conducted experimental trials to examine travel speeds in a smoke-filled environment that employed different lighting and wayguidance systems. The trials were performed in a two-storey test facility that included a 13 m corridor. Non-toxic white smoke was generated from a mineral-based fluid. A single smoke density (extinction coefficient 2.5 m⁻¹, OD 1.1 m⁻¹) was maintained in the corridor during the trials. The trials therefore focused on the performance of different systems at a given extinction coefficient. Eighteen participants (7 men and 11 women, ranging from 23 to 63 years of age, mean 46 years) took part in the trials and experienced the same set of six scenarios; i.e. five way guidance systems separately employed and a separate scenario with normal overhead lighting only. The speeds recorded ranged from 0.61 to 0.71 m/s for normal lighting [192], producing an average speed of 0.5 m/s. A range of other scenario conditions, employing two different LED systems, incandescent lighting, electroluminescent lighting and overhead emergency lighting were also examined. These produced mean speeds of 0.85 m/s (range 0.79–0.91 m/s) for LED1, 0.82 m/s (range 0.76–0.88 m/s) for LED2, 0.84 m/s (range 0.78–0.91 m/s) for Incandescent lighting, 0.75 m/s (range 0.69–0.81 m/s) for electroluminescent lighting and 0.53 m/s (range 0.48–0.58 m/s) for overhead emergency lighting. Other trials were conducted by Wright et al. and are described in detail elsewhere [192].

Galea et al. [193] also conducted a series of experiments to collect data on human movement and behavior that might occur during a maritime evacuation. The experiments were conducted in a specially designed Ship Evacuation Behaviour Assessment Facility (SHEBA), and involved an assessment of the walking speeds of 360 participants as they traversed an 11 m corridor. Measurements of participant speed were performed in a clear (smoke-free) environment and in a smoke-filled environment with extinction coefficients of 0.0, 0.23, 1.15, and 2.30 m⁻¹. These were conducted at various degrees of heel and trim (i.e. pitching perpendicular to the direction of movement or in the direction of movement), in order to represent various sea states and scenario conditions; however, only the trials conducted on the flat are reported here, given their potential applicability to the built environment.

According to the SHEBA data-set, the participants’ average unimpeded walking speed measured in a smoke free environment was approximately 2.2 m/s. This speed is much higher than the unimpeded walking speeds suggested in the Jin [189] and Frantzich [191] data-sets (see Chap. 61 and Frantzich and Nilsson [66, 190, 191]) and has been attributed [193] to a combination of factors, i.e. the participants in the SHEBA data set were highly motivated, familiar with the environment, required to perform relatively straightforward tasks and traverse a shorter route. Notwithstanding, the results indicate a reduction in walking speed with increased smoke density which is consistent with other studies, see Fig. 64.18.

Fig. 64.18

Relationships between walking speed and extinction coefficient from a number of researchers derived by Xie Hiu [182]

The SHEBA study also defined ‘mobility’ as the walking speed in smoke as a fraction of the walking speed measured in a smoke free environment (i.e., the speed as a fraction of the speed attained in a clear environment). Figure 64.19, compiled by Hui Xie [182] illustrates how the mobility varied with different smoke densities for the SHEBA data set, and how this compares with the derived ‘mobility’ from the Jin and Frantzich and Nilsson studies. It is apparent that the relative reduction factor is broadly comparable across the three studies examined.

Fig. 64.19

Relationships between speed reduction factor and extinction coefficient derived by Xie Hiu [182]

Xie Hui [182] performed a further analysis on the results from Jin’s work (as part of his model development), and reported the relationships between extinction coefficient, K, and the reduction in walking speed as being (i.e. relative to performance in smoke-free environment):

$$ Reduction\ Factor = -0.16{K}^2-0.488K+1.105 $$

between extinction coefficients of 0.2 and 1.0 m⁻¹ for non-irritant smoke and

$$ Reduction\ Factor = -2.08{K}^2-0.38K+1.06 $$

between extinction coefficients of 0.1 and 0.5 m⁻¹ for irritant smoke.

These equations might then be applied to an individual’s initial travel speed to estimate the impact of the smoke conditions upon performance. This assumes that the effect is relative to an individual’s initial starting performance rather than producing an absolute effect upon all individuals in the same manner and then determining a specific walking speed for a given extinction coefficient; i.e., the impact gets progressively more severe in relation to the individual’s initial performance rather than being independent of the individual involved. These two assumptions are discussed in detail by [183].

Crawling Movement in Smoke

As discussed in Chaps. 61 and 63, the physiological effects of exposure to fire can influence evacuees’ decision making and performance. The descending hot smoke layer can also lead occupants to crawl in an attempt to evacuate the building safely—either forcing them to do so in an attempt to avoid the worsening conditions or instigating a planned response to the presence of smoke through some procedural instruction or training. It may therefore be important, depending on the complexity of the engineering analysis being performed, to refer to data on crawling speeds and to understand the interrelationships between crawling speed and density. This is discussed in the paragraphs below.

Compared to the research on the upright walking speeds, there has been relatively little research conducted in this area until recently. The most significant being the work of Muhdi et al., Nagai et al., and Kady [170, 198], Kady et al. [187, 199].¹ The study by Muhdi et al. [200] was designed to measure and compare individual maximum and normal walking and crawling speeds, and concluded that (assuming a comparable level of effort), crawling results in a significant reduction in speed compared to walking. His later study [187] compared the walking speed and crawling speed of persons of different body composition (normal, overweight and obese) and concluded that gender and body composition were major determinants of occupant normal crawling speed, accounting for 80 % of the variance. He again showed that the mean individual normal crawling speed was significantly less than the mean walking speed, with the mean crawling speeds of 0.79 m/s being comparable to that of 0.71 m/s obtained in his earlier study [200]. The study by Nagai et al. [170] also compared speeds of walkers and crawlers through a corridor and exit, and examined the impact of crowd density on crawling and walking speeds. The average individual normal crawling speeds measured in these experiments was 0.73 m/s which is also comparable to that obtained by Kady and Muhdi [187, 200]. Kady [199] further investigated the relationship between speed and density and concluded that the location of crawlers relative to the exit affects the density of crawlers produced and subsequently the speed of the individual crawlers and the population movement speed. The relationship derived between crowd crawling speed and crawling crowd density is shown in Fig. 64.20 and can be expressed by:

Fig. 64.20

Relation between crawling speed and density on a flat surface (From [199])

$$ Crawling\ Speed=0.7973+0.2909D-0.1503{D}^2 $$

It appears that this quadratic model provides a good fit to the data (p = 0.004), whilst the R² value indicates that crowd density accounts for 42.7 % of the variability in crowd crawling speed.

A summary of the current data on crawling speeds is given in Table 64.23. In addition to the crawling speed data (means, standard deviation and ranges), Table 64.23 provides information on the Source Observational Conditions (Location, Nature of the experiment, Spatial Configuration (details of the test route), Participants and Variables). An Additional Information column provides other relevant information related to the instructions to the subjects and precision of measurement. In all of the experiments, crawling was considered to be achieved when a subject rested on their knees and flattened palms with arms and thighs perpendicular to the floor and feet comfortably extended and spaced [201]. It should be noted that in each of the experiments, subjects utilized adjustable knee pads and gloves which would not be expected to be available under normal emergency conditions.

Table 64.23 Summary crawling speed data

Vulnerabilities: Innate: Impaired Movement

During the last three decades, there has been a growing effort to improve access to buildings for the entire population. It is now reasonably established in the regulation of developed countries that all new or refurbished public buildings should provide access for all and that existing buildings should make ‘reasonable’ provision. This improved access means that building populations are now more diverse and span a spectrum of movement abilities. The fast-changing anthropometric profile of occupants, as well as the large number of mobility-limiting diseases such as asthma and heart conditions, may have an impact upon issues relating to ingress and egress from a building. Although, the exact nature of this impact is not well understood at the moment, this still needs to be taken into account in fire safety design.

As noted previously, significant percentages of national and international populations have disabilities/impairments that impact on their ability to evacuate buildings [202]. These include those with physical and mental impairments, including the elderly and the obese. Since the prevalence of disability increases with age, and we are an increasingly aging society, it is expected that building populations will become even more diverse in terms of the physical and mental capabilities in the future. One potential impact on the engineering timeline is with respect to the t _trav component and it is therefore important to understand the range of capabilities of people with disabilities in terms of movement on both flat and inclined surfaces; i.e. stairs and ramps. It is also important to recognize that the percentage of those with mobility impairments may increase during an incident where injuries may occur leading to immediate limitations that cannot be immediately addressed through the presence of movement aids/devices. Limitations and impairments are certainly not confined to movement; for instance, sensory and cognitive impairments may also influence the performance of an evacuee. Those with visual impairments may experience difficulty in wayfinding, including locating and reading exit signage and indeed also may have reduced movement speeds. Those with hearing impairments may find it difficult to hear alarms, those with mental impairments may experience difficulty in recognizing alarms and understanding the need to evacuate, and those with limited dexterity may experience difficulty using particular types of door furniture to facilitate their escape. Although some [203–207] have researched the latter the body of research is limited and therefore this section focuses upon data specifically related to movement.

Although design and procedural measures are frequently implemented in multi-storey buildings to aid those who are unable or find movement (particularly stair movement) difficult e.g. use of refuges, elevators, assistive evacuation devices, buddy systems, it is clear that many people with disabilities will attempt to use stairs irrespective of their limitations. This was particularly noticeable in the WTC evacuation on 9/11, where the evacuation of mobility impaired occupants within the entire evacuating population was described as “slow and arduous” [145].

It is important to note that those with mobility impairments often use devices such as walking sticks/canes, crutches, rollators, frames or manual or electric wheelchairs to aid their movement and may also be assisted by friends/family. Therefore, in addition to expected lower movement speeds people with such impairments may often require more space (both in a stationary position and in movement) than a person evacuating without such aids. Clearly, if the routes that comprise the means of escape are not of sufficient width to facilitate passing in such cases, this may impact the movement of the entire evacuating population using that particular component. This was evident in the evacuation of WTC where it was reported that 51 % of the occupants of WTC1 and 33 % of the occupants of WTC2 indicated that injured and disabled people in the stairwell were a ‘constraint to evacuation’ [145], and where the evacuating population often reported having to wait until reaching the landings before being able to overtake slower moving individuals on stairs [208].

Design guidance regarding exit and stair sizing generally makes assumptions of optimum flows derived from data on largely able-bodied populations. It is therefore essential to understand the capabilities of those with disabilities in the movement component of means of escape, namely their speeds on the horizontal, on ramps and on stairs.

Summaries of the data related to both unassisted and assisted unimpeded speeds on the horizontal, ramps and stairs (ascent and descent) and door traversal speeds are given in Tables 64.24, 64.25, 64.26, 64.27, 64.28, and 64.29. Tables 64.24, 64.25, 64.26, 64.27, 64.28, and 64.29. include, in addition to descriptive statistics (mean, standard deviation and range) for each data set, the Source of the data as well as the conditions under which the measurements were made; i.e., Observational Conditions (Location, Nature of the study, Spatial Configuration of the space in which measurements were made, Participants and the Variables). Separate tables are provided according to whether the subject group was moving with assistance of another person or without assistance, since this may have affected the movement speed. It is important to note, however, that across the studies, the nature and level of assistance varied. In all tables the conditions under which the observations/measurements were made are noted. In most cases the research comprised of an experimental study involving persons moving as individuals within a building or space; however, in some studies the measurements related to individuals moving in a more general evacuation involving others. In the former, the speeds are presented as in the original papers as unimpeded movement speeds, while in the latter, although densities were described as low, there may have been some influence of others on an individual’s movement and so additional information is provided. Attention should be paid to the Additional Information columns in the tables which describe these conditions.

Table 64.24 Travel data—unassisted horizontal movement speeds (m/s) for people with disabilities

Table 64.25 Travel data—assisted horizontal movement speeds (m/s) for people with disabilities

Table 64.26 Travel data—unassisted/assisted movement speeds (m/s) for people with disabilities on ramps

Table 64.27 Travel data—unassisted movement speeds on stairs (m/s) for people with disabilities/elderly

Table 64.28 Travel data—assisted movement speeds on stairs (m/s) for people with disabilities/elderly

Table 64.29 Travel data—door traversal speeds for people with disabilities/elderly (unassisted and assisted)

Where known, the distances over which measurements were made are also noted (under Spatial Configuration); in the case of horizontal movement this is simply a distance covered; in the case of stairs, information is in the form of a distance (if available) or the number of storeys over which the measurements were made. The distance over which movement is measured is particularly important when choosing speed data since individuals with reduced mobility may be expected to experience some reduction in speed as they tire over longer distances, i.e. the development of fatigue. The study by Kuligowski [220] reported in Table 64.27 is the only study to consider the variation in speed achieved over subsequent portions of the escape route and reference should be made to the original source for further information.

It is also important to mention in this respect that the movement speeds presented in the tables do not generally include periods of rest, unless otherwise noted. The exception is the study by Hunt et al. [214] which did include stops in the calculation of movement speed. The study by Boyce et al. [203, 204] noted that 14 % of those with a ‘locomotion disability’ required at least one rest over the 50 m horizontal route. Analysis of the WTC evacuation on 9/11 [145, 208] also suggested that some individuals with impairments had to stop to rest for short periods of time. The need to rest is likely to be dependent upon the distance travelled and exertion required relative to the severity of one’s disability, although no studies have specifically looked at the impact of fatigue on the movement of people with disabilities over longer distances. The behavior of individuals with disabilities evacuating may also be important for evacuation modeling. For example, the Boyce et al. study [203, 204] noted that the majority of persons with disabilities sought support from handrails; this was particularly evident when ascending and descending stairs where 91 % and 94 % respectively of those moving unassisted utilized the hand-rail for support.²

The stair incline is also important in relation to stair movement and is described in Table 64.27 if reported in the original source or if it could be derived. The study by Fujiyama and Tyler [212] is interesting in that it is the only study presented here which compared the movement speeds for the same subject group across different stair geometries. Tables 64.24, 64.25, 64.26, and 64.27 also include a brief description of the floor covering if noted in the original source; different floor coverings may provide different resistances to movement and therefore may have an influence on speeds achieved, particularly for wheelchair users, and this should be taken into account when considering the data relative to the intended application. It should also be noted that different researchers have used different terminology to denote the subject groups, e.g., those who have difficulty walking in some studies are described as ‘mobility impaired’, in others as having a ‘locomotion disability’. The terminology used in the tables is simply that which was adopted in the source material and users should refer to the source for a fuller understanding of the definitions of these terms. Care should also be taken to note the instructions given to participants of the experimental studies as they began their movement. Instructions ranged from moving in a ‘prompt manner’ to ‘normal speed’ and ‘fast speed’. The nature of the instructions provided is important as it might have influenced the efforts of the participants and subsequently the speeds achieved.

It will be noted that few studies have been designed specifically to consider speeds of movement that can be achieved by those using different assistive techniques/devices (carry chairs, evacuation chairs, stretchers, drag mattress). In this respect the reader is directed to the studies of by Hunt et al. [214], Kuligowski et al. [21], and Lavender et al. [217]; the former study investigated speeds of assistive movement of highly trained personnel in a hospital environment; the latter the speeds which could be achieved by professional fire fighters in assisted escape. Data from these studies is presented in Table 64.18.

It is clear from Tables 64.24, 64.25, 64.26, and 64.27 that the abilities of people with disabilities and the elderly cover a wide spectrum with respect to horizontal and vertical movement. It is important in egress analysis to consider not only the variance in speed but also variation in spatial requirements of different individuals, since this may have a significant impact on flow performance produced. Indeed Boyce et al. [203, 204] has suggested that for analytical purposes, individuals should be categorized according to the mobility aid used and whether or not they require assistance so that potential interactions with the evacuating population can be fully realized.

The ability of individuals with disabilities and those assisting persons with disabilities to negotiate doors has also been studied by a number of researchers. Boyce et al. [203, 204], for example, examined the necessary time for people with disabilities to go through a door by pulling and pushing the door, which was subjected to a range of closing forces. The analysis of this data suggested that the ability of people with disabilities to negotiate doors subjected to a range of closing forces may depend on different factors including: the type of aid used (since it is implies a movement speed and particular technique in maneuvering the technical aid though the door), how old the participant was (since this is inherently related to strength) and the presence and severity of a dexterity or reaching and stretching disability. Hunt et al. [214] also investigated the time for teams of well-trained hospital staff to negotiate doors whilst handling different evacuation devices, i.e., a stretcher, rescue sheet and Evac+chair. They found that female handling teams took longer than male handling teams to maneuver through closed doors and found that generally it was easier to negotiate doors which opened away from the handlers. The data from both studies is presented in Table 64.25.

Identifying Applicable Data-Sets

Egress analysis is frequently presented to a third party for scrutiny as part of the performance-based design process (see Chap. 57). As mentioned previously, egress analysis is used to provide an estimate of the Required Safe Egress Time (t _RSET ). This is then compared against the Available Safe Egress Time (t _ASET ). Egress analysis is typically achieved through the use of engineering calculations or computational tools (see Chaps. 59 and 60), both of which require support from the use of data. In this process, one group of people perform the analysis while another group (e.g. AHJs, code enforcers, etc.) judge the approach adopted, including the relevance and credibility of the data employed. For this judgment to be fair, accurate and credible, a comprehensive and detailed description of the key elements involved is required—including the data employed. Just as this assessment is required in the presentation of any results to a third party, an equivalent description (and understanding) is required on the part of the engineer in initially selecting the data. This is not trivial; indeed, given the immature nature of the field, it can be problematic. Given this, it is important that the engineer poses a series of questions of the data-sets to be used to establish if the background information related to the data is to their satisfaction. The answers to these questions will then determine whether the data is appropriate for the current application and the same answers should also be reflected in the presentation of any results produced. The reader should also examine the material provided in Chap. 57, as the answers to these questions shown below may complement the material presented in that chapter.

Below is a list of questions that the engineer should consider when identifying data sets for use. Although not a definitive list, these questions should assist the engineer in the selection of the most appropriate data-sets. Indeed, these questions formed the basis of the table format used when presenting the various data-sets provided earlier in the chapter. The engineer should consider the answers to these questions whilst choosing, interpreting and using the data and the answers to these questions should also be reflected in the eventual presentation of the results to the third parties, i.e. the background information that places the data (and the egress results) into context.

What is the origin of the data? For instance, is it derived from:

• Non-emergency movement,

• Evacuation drills or unannounced evacuations

• Experimental work,

• A real incident,

• Survey?

Is this appropriate and valid for the current project?

What is the nature of this data? For instance, is the format

• Numerical: Raw/Compiled/Composite/Extrapolated,

• Descriptive: reports/anecdotes/journalism,

• Graphical,

• Function/Relationship-based,

• Primary/Secondary, and

Is the sample large/small, representative/unrepresentative?

Does the data format support the current application? If not, is the engineer able to derive the necessary data from it? Is the data sufficiently representative to be used in the current application?

Is this source considered appropriate, valid and reliable?

• Is it adequately documented?

• Who collected it?

• How was it collected?

• When was it collected?

• What data collection techniques were employed?

• What research methods were employed to collect and analyse the data?

• What was the observed scenario?

Is there enough known about the data and the original event scenario?

Population characteristics/performance issues

• How many people were involved in the event and in the sample?

• Who was involved?

• Were they familiar with the structure?

• Where was the population located?

• What were they doing at the time of the incident?

• Are there sub-populations of particular concern within the space e.g., an impaired population?

• What attributes/issues might influence the results produced?

  • Physical—age/gender/health/fatigue/impairment/weight/encumbrance/children/elderly

  • Psychological/Behavioral—exposure to information/familiarity/cognitive abilities/experience/motivation/status

  • Social—role/hierarchy/relationships (employment/social/familial)/responsibilities/(alone/groups/crowd)/affiliation/culture

  • Situational—activities/engagement/commitment/alertness/location/proximity to incident/intoxication

Procedural (organizational) characteristics

• What was the emergency procedure employed?

• Was the population aware that the incident was going to take place?

• What notification systems were in place?

• How many staff members were actively engaged in the emergency procedure?

• What was the nature of the message/information being provided?

• Were people expected to assemble and, if so, where were the assembly points?

• What non-emergency procedures were in place that might have influenced the effectiveness of the emergency procedure?

Structural characteristics

• What was the structural configuration?

  • Existence of egress routes, doors, etc.,

  • Number of floors,

  • Use of the structure,

  • Dimensions of the space and egress components.

Environmental characteristics

• What was the cause of the incident?

• How did the incident develop?

• Where was the incident located?

• Did it spread beyond the room of origin?

• What cues were present?

• Did this influence the availability of egress routes?

• Did the spaces have environmental pollution/noise (visual/aural) that might have influenced evacuee performance?

• Were the spaces cluttered/confined?

• What were the expected lighting levels?

• What impact did the environmental conditions have on the evacuating population?

  • Physiological,

  • Psychological,

  • Behavioral.

• Were there external conditions (e.g. weather conditions) that influenced the performance of the evacuating population?

It would be impossible in this chapter to provide this level of detailed information for all of the data sets that exist in the public domain. However, the tables are structured such that space is provided for the key elements for each type of data to be included (see section “Structure of Data Presentation”). The tables presented in this chapter should therefore allow the engineer to answer the majority of these questions, or at least establish, with additional reference to the original source material, where insufficient information is provided to answer them; i.e. the omission of information can be established.

Using the Data

It is acknowledged that the presentation and availability of the data is only a small part of the engineering process. Another key step is for the engineer to select an appropriate data-set (or more likely data-sets) for the scenario at hand. Given the nature of this scenario, it may be that the scenario conditions do not exactly match those of the data presented; i.e., that the engineer is required to potentially select and use several data-sets to represent a single real-world scenario. This might occur where the key factors in the scenario are not adequately addressed by a single data-set, where the data is not considered reliable enough, where the data is not adequately described and/or where the data is considered too old.

In addition, the engineer may also be required to manipulate single/combined data-sets in order to reflect specific aspects of the scenario, even where the original data-set is deemed to be an ideal match. The engineer might therefore be required to combine whole data-sets, splice parts of data-sets together, or manipulate the manner in which a data-set is applied across the area/population/timeline being examined. For this to occur the engineer would need to be able to match the data-sets against the scenario factors being represented and ensure that the manipulation of the data-sets adequately represented the real-world conditions while not undermining the original meaning of the data; i.e., the appropriateness of the original data selection.

As an example, a workflow of such data manipulation is shown in Fig. 64.21. This is intended to outline the process through which an engineer might pass and is but one of many that might occur. Stages (4, 7, 8) outline the initial definition of the quantitative component of the data-sets (i.e., the extent of certain factors), while Stages (5, 6, 9, 10) outline the qualitative aspects that help define the scenario and influence the manipulation of the eventual data-set generated (i.e., the nature of certain factors).

Fig. 64.21

Manipulation of multiple data-sets to fit the scenario

Initially, the engineer describes the scenario, S, given the real world situation at hand and the factors and conditions associated with it. For instance, the engineer has to calculate the egress time for a population in a hotel with a fire assumed to start in one of the rooms on an upper floor.

The engineer first outlines the scenario in question in as much detail as possible (Stage (1)). The engineer then (Stage (2)) determines whether specific data and supporting information is available to enable the examination of the scenario described, S. For instance, a previous incident may have been documented at that site or a number of evacuation trials may have been recorded from that structure (or similar structures). Given that this data is not available, the engineer consults the available guidance and literature (Stage (3)), to identify data-sets and information, L, that may relate to the scenario. For instance, the data presented in Table 64.4. In Stage (4), the quantitative data included in L is collated to produce a number of data-sets, E, that are deemed to relate to the scenario. Similarly in Stage (5), given the issues highlighted in L, a set of behavioral and procedural issues, I, are identified that need to be accounted for in the representation of the scenario—in the ‘model’ of the scenario.

In Stage (6), these issues, I, are coupled together—integrated to establish a schema describing how they might interact to produce effects during the simulated scenario. In other words, how do the issues identified combine to produce the modeled scenario of interest? This schema may take the form of a number of questions that guide the engineer in what might be expected. For instance, how might intoxication affect performance of guests in a hotel given the use of voice alarm? In Stage(7), given the data-sets, E, the real-world scenario, S and the schema describing expected performance, F, the data-sets are compiled such that the data-sets selected are clearly associated with key elements in the schema in order to represent the scenario, E*. Here, the interaction between key factors is quantified; for instance, the pre-evacuation times of the intoxicated, in a hotel given the use of voice alarm. Whereas E represents data-sets that address factors associated with the scenario, S, E* represents how these may be spliced and combined to specifically address aspects of the behavioral schema produced, F.

In Stage (8), the refined data-sets are manipulated such that a final distribution of data is produced for the factors highlighted, Q. This reflects the range of values that might represent the key factors and interactions in F. For instance, several functions are produced to represent the influence of intoxication and voice alarm in a hotel setting. In Stage (9), the data distribution, Q, is compared against the schema, F, to identify shortfalls in the quantitative representation of the schema—of the qualitative factors that are expected to influence performance—to generate a schema that is numerically represented to some degree, F^D. This represents the factors that will be depicted in the ‘model’ scenario. In Stage (10), this schema, F^D, is then applied across the space and time being represented in the scenario, S, to determine whether modifiers need to be applied, producing F*. For instance, whether proximity to the incident, alarm coverage/notification, etc., need to be taken into account when applying the data to different areas/times/populations during the scenario. Will someone in the room of origin respond in the same way as someone with no visual access to the incident given that they are both intoxicated, in a hotel, awake, etc.?

In Stage (11), the refined schema, F*, representing both the quantifiable issues and the manner in which is applied across the temporal/spatial environment is then applied to the data distribution, Q. This then produces a quantified definition of the scenario, S*, that is used to configure the model employed in the egress analysis.

Obviously, the process is over-simplified and presented in a linear manner when in fact it may be a highly iterative process. In reality, the engineer may iterate between these stages, ignore some and/or introduce others. However, the example is presented to demonstrate that the data presented may need significant manipulation both to broadly address the factors present in the scenario being addressed and then to customize the derived data-sets to represent local situations within the scenario. This will require skill and judgment. This process should be documented, not only to identify the data-sets and the steps involved, but also to make the assumptions and actions clear to third party viewers.

Summary

This chapter has presented a range of data-sets relative to the engineering time line commonly used to represent response and evacuation behaviors. Although these may inform a number of research, engineering and regulatory practices, the objective of this chapter has been to facilitate more effective, reliable and informed egress analysis as part of the performance-based design process. This data might be used to therefore to develop, configure, employ and validate computational and engineering egress tools as part of this effort.

Given the range of data types provided, the data have been provided in a number of tables reflecting the different evacuation phases, behavioral elements and influential factors involved. In each case, the data is accompanied by as much relevant background information as possible. Where space was an issue, background information has been abbreviated, and, in all instances by necessity, background information that would ideally have been included has been excluded due to space and time limitations.

The data-sets are not, therefore, provided in sufficient detail for the reader to make a definitive selection. Instead, the data and associated description should provide sufficient information for the reader to narrow down their review and focus on the most relevant data-sets for their particular application. This emphasizes the importance of the user following up on data-sets of interest by reference to the original sources. It is hoped that this approach should save them time and ensure that the most credible data is employed.