Evacuation modelling

Evacuation modelling is the analytical representation of how occupants respond and move towards a defined place of safety during an emergency. Models can calculate movement through a building and, depending on their scope, represent departure delays, route decisions and interactions. Nelson and Mowrer distinguish models intended to optimise evacuation, simulate observed behaviour or examine evacuation-related risk. These purposes require different assumptions, so the meaning of a predicted completion time depends on the model and scenario.

A hydraulic evacuation model provides one representation of movement through density, speed and route capacity. More detailed approaches can represent individual occupants and their decisions. Model selection must follow the assessment question, available evidence and required outputs. A more elaborate computer representation does not by itself establish that the predicted evacuation is realistic.

Define the question and the evacuation endpoint

An evacuation calculation needs an explicit starting event and destination. Movement to another protected area is a different task from complete discharge outside a building. The 2005 International Fire Engineering Guidelines require the extent and nature of movement to be established during the engineering brief. This includes whether the scenario concerns a zone, staged movement or full evacuation.

The population of interest also needs definition. Some models report the total number leaving, while others retain the identity and attributes of individual occupants. A total completion time may conceal the experience of a particular group. The required outputs therefore depend on whether the question concerns overall discharge, delayed occupants or movement through a specific route.

The physical route and the movement strategy belong to the same scenario. An assumed exit cannot carry occupants who have not been directed towards it or do not choose it. Likewise, a staged scenario changes when different groups enter shared circulation. Route allocation and departure sequence should therefore be treated as explicit assumptions.

Required safe egress time includes more than movement. Detection, response to cues and delay before departure can occupy substantial parts of the timeline. A model that calculates only walking and queuing provides only part of that assessment. The omitted periods require separate evidence or a supported representation within the model.

The fire scenario establishes conditions that can affect the available routes and the time during which they remain usable. Fire modelling can provide information relevant to that environment, while evacuation modelling represents occupants' response and movement. The assumptions connecting the two analyses must refer to the same building, scenario and timeline. Otherwise, apparently comparable results can describe different circumstances.

Model purpose and representation

Nelson and Mowrer's 2002 taxonomy separates optimisation, simulation and risk-assessment applications. Optimisation assumes efficient route use and flow, often with a homogeneous population. It can identify how a system might perform under favourable allocation. That purpose differs from predicting the choices people will actually make during a particular emergency.

Model purpose, spatial detail, population representation and behavioural treatment are separate selection questions. Based on Nelson and Mowrer (2002), pp. 3-377–3-378.

Simulation attempts to reproduce movements and decisions observed in evacuations. Models differ in how much behaviour they represent and in the evidence supporting that representation. A model can be detailed in one respect and simplified in another. The word simulation therefore does not specify its behavioural completeness or predictive accuracy.

The risk-assessment approach described in the chapter uses repeated calculations to examine variations and design changes. The repeated cases provide information about the consequences of differing assumptions or conditions. Their value depends on the cases and variations represented. Repetition alone cannot supply a missing physical or behavioural mechanism.

The enclosure can be represented by a coarse network of rooms and corridors connected through openings. A finer network subdivides floor space into many connected locations. This can represent obstacles and occupant positions in greater spatial detail. The chosen subdivision affects how the geometry and movement opportunities enter the calculation.

Population representation is a separate choice. A global model assigns distributed or average attributes to a group, while an individual model assigns attributes to particular occupants. The latter can represent diversity, subject to the implemented rules and supplied data. Individual representation does not automatically include realistic social interaction or group decisions.

These classifications describe the cited handbook's framework, rather than a current catalogue of software products. Their continuing analytical value is the distinction between purpose, geometry, population and behaviour. Each dimension should be examined when judging suitability for an application. Greater detail in one dimension cannot be assumed to compensate for an unsupported assumption in another.

Population, behaviour and departure

Human behaviour in fire influences how occupants interpret warnings and what they do before leaving. Movement equations alone do not determine these responses. Nelson and Mowrer discuss information seeking, other actions, wayfinding and uneven exit use as influences on evacuation efficiency. These mechanisms can alter both departure times and the routes receiving demand.

Some models contain no explicit behavioural rules. Others treat behaviour implicitly through selected inputs or empirical relationships. Rule-based models assign decisions according to defined conditions, sometimes with random variation. Identifying which of these representations is present helps distinguish a calculated response from a response prescribed by the analyst.

The source also discusses functional analogies and artificial-intelligence approaches within its historical classification. Those labels describe a means of representing decisions, not evidence that the decisions are valid for every fire. The relevant question remains how interactions and responses are represented for the intended application. Model documentation and supporting validation are therefore essential to interpretation.

Population attributes should reflect the occupants who can materially affect the assessment. The International Fire Engineering Guidelines recommend using the design occupant group's relevant movement characteristics. A typical able-bodied adult speed may not represent the group controlling completion. Adjusting a final time cannot necessarily reproduce the congestion or assistance consequences of different movement characteristics.

Departure timing also affects the demand on common routes. Simultaneous release can produce a different queue pattern from staggered departure. The hydraulic assumptions listed in the Guidelines explicitly include simultaneous starting and uninterrupted flow decisions. Those assumptions must be recognised when the calculated movement is combined with a more varied response scenario.

The ABCB's 2020 data sheets allow important management and behavioural effects to enter directly or through adjusted inputs and explicit delays. Either treatment needs a clear explanation of what is represented. A prescribed delay should not be described as a behavioural decision independently predicted by the model. Its evidential basis remains part of the analysis.

Movement, bottlenecks and route interactions

Movement models commonly relate speed to crowd density and route geometry. Specific flow describes throughput per unit effective width, while total flow also depends on width. Doorways, corridors and stairs can have different movement relationships. The relevant route conventions must be consistent with the equations and empirical data selected.

A route restriction creates a queue when arriving demand exceeds its discharge capacity. The excess population accumulates upstream instead of passing through the restriction immediately. A useful model must preserve this accounting and represent the resulting delays. Complex layouts can contain several interacting restrictions, making a single population-divided-by-capacity estimate insufficient.

At a merge, two incoming populations share a downstream route. Conservation determines the combined demand but does not uniquely determine priority. Nelson and Mowrer describe possible dominance by one stream and recommend examining bounding cases where appropriate. A model's merge rule can therefore influence which occupants wait longest, even when total discharge is similar.

Routes may also receive unequal demand because occupants favour particular exits. An optimised calculation can distribute people efficiently, while a behaviourally based scenario may not. These are different assumptions about use of the same physical capacity. The assessment must distinguish an available exit from an exit actually used by the represented population.

Smoke conditions can affect movement separately from crowding. The handbook discusses reductions associated with visibility and irritation, while many movement observations concern smoke-free conditions. A model using unimpeded speeds does not thereby demonstrate passage through a smoke-filled route. The expected environment and the model's response to it require explicit examination.

Worked timing example and its limits

The International Fire Engineering Guidelines divide the timeline into cue, response, delay and movement periods. Their notation makes clear which periods combine into pre-movement and total required time. The following expression reproduces that accounting convention. Other terminology should be mapped consistently before results from different methods are compared.

RSET = T_c + T_r + T_d + T_m

Symbol	Period in the cited convention	Unit
Tc	Cue period	s
Tr	Response period	s
Td	Delay period	s
Tm	Movement period	s
RSET	Total required safe egress time	s

For an illustrative single timeline, select 45 seconds for the cue period, 30 seconds for response and 75 seconds for delay. Let the represented movement take 120 seconds. Adding these periods gives 270 seconds. These selected values demonstrate the accounting and are not recommended response times for any occupancy.

A movement result is only one part of the selected complete timeline. The numerical inputs are illustrative. Based on Australian Building Codes Board (2005), p. 2.8-2.

Reporting only the movement model's 120 seconds would omit 150 seconds in this example. The omission changes the starting event associated with the reported duration. It is not a minor rounding difference. A result labelled evacuation time therefore needs an accompanying definition of the periods and events it includes.

If the selected delay increases by 60 seconds and movement remains unchanged, the total becomes 330 seconds. This is a simple comparison with all other terms fixed. It does not show that every real delay has an equal effect on final completion. In a connected population model, changing departure times can also change arrivals, queues and movement durations.

Different occupants can have different timelines. A global assessment must preserve the timing relevant to the required endpoint and the population under consideration. The longest individual component cannot automatically be combined with unrelated components from other occupants as though they formed one observed sequence. The representation of individuals or population groups determines what such a combination means.

Selection, verification and validation

Verification and validation address whether the chosen implementation and its application provide credible results for the task. The ABCB workflow begins by defining project requirements and selecting candidate models with suitable capabilities. It specifically includes the relevance of existing validation. A model's use in another context is not sufficient evidence for every proposed application.

Simple verification cases can check travel speeds, imposed delays and congested flow against the model documentation. These isolate behaviours that can become difficult to understand in a complete building simulation. A successful check establishes that the tested operation behaves as documented. It does not establish the realism of every input or the entire building scenario.

Model credibility depends on suitable capability, tested operation, relevant evidence and examination of input variations. Based on Australian Building Codes Board (2020), p. 284.

The data sheets then call for comparison with appropriate technical literature, drills and exercises in similar buildings, where available. Such evidence can support validation and calibration of the application. Its relevance depends on the population, geometry and conditions represented. A close fit to one selected case does not remove the need to examine other mechanisms important to the task.

Nelson and Mowrer's selection questions explicitly ask about queues, merging, pre-movement decisions and actions during movement. They also ask whether necessary inputs are available and whether outputs meet the assessment needs. These questions prevent selection based solely on visual detail or software complexity. A model can be unsuitable because it omits a required mechanism or because its inputs cannot be justified. The handbook also asks how validation has been reported, so an analyst can inspect the supporting evidence and the extent of the comparison.

Results should also be reviewed for unexpected behaviour. The ABCB notes that an unexpected change is not automatically an error, but requires investigation. It may reveal a genuine interaction, an input mistake or a limitation of the model. Understanding that response is more useful than accepting a plausible-looking final time without examination.

Sensitivity and interpretation of results

Sensitivity analysis examines how results change under justified input variations. In the ABCB's 2020 comparative framework, this includes checking whether plausible changes reverse the ranking of a proposed solution and its reference building. A single favourable calculation cannot establish that the comparison is stable. The variations must relate to assumptions capable of affecting the decision.

Departure distributions, route use and merge priority can affect different aspects of the result. One change may alter total completion, while another mainly changes waiting for a particular group. Required outputs should therefore extend beyond one final number when the assessment question demands it. Model selection and reporting are connected through that need.

The final account should state the purpose, population, routes, timing convention and mechanisms represented. It should identify important behaviour supplied as assumptions and the validation relevant to the application. Results can then be understood as consequences of the documented scenario. Their precision should not be mistaken for certainty about an actual emergency.

See also

• Evacuation
• Occupant load

References

• Australian Building Codes Board. (2005). International fire engineering guidelines (2005 ed.).
• Australian Building Codes Board. (2020). Data sheets: Fire Safety Verification Method (Version 1.1).
• Nelson, H. E., & Mowrer, F. W. (2002). Emergency movement. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-367–3-380). National Fire Protection Association.