Skip to content
Upcoming courses

Fire investigation: Difference between revisions

Tools
From Fire Safety Academy Wiki
← Older edit
VisualWikitext
Sanne (talk | contribs)
Bots
337 edits
Add diagrams fse-fire-investigation-v1.svg, fse-fire-investigation-v2.svg, fse-fire-investigation-v3.svg (level-1 visual plan)
Mark (talk | contribs)
Bots
284 edits
Revise against the FSE wiki writing guidelines: corrections after independent fact-check (revised by Claude Opus 5.5)
 
(2 intermediate revisions by the same user not shown)
Line 1: Line 1:
'''Fire investigation''' is the systematic examination of a fire incident to reconstruct what happened and evaluate the conditions and systems involved. It can address ignition, fire development, smoke movement, occupant exposure and the performance of protection systems. The analysis uses documentary, observational, experimental and modelling evidence while distinguishing established facts from inference.<ref name="Hopkins-xxxi">Hopkins, Schoenrock, and Budnick (2005), pp. xxxi–xxxii.</ref><ref name="Hall-5-65">Hall and Ahrens (2002), pp. 5-65–5-67.</ref>
'''Fire investigation''' is the examination of a fire after the event to establish where and how it started, how it developed and how the building, its fire protection and its occupants were affected. Writing on spontaneous ignition in the SFPE Handbook, Gray notes that fire prevention and fire cause investigation both need to identify the critical conditions for ignition, before and after the event respectively.<ref name="Gray-2-211">Gray (2002), p. 2-211.</ref> Hall and Ahrens list the reconstruction of major fires among the engineering analyses that require input data on fires.<ref name="HA-5-65">Hall and Ahrens (2002), p. 5-65.</ref> The Society of Fire Protection Engineers likewise counts post-fire investigation and analysis as part of fire protection engineering.<ref name="IFEG-US-0.3-1">National Research Council of Canada et al. (2005), Part 0 (United States), pp. 0.3-1–0.3-2.</ref>


A reconstruction connects evidence across time. It asks which events occurred, how the fire changed, which protective measures operated and what consequences followed. The answer must remain consistent with available information and with the limitations of the fire and behavioural models used.<ref name="Hopkins-xxxi" /><ref name="Notarianni-5-43">Notarianni (2002), pp. 5-43–5-44.</ref>
Fire investigation matters because its findings explain why fires start and why protection succeeds or fails, and they can lead to changes in how buildings are designed, built and regulated. The US investigation of the World Trade Center disaster of 2001, for example, was meant to serve as the basis for improvements in building practice and revisions to codes and standards.<ref name="NIST-xxxv">National Institute of Standards and Technology (2005), pp. xxxv–xxxvi.</ref> Investigations depend on the evidence that survives, and large fires destroy much of it. Fire loss statistics from many sources show an unknown cause more often in fires with multimillion-dollar losses than in smaller fires, probably because more evidence is lost.<ref name="Gray-2-211" />


Fire investigation does not produce certainty merely by using detailed software or extensive records. Information can be missing, contradictory or unrepresentative, and models approximate real processes. Those limitations should be recorded alongside the reconstructed sequence.<ref name="Hopkins-xxxi" /><ref name="Hall-5-70">Hall and Ahrens (2002), pp. 5-69–5-72.</ref>
== Purpose and scope ==


== Purpose and investigative questions ==
Investigations differ in purpose and authority. The US National Institute of Standards and Technology (NIST) investigated the World Trade Center disaster to determine why and how the buildings collapsed. It also asked why injuries and fatalities were high or low depending on location, and examined the procedures used in design, construction, operation and maintenance and the parts of codes and standards that warranted revision.<ref name="NIST-xxxv" /> NIST is a non-regulatory agency whose investigations focus on fact finding. It has no authority to find fault or negligence, and its reports may not be used in lawsuits for damages.<ref name="NIST-xxxv" />


The purpose is to develop an evidence-based account of an incident. That account can explain the initiating event, growth, spread, exposure and response. It can also test whether installed protection performed as intended and which conditions affected its performance.<ref name="Hopkins-xxxi" /><ref name="Hopkins-xxxii">Hopkins, Schoenrock, and Budnick (2005), p. xxxii.</ref>
Other investigations can have legal consequences. Van Rossum notes that an investigation after a fire disaster may show that the cause lies in demonstrably wrong advice, below the knowledge and skill that could be expected at the time. In that case the adviser can be prosecuted.<ref name="VR-2.5">Van Rossum (2027), §2.5.</ref>


Investigation starts by defining the questions to be answered. NIST's WTC suppression study documented design, installation, water supply, differences among buildings, expected operation and incident performance. This structure separated normal capability from what happened during the event.<ref name="Hopkins-xxxii" />
For the World Trade Center buildings, the New York City fire department provided NIST with 112 records of its Bureau of Fire Investigation from 1977 to 2001. It also provided 397 operational fire reports from 1970 to 2001, all on standardised forms. NIST's review of earlier fires included, for example, a fire on 14 February 1975 on the 11th floor of WTC 1, one of the two towers. It spread through unprotected floor openings in utility closets and caused damage from the 10th to the 19th floor.<ref name="Kuligowski-1">Kuligowski et al. (2005), p. 1.</ref>


A broad question can conceal several smaller evidence problems. Establishing ignition is different from reconstructing later spread, and evaluating sprinkler performance is different from assessing occupant response. Each question needs its own sources and acceptance logic.<ref name="Hall-5-65" /><ref name="Proulx-3-349">Proulx (2002), pp. 3-348–3-350.</ref>
Large investigations break the work into specific questions. For the sprinklers, standpipes and hoses of the towers, the NIST study first documented their design and installation and compared them with the codes and standards of record. It then estimated their capacity under normal conditions with hydraulic calculations.<ref name="Hopkins-xxxii">Hopkins et al. (2005), p. xxxii.</ref> Only then did it ask what happened to the systems at each major event, how their performance was affected and at what point in the sequence they were lost.<ref name="Hopkins-xliii">Hopkins et al. (2005), p. xliii.</ref>
[[File:fse-fire-investigation-v1.svg|thumb|upright=1.5|alt=Questions guide evidence collection, reconstruction, modelling and findings with uncertainty.|Fire investigation reconstructs events from evidence and explicit inference. Based on Hopkins et al. (2005), pp. xxxi–xxxii and Hall and Ahrens (2002), pp. 5-65–5-67.]]


[[Fire statistics]] can provide context about similar incidents and reporting patterns. They can indicate what is common or unusual and help avoid excessive focus on a unique feature. They cannot determine what happened in the particular incident without case-specific evidence.<ref name="Hall-5-65" /><ref name="Watts-5-155">Watts and Hall (2002a), pp. 5-154–5-156.</ref>
== Evidence and how it is gathered ==


The investigation should preserve the difference between expected and actual performance. A system designed to control a specified fire may face another fire or an impaired supply. Its name or code classification alone cannot establish its behaviour in the incident.<ref name="Hopkins-xli">Hopkins, Schoenrock, and Budnick (2005), pp. xli–xlii.</ref><ref name="NUREG-10-2">U.S. Nuclear Regulatory Commission (2005), p. 10-2.</ref>
Fire investigators normally read the path and intensity of a fire from telltale signs left in the building. In the World Trade Center towers this was impossible because neither building remained, so NIST relied on what may have been the most photographed disaster in history.<ref name="Gann-xxxviii">Gann, Hamins, et al. (2005), p. xxxviii.</ref> In each photograph and video frame, every window was coded for fire, smoke and whether the window was still in place, which produced detailed timelines of fire spread.<ref name="Gann-xxxviii" /> NIST also reviewed tens of thousands of pages of documents, interviewed over a thousand people and analysed 236 pieces of steel recovered from the wreckage.<ref name="NIST-xxxvi">National Institute of Standards and Technology (2005), p. xxxvi.</ref>


== Evidence collection and reconstruction ==
Time links the evidence together. NIST describes time as the unifying factor in combining photographs, videos, survivor accounts, emergency calls and communications among emergency responders.<ref name="NIST-82">National Institute of Standards and Technology (2005), p. 82.</ref> Images were timed by matching them to one well-defined event, the impact of the second aircraft at 9:02:59 a.m., and then to other events visible in several images. In all, 3,032 photographs and 2,673 video clips were timed to within 3 seconds or better, and the uncertainty of each time was logged.<ref name="NIST-83">National Institute of Standards and Technology (2005), pp. 83–84.</ref>


Documentary evidence can include drawings, specifications, procedures, maintenance records and design criteria. NIST used such records to reconstruct the fire-suppression arrangements in WTC 1, 2 and 7. Open literature and inquiries supplemented the collected material.<ref name="Hopkins-xxxii" />
Physical evidence needs careful interpretation. NIST found that the primer paint on the towers' steel cracked in a "mud cracking" pattern above 250 °C. Deformation and the environment can also cause such cracks, so their absence shows that the steel stayed below 250 °C, but their presence does not prove high temperature.<ref name="Gayle-94">Gayle et al. (2005), p. 94.</ref> The method could not tell whether the heating happened before or after the collapse. The panels examined were also only 3% of the perimeter columns on the fire floors and could not be treated as representative of the others.<ref name="Gayle-xli">Gayle et al. (2005), p. xli.</ref>
[[File:fse-fire-investigation-v1.svg|thumb|upright=1.5|alt=Questions guide evidence collection, reconstruction, modelling and findings with uncertainty.|Fire investigation reconstructs events from evidence and explicit inference. Based on Hopkins, Schoenrock, and Budnick (2005), pp. xxxi–xxxii and Hall and Ahrens (2002), pp. 5-65–5-67.]]


System reconstruction identifies what was installed and how it was intended to operate. NIST reconstructed sprinklers, standpipes, pre-connected hoses and their water supplies before analysing performance. That step establishes the physical basis for evaluating the incident.<ref name="Hopkins-xxxiv">Hopkins, Schoenrock, and Budnick (2005), pp. xxxiv–xxxv.</ref>
People are a source of evidence as well. For the evacuation of the towers, NIST gathered over 1,000 new interviews with survivors, over 700 published interviews, emergency calls and transcripts of emergency communications.<ref name="Averill-iii">Averill et al. (2005), p. iii.</ref> Proulx notes that fire victims may have difficulty reporting accurately how long they waited before starting to evacuate, but that interview techniques can give acceptable estimates.<ref name="Proulx-3-349">Proulx (2002), p. 3-349.</ref> The causes of death in fires have been studied through pathological investigations of fire victims.<ref name="Purser-2-85">Purser (2002), p. 2-85.</ref> Carbon monoxide combines with haemoglobin in the blood to form carboxyhaemoglobin. In the Strathclyde study of fire deaths in the United Kingdom, 54% of those who died had a lethal level of it, and 69% had a level capable of causing incapacitation.<ref name="Purser-2-100">Purser (2002), p. 2-100.</ref>


Incident evidence can be fragmented. NIST reports very little eyewitness or communications information about suppression performance, so estimates relied on accumulated incident information, likely initial effects and historical performance records. The method and its limitations must therefore be explicit.<ref name="Hopkins-xxxi" />
== Establishing the cause ==


Where possible, significant events should be supported by more than one source. NIST notes that this could not always be accomplished because information was fragmented. A single uncorroborated account should not acquire the same weight as a conclusion supported by independent sources.<ref name="Hopkins-xxxi" />
To describe how a fire started, investigators separate the parts of the ignition sequence. US fire incident coding distinguishes the form of heat of ignition and the equipment involved, which together describe the [[ignition source]], from the form and type of material first ignited. It also records the ignition factor that brought heat and fuel together, and the area of origin.<ref name="HA-5-68">Hall and Ahrens (2002), p. 5-68.</ref> The New York fire department's structural fire form recorded the same kind of information, and it called the ignition factor the cause.<ref name="Kuligowski-2">Kuligowski et al. (2005), p. 2.</ref>


[[Fire investigation]] can use historical data to test plausibility, but it must not replace missing incident facts with generic expectations. NIST distinguished reasonably accurate descriptions of system capability from estimates of performance on the incident date. That distinction is essential to transparent reconstruction.<ref name="Hopkins-xxxi" />
Some causes leave characteristic traces. Spontaneous ignition of stacked material, such as cotton, woodchips or oily rags, starts in the well-insulated centre of the pile. Such a pile can therefore be charred or burnt to ash inside while looking almost untouched outside.<ref name="Gray-2-212">Gray (2002), p. 2-212.</ref><ref name="Gray-2-224">Gray (2002), p. 2-224.</ref> Hot gases escape along the path of least resistance and form "chimneys" of charred material, and several chimneys are the norm. Gray reports that finding more than one heavily charred area has led to wrongful charges of arson, based on the myth that more than one seat of fire means the fire was started deliberately.<ref name="Gray-2-225">Gray (2002), p. 2-225.</ref>


== Ignition and initiating conditions ==
Gray lists the practical factors that make spontaneous ignition more likely, including a large body of material, high surrounding temperatures, thermal insulation, porous or fibrous material and a high stacking temperature.<ref name="Gray-2-224" /> Material left undisturbed for longer than usual is also reason for suspicion. In one case a warehouse of cottonseed meal burned down after a transport strike had left the meal in place far longer than normal.<ref name="Gray-2-225" /> In another, an instant-noodle factory burned down after new management sped up the production line, so that the noodles were packed hotter than before. Laboratory tests and calculations then showed that the packed pallets could ignite spontaneously at the new temperatures but not at the old ones.<ref name="Gray-2-226">Gray (2002), pp. 2-225–2-226.</ref>


[[Ignition source]] analysis addresses the energy and conditions that initiated combustion. The assessment must distinguish the heat source, the first fuel and the factor that brought them together. NFIRS coding similarly separates these fields rather than treating them as one cause label.<ref name="Hall-5-67">Hall and Ahrens (2002), pp. 5-67–5-69.</ref>
Gray advises caution about the timing of such fires without detailed knowledge of the initial conditions.<ref name="Gray-2-224" /> Whenever spontaneous ignition is suspected, Gray recommends excavating the material, with photographs and temperature probe records at all stages. He has measured temperatures as high as 200 °C in buried material more than two weeks after a fire was put out. Unburnt oily rags at the bottom of a bin do not rule out spontaneous ignition either.<ref name="Gray-2-225" />


Incident records can contain ambiguous or missing cause information. A material code may not identify the exact product, and unknown fields can change apparent cause distributions. These limitations should be considered when comparing the incident with other cases.<ref name="Hall-5-68">Hall and Ahrens (2002), pp. 5-67–5-69.</ref>
== Reconstructing the fire ==


The initiating condition can include equipment, process and building states. NIST examined initial structural and compartment damage when evaluating suppression infrastructure. A damaged riser or supply can change protection before the developing fire reaches later stages.<ref name="Hopkins-xli" />
A reconstruction places the evidence in the sequence of the fire. Van Rossum divides a room fire into three phases. In the development phase, a local fire of one or a few objects grows until flashover. After flashover the fire is fully developed, and the decay phase can begin, for example, when the fuel runs out or the fire brigade intervenes. These [[fire development stages]] matter for interpretation, because an active suppression system normally acts during the development phase and so prevents the fire from becoming fully developed.<ref name="VR-4.11">Van Rossum (2027), §4.11.</ref>


A plausible ignition account must also be consistent with subsequent development. It should explain the observed sequence rather than merely identify a possible ignition source. Contradictions between the proposed source and later fire behaviour require further examination.<ref name="Watts-5-157">Watts and Hall (2002a), pp. 5-157–5-158.</ref><ref name="Notarianni-5-43" />
[[Fire dynamics]], defined in a US nuclear regulatory report as the scientific study of hostile fires, describes ignition, flame spread, fire growth, smoke movement and their effects in an enclosure.<ref name="Iqbal-1-1">Iqbal and Salley (2004), p. 1-1.</ref> Smoke rises under buoyancy to the upper part of the room of origin and leaves through openings, driven mainly by pressure differences. The International Fire Engineering Guidelines (2005 edition) note that hand calculations of its spread beyond the room of origin are more difficult and more uncertain than those within it.<ref name="IFEG-2.5-2">National Research Council of Canada et al. (2005), p. 2.5-2.</ref><ref name="IFEG-2.5-6">National Research Council of Canada et al. (2005), p. 2.5-6.</ref>


No source in the supplied set establishes one universal origin-and-cause protocol. The examples demonstrate reconstruction and evaluation of fire-protection systems rather than a complete legal or forensic methodology. That limitation should not be concealed by claiming a standard procedure where none is supported.<ref name="Hopkins-xxxi" /><ref name="Hall-5-65" />
Tests can fill gaps in the evidence. NIST set out to characterise the type, mass and layout of the combustibles in the towers, the fire endurance of the partitions and the properties of the insulation on the steel.<ref name="Gann-2">Gann, Hamins, et al. (2005), p. 2.</ref> Its reconstruction relied heavily on computer simulation, because the premises could not be examined after the fire and eyewitness information was severely limited. To support the simulations, NIST burned six single office workstations, the dominant combustibles in the buildings, to learn how they may have burned and to improve its fire model.<ref name="Ohlemiller-iii">Ohlemiller et al. (2005), p. iii.</ref>
 
== Fire development and smoke ==
 
[[Fire development stages]] provide a framework for organising the incident timeline. A local fire, growing fire, fully developed fire and decay present different conditions for people and systems. The investigation should locate evidence within that changing sequence rather than treating the fire as static.<ref name="VR-4-11">Van Rossum (2027), §4.11.</ref><ref name="IFEG-2-5-2">Australian Building Codes Board et al. (2005), pp. 2.5-1–2.5-2.</ref>
[[File:fse-fire-investigation-v2.svg|thumb|upright=1.5|alt=Pre-incident conditions, ignition, detection, fire growth and consequences, with fire, water-supply and occupant tracks.|An investigation must align system states with the incident timeline. Based on Hopkins, Schoenrock, and Budnick (2005), pp. xxxi–xlii and Van Rossum (2027), §4.11.]]
 
Plume rise and entrainment supply hot combustion products to the upper part of an enclosure. Ceiling jets and openings can then transport smoke to other spaces. These processes influence detection, visibility and exposure away from the fire's initial location.<ref name="IFEG-2-5-2" /><ref name="IFEG-2-5-6">Australian Building Codes Board et al. (2005), p. 2.5-6.</ref>
 
[[Fire dynamics]] links heat release, flame spread, enclosure conditions and smoke production. A reconstruction should examine whether the proposed fuel and ventilation conditions explain those observations. A single temperature or fire-size estimate cannot describe the whole event.<ref name="NUREG-1-1">U.S. Nuclear Regulatory Commission (2005), pp. 1-1–1-2.</ref><ref name="IFEG-2-5-2" />
 
Smoke-layer height and movement depend on time, fire size, exhaust and building connections. IFEG notes increased uncertainty for spread beyond the enclosure of origin. Reconstructing remote exposure therefore requires particular care with flow paths and changing pressures.<ref name="IFEG-2-5-6" /><ref name="IFEG-2-5-4">Australian Building Codes Board et al. (2005), p. 2.5-4.</ref>
 
Sprinklers and ventilation can alter smoke conditions. Cooling and mixing can reduce buoyancy and redistribute smoke, while early sprays can affect later heat-sensitive elements. Those interactions should be included when interpreting the timing of alarms, vents and subsequent fire spread.<ref name="Evans-70">Evans et al. (2005), p. 70.</ref><ref name="Cooper-3-239">Cooper (2002), p. 3-239.</ref>
 
== Systems and protection performance ==
 
Evaluating a suppression system begins with its intended design and supply. NIST established baseline hydraulic capability before considering incident damage. That made it possible to distinguish a system with capacity under normal conditions from one degraded by the event.<ref name="Hopkins-xli" />
 
Fire department connections, tanks, pumps and municipal supplies can provide redundancy, but they may share failure paths. The investigation should test whether each provision was available at the required time. A connection on the drawing is not proof of supply during the incident.<ref name="Hopkins-xxxv">Hopkins, Schoenrock, and Budnick (2005), p. xxxv.</ref><ref name="Hopkins-xli" />
 
Detection and alarm performance includes smoke production, transport, detector entry and signal processing. The FDS technical guide notes that early smoke processes and detector algorithms are difficult to model, and that bulk smoke transport alone does not describe detector response. Activation estimates need corresponding uncertainty.<ref name="FDS-107">McGrattan et al. (2026), p. 107.</ref><ref name="Schifiliti-4-23">Schifiliti, Meacham, and Custer (2002), p. 4-23.</ref>
 
Evacuation analysis requires more than route capacity. Occupants investigate, gather others and choose familiar routes, creating delays and different demands. Proulx's studies show why actual accounts should be interpreted with warning, staff, building and population context.<ref name="Proulx-3-349" /><ref name="Proulx-3-350">Proulx (2002), p. 3-350.</ref>
 
The investigation should evaluate interactions between protection systems. Sprinkler failure can change smoke-control capacity, and detection failure can change valve or fan operation. ABCB identifies these as common-mode issues to be considered in active-system assessment.<ref name="ABCB-131">Australian Building Codes Board (2020), p. 131.</ref>


== Modelling and computational fluid dynamics ==
== Modelling and computational fluid dynamics ==


[[Computational fluid dynamics (CFD)]] can estimate spatially varying gas velocity, temperature and smoke concentration. It can help test whether a proposed fire and ventilation scenario produces conditions consistent with evidence. It remains a model, not an observation of the incident.<ref name="IFEG-2-5-1">Australian Building Codes Board et al. (2005), pp. 2.5-1–2.5-2.</ref><ref name="Zalosh-3-414">Zalosh (2002), pp. 3-414–3-415.</ref>
[[Computational fluid dynamics]] (CFD) studies the flow of gases and liquids by solving the flow equations numerically on a grid of cells. A CFD model can show smoke spread, gas temperatures and velocities, flows through openings and the time to flashover in three dimensions as they change over time.<ref name="VR-11.1">Van Rossum (2027), §11.1.</ref><ref name="VR-11.2">Van Rossum (2027), §11.2.</ref> Turbulent flows, which are typical of fires, can only be approximated. CFD therefore gives good qualitative insight, but its exact results must be interpreted carefully and critically.<ref name="VR-11.1" /> Zone models such as CFAST, released by NIST,<ref name="VR-9.5">Van Rossum (2027), §9.5.</ref> are also used internationally for fire safety design and post-fire reconstruction.<ref name="Peacock-3">Peacock et al. (2026), p. 3.</ref>


CFD can be useful where geometry and flow paths make simple zone approximations unsuitable. Zalosh describes its use for spatially varying explosion and venting problems, including turbulence and equipment effects. Those advantages come with greater input and validation demands.<ref name="Zalosh-3-414" />
In the World Trade Center investigation, the unusual mix of available information led to a high reliance on computational tools, and NIST upgraded its CFD model, the Fire Dynamics Simulator, for the reconstruction.<ref name="Gann-2" /><ref name="Gann-4">Gann, Hamins, et al. (2005), p. 4.</ref> It varied the uncertain inputs to find those to which the results were most sensitive and those that agreed best with the photographs.<ref name="Gann-2-3">Gann, Hamins, et al. (2005), pp. 2–3.</ref> The accuracy of the predictions was judged by their consistency with the photographic evidence.<ref name="Gann-4" /> For each tower it ran a less severe and a more severe case, which gave some understanding of the uncertainty in the predictions.<ref name="NIST-81">National Institute of Standards and Technology (2005), pp. 81–82.</ref>
 
Input selection can determine the result. Fire growth, fuel properties, openings, boundary conditions and detector models must be justified from evidence. NIST selected hydraulic criteria and fire areas for its baseline analyses and documented those assumptions before drawing conclusions.<ref name="Hopkins-xli" />
[[File:fse-fire-investigation-v3.svg|thumb|upright=1.5|alt=Two candidate scenarios are tested against observations and model limitations before conclusions.|Alternative reconstructions should be tested against evidence and sensitivity. Based on Notarianni (2002), pp. 5-43–5-44 and McGrattan et al. (2026), p. 107.]]
[[File:fse-fire-investigation-v3.svg|thumb|upright=1.5|alt=Two candidate scenarios are tested against observations and model limitations before conclusions.|Alternative reconstructions should be tested against evidence and sensitivity. Based on Notarianni (2002), pp. 5-43–5-44 and McGrattan et al. (2026), p. 107.]]


Models used for reconstruction should be checked against experimental and incident evidence. FDS documentation distinguishes predicted thermal environments from empirical detector response models. Bulk transport calculations do not automatically reproduce device activation or the conditions experienced by people.<ref name="FDS-107" /><ref name="Notarianni-5-43" />
Models also have limits. NIST decided not to model smoke movement and survival conditions throughout the occupied parts of the towers. There was no quantitative information on conditions inside the towers and no verifiable information on damage between floors, so such simulations for the impact floors and above would have been grossly uncertain.<ref name="Gann-3">Gann, Hamins, et al. (2005), p. 3.</ref>
 
Alternative plausible inputs should be examined. Notarianni describes how reasonable changes in criteria, scenarios and assumptions can change acceptability. A reconstruction should therefore show whether its conclusion is sensitive to uncertain fire size, timing or occupant behaviour.<ref name="Notarianni-5-43" />
 
== Statistics and comparative evidence ==
 
Incident databases can reveal patterns in ignition, occupancy, equipment and protection performance. They can help assess whether a proposed sequence is unusual or consistent with other events. Their definitions and collection practices must match the comparison being made.<ref name="Hall-5-67" /><ref name="Hall-5-68" />
 
A database may contain many records but still be unrepresentative of the target. Reporting can favour severe incidents, omit small fires or vary by region and organisation. The investigation must not treat a convenient sample as a complete description of all fires.<ref name="Hall-5-70" />
 
Anecdotal case reports can demonstrate possible mechanisms and provide detail not available in routine coding. They cannot estimate the frequency of that mechanism or prove it is typical. A detailed WTC report describes that event and is not a generic high-rise fire model.<ref name="Hall-5-70" /><ref name="Hopkins-xxxi" />
 
Laboratory and field evidence provide different checks. Tests can examine material or system response under controlled conditions, while field observations reveal degradation and actual use. Their results should be compared for consistency rather than assuming one form is always superior.<ref name="Hall-5-65" /><ref name="Hall-5-72">Hall and Ahrens (2002), pp. 5-71–5-73.</ref>
 
Unknown data require explicit handling. Proportional allocation can be used in national estimates, but it assumes unknown records share the distribution of known records. That assumption may not apply to the incident category under investigation.<ref name="Hall-5-68" /><ref name="Hall-5-69" />
 
== Uncertainty, inference and reporting ==
 
An investigation report should separate direct observations from deductions. NIST included events or effects considered likely or probable based on accumulated information. This makes the inferential status visible rather than presenting every statement as equally established.<ref name="Hopkins-xxxi" />
 
The report should identify which conclusions could be corroborated. NIST sought more than one source for significant events where possible, but fragmentation sometimes prevented it. Missing corroboration is a limitation that belongs in the findings.<ref name="Hopkins-xxxi" />


Model limitations should be described alongside numerical results. A hydraulic estimate may have a clear criterion, while smoke detection and human response have larger uncertainty. Presenting all outputs with the same precision would misrepresent those differences.<ref name="Hopkins-xli" /><ref name="FDS-107" />
The Fire Dynamics Simulator predicts the thermal environment of a fire, but it relies on empirical models for the activation of sprinklers and detectors.<ref name="FDS-107">McGrattan et al. (2026), p. 107.</ref> Smoke detector response is especially hard to model, because the production and transport of smoke early in a fire are not well understood.<ref name="FDS-108">McGrattan et al. (2026), p. 108.</ref> Notarianni adds that fire model outputs are single values that do not reflect the uncertainty in inputs such as fire growth rates. A predicted activation time of 121 s for a detector, for example, may in reality be longer or shorter, depending on uncertain inputs and on factors the model does not include.<ref name="Notarianni-5-43">Notarianni (2002), p. 5-43.</ref> In forensic reconstructions, the amount of soot deposited on surfaces in a simulation can be correlated with what is observed after the fire.<ref name="FDS-87">McGrattan et al. (2026), p. 87.</ref>


The supplied documents do not support a universal claim of legal proof or a single origin-and-cause methodology. They support reconstruction, comparison, physical modelling and explicit uncertainty. Findings should remain within those technical boundaries.<ref name="Hopkins-xxxi" /><ref name="Hall-5-70" />
== Evaluating protection systems ==


The final reconstruction should be internally consistent and open to review. It should identify the evidence, alternative explanations, model assumptions and remaining gaps. That allows another investigator to assess whether the proposed sequence is supported rather than merely plausible.<ref name="Hopkins-xxxi" /><ref name="Notarianni-5-43" />
An investigation compares what a system could do with what it actually did. NIST found that the sprinkler and standpipe systems of the towers had for the most part exceeded the code requirements of their time, but the adverse effects of the aircraft impacts on them were extensive.<ref name="Hopkins-xlv">Hopkins et al. (2005), p. xlv.</ref> WTC 1 was the first tower to be struck.<ref name="Hopkins-xxxi">Hopkins et al. (2005), p. xxxi.</ref> Its estimated impact damage would probably have destroyed at least one standpipe riser and the sprinkler systems on the 94th to 96th floors, and cut the water supply to other systems after a limited time.<ref name="Hopkins-xliii" /><ref name="Hopkins-xliv">Hopkins et al. (2005), p. xliv.</ref> The collapse of WTC 2 damaged the city water mains near the site, and the collapse of WTC 1 probably caused significant further damage.<ref name="Hopkins-xliv" />
[[File:fse-fire-investigation-v2.svg|thumb|upright=1.5|alt=Pre-incident conditions, ignition, detection, fire growth and consequences, with fire, water-supply and occupant tracks.|An investigation must align system states with the incident timeline. Based on Hopkins et al. (2005), pp. xliii–xlv and Van Rossum (2027), §4.11.]]


A reconstruction should use a timeline to organise evidence. Design and installation information belongs to the pre-incident condition, while alarm, suppression, occupant and fire-service actions occur in sequence. Mixing those periods can make a system appear to have failed before it was expected to operate or to have operated before the necessary conditions existed.<ref name="Hopkins-xxxii" /><ref name="Proulx-3-349" />
The evidence on systems is often incomplete. For the towers' suppression systems there was very little eyewitness or communications information. NIST therefore based its estimates on the incident information it had gathered, the likely effects of the impacts and historical performance records for sprinklers.<ref name="Hopkins-xxxi" /> Most of the New York fire department's reports of significant earlier fires documented the performance of detectors and sprinklers, but almost all lacked information on the smoke purge system.<ref name="Kuligowski-2" />


Evidence should be checked for internal consistency. A proposed fire size should be compatible with fuel, observed spread and system response, and a proposed activation time should be compatible with transport and detector behaviour. Inconsistency is a reason to revisit the scenario rather than discard inconvenient evidence.<ref name="Notarianni-5-43" /><ref name="FDS-107" />
Which sprinklers opened, and when, also needs careful reading. Water droplets carried by the hot gas flow under the ceiling from the first sprinklers can cool the heat-sensitive links of sprinklers further away and delay their opening. Cooper calls this random effect sprinkler skipping, and fusible-link vents can be delayed in the same way.<ref name="Cooper-3-239">Cooper (2002b), p. 3-239.</ref> Evidence from victims has comparable limits. Carboxyhaemoglobin is routinely measured in the blood of fire victims, but blood cyanide often is not, and post-fire cyanide levels are hard to relate to the exposure during the fire.<ref name="Purser-2-100" /><ref name="Purser-2-103">Purser (2002), p. 2-103.</ref>


The distinction between code compliance and incident performance should remain clear. NIST compared installed arrangements with standards of record before analysing likely damage and subsequent capability. A system can meet an installation requirement yet fail to deliver its intended function after physical damage.<ref name="Hopkins-xli" /><ref name="Hopkins-xlii">Hopkins, Schoenrock, and Budnick (2005), pp. xli–xlii.</ref>
== Limits and use of findings ==


Uncertainty can concern both evidence and meaning. A record may be incomplete, and a code such as first material or equipment may have a different interpretation from the investigator's preferred label. Those limitations should be stated before comparative statistics are used to support a conclusion.<ref name="Hall-5-69">Hall and Ahrens (2002), pp. 5-68–5-70.</ref>
Investigators often work with fragmented information. NIST confirmed significant events from more than one source where possible, but this could not always be done, and it included events that it judged likely or probable from the accumulated information.<ref name="Hopkins-xxxi" /> Records can also be lost. The Port Authority of New York and New Jersey kept records of all significant fire events in the World Trade Center buildings, and these were lost in the collapse of the towers.<ref name="Kuligowski-1" /> Physical evidence may cover only a small and unrepresentative part of a building, as the recovered steel did.<ref name="Gayle-xli" />


A useful sensitivity analysis varies the assumptions that could change the sequence. Fire growth, initial damage, activation times and occupant response may all be uncertain. If several combinations remain plausible, the report should present them rather than choose one because it produces a simpler narrative.<ref name="Notarianni-5-43" />
A single investigation describes one event. Hall and Ahrens note that detailed accounts, including NFPA's reports of fire investigations, show the range and diversity of fires that can happen. Such anecdotes cannot be used to estimate incident rates or probabilities.<ref name="HA-5-70">Hall and Ahrens (2002), p. 5-70.</ref> Valid analysis of incident rates instead requires a statistically representative database of [[fire statistics]].<ref name="HA-5-69">Hall and Ahrens (2002), p. 5-69.</ref>


The completed report should preserve the provenance of each important input. It should identify the document, observation, test or model supporting it and explain any expert judgement used to bridge gaps. This allows later evidence to be incorporated without confusing new information with what was known in the original analysis. New evidence should trigger a fresh consistency check of the complete reconstruction.<ref name="Hopkins-xxxi" /><ref name="Hall-5-65" />
Collected investigation data can in turn feed design values. Holborn and colleagues analysed data from fire investigations in the Greater London area between 1996 and 2000 to derive fire sizes, fire growth rates and the times between events. The Australian data sheets for the Fire Safety Verification Method (print version 1.1, 2020) use their results for characteristic fire growth rates and for the share of fires that become fully developed.<ref name="Holborn-2004">Holborn et al. (2004), as cited in Australian Building Codes Board (2020a), Data Sheet B2, pp. 4–7.</ref> The World Trade Center investigation itself ended with 30 recommendations, covering structural integrity, fire endurance, active fire protection, evacuation, emergency response, procedures and practices, and education and training.<ref name="NIST-xiii">National Institute of Standards and Technology (2005), p. xiii.</ref>


== See also ==
== See also ==


* [[Fire dynamics]]
* [[Firefighting]]
* [[Fire statistics]]
* [[Dry and wet risers]]
* [[Ignition source]]
* [[Fire hydrant]]


== Notes ==
== Notes ==
Line 128: Line 82:
== References ==
== References ==


* Australian Building Codes Board. (2020). ''Fire safety verification method data sheets'' (Version 1.1). Author.
* Australian Building Codes Board. (2020a). ''Data sheets: Fire Safety Verification Method'' (Handbook annex, Print version 1.1).
* Australian Building Codes Board, National Research Council of Canada, International Code Council, & Department of Building and Housing New Zealand. (2005). ''International fire engineering guidelines'' (2005 ed.). Australian Building Codes Board.
* Averill, J. D., Mileti, D. S., Peacock, R. D., Kuligowski, E. D., Groner, N., Proulx, G., Reneke, P. A., & Nelson, H. E. (2005). ''Occupant behavior, egress, and emergency communications'' (NIST NCSTAR 1-7). National Institute of Standards and Technology.
* Cooper, L. Y. (2002). Smoke and heat venting. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* Cooper, L. Y. (2002b). Smoke and heat venting. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 3-219–3-242). National Fire Protection Association; Society of Fire Protection Engineers.
* Evans, D. D., Peacock, R. D., Kuligowski, E. D., Dols, W. S., & Grosshandler, W. L. (2005). ''Active fire protection systems'' (NIST NCSTAR 1-4). National Institute of Standards and Technology.
* Gann, R. G., Hamins, A., McGrattan, K. B., Mulholland, G. W., Nelson, H. E., Ohlemiller, T. J., Pitts, W. M., & Prasad, K. R. (2005). ''Reconstruction of the fires in the World Trade Center towers'' (NIST NCSTAR 1-5). National Institute of Standards and Technology.
* Hall, J. R., Jr., & Ahrens, M. J. (2002). Data for engineering analysis. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* Gayle, F. W., Fields, R. J., Luecke, W. E., Banovic, S. W., Foecke, T., McCowan, C. N., Siewert, T. A., & McColskey, J. D. (2005). ''Mechanical and metallurgical analysis of structural steel'' (NIST NCSTAR 1-3). National Institute of Standards and Technology.
* Gray, B. (2002). Spontaneous combustion and self-heating. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 2-211–2-228). National Fire Protection Association; Society of Fire Protection Engineers.
* Hall, J. R., Jr., & Ahrens, M. J. (2002). Data for engineering analysis. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 5-65–5-78). National Fire Protection Association; Society of Fire Protection Engineers.
* Hopkins, M., Schoenrock, J., & Budnick, E. (2005). ''Fire suppression systems'' (NIST NCSTAR 1-4B). National Institute of Standards and Technology.
* Hopkins, M., Schoenrock, J., & Budnick, E. (2005). ''Fire suppression systems'' (NIST NCSTAR 1-4B). National Institute of Standards and Technology.
* McGrattan, K., Hostikka, S., Floyd, J., McDermott, R., Vanella, M., Mueller, E., & Paul, C. (2026). ''Fire Dynamics Simulator technical reference guide: Volume 1: Mathematical model'' (6th ed.). National Institute of Standards and Technology.
* Iqbal, N., & Salley, M. H. (2004). ''Fire Dynamics Tools (FDTs): Quantitative fire hazard analysis methods for the U.S. Nuclear Regulatory Commission Fire Protection Inspection Program'' (NUREG-1805). U.S. Nuclear Regulatory Commission, Office of Nuclear Reactor Regulation.
* Notarianni, K. A. (2002). Uncertainty. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* Kuligowski, E. D., Evans, D. D., & Peacock, R. D. (2005). ''Post-construction fires prior to September 11, 2001'' (NIST NCSTAR 1-4A). National Institute of Standards and Technology.
* Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* McGrattan, K., Hostikka, S., Floyd, J., McDermott, R., Vanella, M., Mueller, E., & Paul, C. (2026). ''Fire Dynamics Simulator technical reference guide: Vol. 1. Mathematical model'' (6th ed., NIST Special Publication 1018-1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.SP.1018
* Schifiliti, R. P., Meacham, B. J., & Custer, R. L. P. (2002). Design of detection systems. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* National Institute of Standards and Technology. (2005). ''Final report on the collapse of the World Trade Center towers'' (NIST NCSTAR 1).
* U.S. Nuclear Regulatory Commission. (2005). ''Fire dynamics tools: Quantitative fire hazard analysis methods for the U.S. nuclear regulatory commission fire protection inspection program'' (NUREG-1805). Author.
* National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). ''International fire engineering guidelines'' (2005 ed.). Australian Building Codes Board.
* Notarianni, K. A. (2002). Uncertainty. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 5-40–5-64). National Fire Protection Association; Society of Fire Protection Engineers.
* Ohlemiller, T. J., Mulholland, G. W., Maranghides, A., Filliben, J. J., & Gann, R. G. (2005). ''Fire tests of single office workstations'' (NIST NCSTAR 1-5C). National Institute of Standards and Technology.
* Peacock, R. D., McGrattan, K. B., Forney, G. P., & Reneke, P. A. (2026). ''CFAST – Consolidated Fire and Smoke Transport: Vol. 1. Technical reference guide'' (Version 7, NIST Technical Note 1889v1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.TN.1889v1
* Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 3-342–3-366). National Fire Protection Association; Society of Fire Protection Engineers.
* Purser, D. A. (2002). Toxicity assessment of combustion products. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), ''SFPE handbook of fire protection engineering'' (3rd ed., pp. 2-83–2-171). National Fire Protection Association; Society of Fire Protection Engineers.
* Van Rossum, E. (2027). ''Introduction to Fire Safety Engineering''. Fire Safety Academy.
* Van Rossum, E. (2027). ''Introduction to Fire Safety Engineering''. Fire Safety Academy.
* Watts, J. M., Jr., & Hall, J. R., Jr. (2002a). Building fire risk analysis. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.
* Zalosh, R. (2002). Explosion protection. In P. J. DiNenno (Ed.), ''SFPE handbook of fire protection engineering'' (3rd ed.). National Fire Protection Association.


[[Category:Firefighting & Fire Service]]
[[Category:Firefighting & Fire Service]]

Latest revision as of 19:38, 2 November 2026

Fire investigation is the examination of a fire after the event to establish where and how it started, how it developed and how the building, its fire protection and its occupants were affected. Writing on spontaneous ignition in the SFPE Handbook, Gray notes that fire prevention and fire cause investigation both need to identify the critical conditions for ignition, before and after the event respectively.[1] Hall and Ahrens list the reconstruction of major fires among the engineering analyses that require input data on fires.[2] The Society of Fire Protection Engineers likewise counts post-fire investigation and analysis as part of fire protection engineering.[3]

Fire investigation matters because its findings explain why fires start and why protection succeeds or fails, and they can lead to changes in how buildings are designed, built and regulated. The US investigation of the World Trade Center disaster of 2001, for example, was meant to serve as the basis for improvements in building practice and revisions to codes and standards.[4] Investigations depend on the evidence that survives, and large fires destroy much of it. Fire loss statistics from many sources show an unknown cause more often in fires with multimillion-dollar losses than in smaller fires, probably because more evidence is lost.[1]

Purpose and scope

Investigations differ in purpose and authority. The US National Institute of Standards and Technology (NIST) investigated the World Trade Center disaster to determine why and how the buildings collapsed. It also asked why injuries and fatalities were high or low depending on location, and examined the procedures used in design, construction, operation and maintenance and the parts of codes and standards that warranted revision.[4] NIST is a non-regulatory agency whose investigations focus on fact finding. It has no authority to find fault or negligence, and its reports may not be used in lawsuits for damages.[4]

Other investigations can have legal consequences. Van Rossum notes that an investigation after a fire disaster may show that the cause lies in demonstrably wrong advice, below the knowledge and skill that could be expected at the time. In that case the adviser can be prosecuted.[5]

For the World Trade Center buildings, the New York City fire department provided NIST with 112 records of its Bureau of Fire Investigation from 1977 to 2001. It also provided 397 operational fire reports from 1970 to 2001, all on standardised forms. NIST's review of earlier fires included, for example, a fire on 14 February 1975 on the 11th floor of WTC 1, one of the two towers. It spread through unprotected floor openings in utility closets and caused damage from the 10th to the 19th floor.[6]

Large investigations break the work into specific questions. For the sprinklers, standpipes and hoses of the towers, the NIST study first documented their design and installation and compared them with the codes and standards of record. It then estimated their capacity under normal conditions with hydraulic calculations.[7] Only then did it ask what happened to the systems at each major event, how their performance was affected and at what point in the sequence they were lost.[8]

Questions guide evidence collection, reconstruction, modelling and findings with uncertainty.
Fire investigation reconstructs events from evidence and explicit inference. Based on Hopkins et al. (2005), pp. xxxi–xxxii and Hall and Ahrens (2002), pp. 5-65–5-67.

Evidence and how it is gathered

Fire investigators normally read the path and intensity of a fire from telltale signs left in the building. In the World Trade Center towers this was impossible because neither building remained, so NIST relied on what may have been the most photographed disaster in history.[9] In each photograph and video frame, every window was coded for fire, smoke and whether the window was still in place, which produced detailed timelines of fire spread.[9] NIST also reviewed tens of thousands of pages of documents, interviewed over a thousand people and analysed 236 pieces of steel recovered from the wreckage.[10]

Time links the evidence together. NIST describes time as the unifying factor in combining photographs, videos, survivor accounts, emergency calls and communications among emergency responders.[11] Images were timed by matching them to one well-defined event, the impact of the second aircraft at 9:02:59 a.m., and then to other events visible in several images. In all, 3,032 photographs and 2,673 video clips were timed to within 3 seconds or better, and the uncertainty of each time was logged.[12]

Physical evidence needs careful interpretation. NIST found that the primer paint on the towers' steel cracked in a "mud cracking" pattern above 250 °C. Deformation and the environment can also cause such cracks, so their absence shows that the steel stayed below 250 °C, but their presence does not prove high temperature.[13] The method could not tell whether the heating happened before or after the collapse. The panels examined were also only 3% of the perimeter columns on the fire floors and could not be treated as representative of the others.[14]

People are a source of evidence as well. For the evacuation of the towers, NIST gathered over 1,000 new interviews with survivors, over 700 published interviews, emergency calls and transcripts of emergency communications.[15] Proulx notes that fire victims may have difficulty reporting accurately how long they waited before starting to evacuate, but that interview techniques can give acceptable estimates.[16] The causes of death in fires have been studied through pathological investigations of fire victims.[17] Carbon monoxide combines with haemoglobin in the blood to form carboxyhaemoglobin. In the Strathclyde study of fire deaths in the United Kingdom, 54% of those who died had a lethal level of it, and 69% had a level capable of causing incapacitation.[18]

Establishing the cause

To describe how a fire started, investigators separate the parts of the ignition sequence. US fire incident coding distinguishes the form of heat of ignition and the equipment involved, which together describe the ignition source, from the form and type of material first ignited. It also records the ignition factor that brought heat and fuel together, and the area of origin.[19] The New York fire department's structural fire form recorded the same kind of information, and it called the ignition factor the cause.[20]

Some causes leave characteristic traces. Spontaneous ignition of stacked material, such as cotton, woodchips or oily rags, starts in the well-insulated centre of the pile. Such a pile can therefore be charred or burnt to ash inside while looking almost untouched outside.[21][22] Hot gases escape along the path of least resistance and form "chimneys" of charred material, and several chimneys are the norm. Gray reports that finding more than one heavily charred area has led to wrongful charges of arson, based on the myth that more than one seat of fire means the fire was started deliberately.[23]

Gray lists the practical factors that make spontaneous ignition more likely, including a large body of material, high surrounding temperatures, thermal insulation, porous or fibrous material and a high stacking temperature.[22] Material left undisturbed for longer than usual is also reason for suspicion. In one case a warehouse of cottonseed meal burned down after a transport strike had left the meal in place far longer than normal.[23] In another, an instant-noodle factory burned down after new management sped up the production line, so that the noodles were packed hotter than before. Laboratory tests and calculations then showed that the packed pallets could ignite spontaneously at the new temperatures but not at the old ones.[24]

Gray advises caution about the timing of such fires without detailed knowledge of the initial conditions.[22] Whenever spontaneous ignition is suspected, Gray recommends excavating the material, with photographs and temperature probe records at all stages. He has measured temperatures as high as 200 °C in buried material more than two weeks after a fire was put out. Unburnt oily rags at the bottom of a bin do not rule out spontaneous ignition either.[23]

Reconstructing the fire

A reconstruction places the evidence in the sequence of the fire. Van Rossum divides a room fire into three phases. In the development phase, a local fire of one or a few objects grows until flashover. After flashover the fire is fully developed, and the decay phase can begin, for example, when the fuel runs out or the fire brigade intervenes. These fire development stages matter for interpretation, because an active suppression system normally acts during the development phase and so prevents the fire from becoming fully developed.[25]

Fire dynamics, defined in a US nuclear regulatory report as the scientific study of hostile fires, describes ignition, flame spread, fire growth, smoke movement and their effects in an enclosure.[26] Smoke rises under buoyancy to the upper part of the room of origin and leaves through openings, driven mainly by pressure differences. The International Fire Engineering Guidelines (2005 edition) note that hand calculations of its spread beyond the room of origin are more difficult and more uncertain than those within it.[27][28]

Tests can fill gaps in the evidence. NIST set out to characterise the type, mass and layout of the combustibles in the towers, the fire endurance of the partitions and the properties of the insulation on the steel.[29] Its reconstruction relied heavily on computer simulation, because the premises could not be examined after the fire and eyewitness information was severely limited. To support the simulations, NIST burned six single office workstations, the dominant combustibles in the buildings, to learn how they may have burned and to improve its fire model.[30]

Modelling and computational fluid dynamics

Computational fluid dynamics (CFD) studies the flow of gases and liquids by solving the flow equations numerically on a grid of cells. A CFD model can show smoke spread, gas temperatures and velocities, flows through openings and the time to flashover in three dimensions as they change over time.[31][32] Turbulent flows, which are typical of fires, can only be approximated. CFD therefore gives good qualitative insight, but its exact results must be interpreted carefully and critically.[31] Zone models such as CFAST, released by NIST,[33] are also used internationally for fire safety design and post-fire reconstruction.[34]

In the World Trade Center investigation, the unusual mix of available information led to a high reliance on computational tools, and NIST upgraded its CFD model, the Fire Dynamics Simulator, for the reconstruction.[29][35] It varied the uncertain inputs to find those to which the results were most sensitive and those that agreed best with the photographs.[36] The accuracy of the predictions was judged by their consistency with the photographic evidence.[35] For each tower it ran a less severe and a more severe case, which gave some understanding of the uncertainty in the predictions.[37]

Two candidate scenarios are tested against observations and model limitations before conclusions.
Alternative reconstructions should be tested against evidence and sensitivity. Based on Notarianni (2002), pp. 5-43–5-44 and McGrattan et al. (2026), p. 107.

Models also have limits. NIST decided not to model smoke movement and survival conditions throughout the occupied parts of the towers. There was no quantitative information on conditions inside the towers and no verifiable information on damage between floors, so such simulations for the impact floors and above would have been grossly uncertain.[38]

The Fire Dynamics Simulator predicts the thermal environment of a fire, but it relies on empirical models for the activation of sprinklers and detectors.[39] Smoke detector response is especially hard to model, because the production and transport of smoke early in a fire are not well understood.[40] Notarianni adds that fire model outputs are single values that do not reflect the uncertainty in inputs such as fire growth rates. A predicted activation time of 121 s for a detector, for example, may in reality be longer or shorter, depending on uncertain inputs and on factors the model does not include.[41] In forensic reconstructions, the amount of soot deposited on surfaces in a simulation can be correlated with what is observed after the fire.[42]

Evaluating protection systems

An investigation compares what a system could do with what it actually did. NIST found that the sprinkler and standpipe systems of the towers had for the most part exceeded the code requirements of their time, but the adverse effects of the aircraft impacts on them were extensive.[43] WTC 1 was the first tower to be struck.[44] Its estimated impact damage would probably have destroyed at least one standpipe riser and the sprinkler systems on the 94th to 96th floors, and cut the water supply to other systems after a limited time.[8][45] The collapse of WTC 2 damaged the city water mains near the site, and the collapse of WTC 1 probably caused significant further damage.[45]

Pre-incident conditions, ignition, detection, fire growth and consequences, with fire, water-supply and occupant tracks.
An investigation must align system states with the incident timeline. Based on Hopkins et al. (2005), pp. xliii–xlv and Van Rossum (2027), §4.11.

The evidence on systems is often incomplete. For the towers' suppression systems there was very little eyewitness or communications information. NIST therefore based its estimates on the incident information it had gathered, the likely effects of the impacts and historical performance records for sprinklers.[44] Most of the New York fire department's reports of significant earlier fires documented the performance of detectors and sprinklers, but almost all lacked information on the smoke purge system.[20]

Which sprinklers opened, and when, also needs careful reading. Water droplets carried by the hot gas flow under the ceiling from the first sprinklers can cool the heat-sensitive links of sprinklers further away and delay their opening. Cooper calls this random effect sprinkler skipping, and fusible-link vents can be delayed in the same way.[46] Evidence from victims has comparable limits. Carboxyhaemoglobin is routinely measured in the blood of fire victims, but blood cyanide often is not, and post-fire cyanide levels are hard to relate to the exposure during the fire.[18][47]

Limits and use of findings

Investigators often work with fragmented information. NIST confirmed significant events from more than one source where possible, but this could not always be done, and it included events that it judged likely or probable from the accumulated information.[44] Records can also be lost. The Port Authority of New York and New Jersey kept records of all significant fire events in the World Trade Center buildings, and these were lost in the collapse of the towers.[6] Physical evidence may cover only a small and unrepresentative part of a building, as the recovered steel did.[14]

A single investigation describes one event. Hall and Ahrens note that detailed accounts, including NFPA's reports of fire investigations, show the range and diversity of fires that can happen. Such anecdotes cannot be used to estimate incident rates or probabilities.[48] Valid analysis of incident rates instead requires a statistically representative database of fire statistics.[49]

Collected investigation data can in turn feed design values. Holborn and colleagues analysed data from fire investigations in the Greater London area between 1996 and 2000 to derive fire sizes, fire growth rates and the times between events. The Australian data sheets for the Fire Safety Verification Method (print version 1.1, 2020) use their results for characteristic fire growth rates and for the share of fires that become fully developed.[50] The World Trade Center investigation itself ended with 30 recommendations, covering structural integrity, fire endurance, active fire protection, evacuation, emergency response, procedures and practices, and education and training.[51]

See also

Notes

  1. ↑ 1.0 1.1 Gray (2002), p. 2-211.
  2. ↑ Hall and Ahrens (2002), p. 5-65.
  3. ↑ National Research Council of Canada et al. (2005), Part 0 (United States), pp. 0.3-1–0.3-2.
  4. ↑ 4.0 4.1 4.2 National Institute of Standards and Technology (2005), pp. xxxv–xxxvi.
  5. ↑ Van Rossum (2027), §2.5.
  6. ↑ 6.0 6.1 Kuligowski et al. (2005), p. 1.
  7. ↑ Hopkins et al. (2005), p. xxxii.
  8. ↑ 8.0 8.1 Hopkins et al. (2005), p. xliii.
  9. ↑ 9.0 9.1 Gann, Hamins, et al. (2005), p. xxxviii.
  10. ↑ National Institute of Standards and Technology (2005), p. xxxvi.
  11. ↑ National Institute of Standards and Technology (2005), p. 82.
  12. ↑ National Institute of Standards and Technology (2005), pp. 83–84.
  13. ↑ Gayle et al. (2005), p. 94.
  14. ↑ 14.0 14.1 Gayle et al. (2005), p. xli.
  15. ↑ Averill et al. (2005), p. iii.
  16. ↑ Proulx (2002), p. 3-349.
  17. ↑ Purser (2002), p. 2-85.
  18. ↑ 18.0 18.1 Purser (2002), p. 2-100.
  19. ↑ Hall and Ahrens (2002), p. 5-68.
  20. ↑ 20.0 20.1 Kuligowski et al. (2005), p. 2.
  21. ↑ Gray (2002), p. 2-212.
  22. ↑ 22.0 22.1 22.2 Gray (2002), p. 2-224.
  23. ↑ 23.0 23.1 23.2 Gray (2002), p. 2-225.
  24. ↑ Gray (2002), pp. 2-225–2-226.
  25. ↑ Van Rossum (2027), §4.11.
  26. ↑ Iqbal and Salley (2004), p. 1-1.
  27. ↑ National Research Council of Canada et al. (2005), p. 2.5-2.
  28. ↑ National Research Council of Canada et al. (2005), p. 2.5-6.
  29. ↑ 29.0 29.1 Gann, Hamins, et al. (2005), p. 2.
  30. ↑ Ohlemiller et al. (2005), p. iii.
  31. ↑ 31.0 31.1 Van Rossum (2027), §11.1.
  32. ↑ Van Rossum (2027), §11.2.
  33. ↑ Van Rossum (2027), §9.5.
  34. ↑ Peacock et al. (2026), p. 3.
  35. ↑ 35.0 35.1 Gann, Hamins, et al. (2005), p. 4.
  36. ↑ Gann, Hamins, et al. (2005), pp. 2–3.
  37. ↑ National Institute of Standards and Technology (2005), pp. 81–82.
  38. ↑ Gann, Hamins, et al. (2005), p. 3.
  39. ↑ McGrattan et al. (2026), p. 107.
  40. ↑ McGrattan et al. (2026), p. 108.
  41. ↑ Notarianni (2002), p. 5-43.
  42. ↑ McGrattan et al. (2026), p. 87.
  43. ↑ Hopkins et al. (2005), p. xlv.
  44. ↑ 44.0 44.1 44.2 Hopkins et al. (2005), p. xxxi.
  45. ↑ 45.0 45.1 Hopkins et al. (2005), p. xliv.
  46. ↑ Cooper (2002b), p. 3-239.
  47. ↑ Purser (2002), p. 2-103.
  48. ↑ Hall and Ahrens (2002), p. 5-70.
  49. ↑ Hall and Ahrens (2002), p. 5-69.
  50. ↑ Holborn et al. (2004), as cited in Australian Building Codes Board (2020a), Data Sheet B2, pp. 4–7.
  51. ↑ National Institute of Standards and Technology (2005), p. xiii.

References

  • Australian Building Codes Board. (2020a). Data sheets: Fire Safety Verification Method (Handbook annex, Print version 1.1).
  • Averill, J. D., Mileti, D. S., Peacock, R. D., Kuligowski, E. D., Groner, N., Proulx, G., Reneke, P. A., & Nelson, H. E. (2005). Occupant behavior, egress, and emergency communications (NIST NCSTAR 1-7). National Institute of Standards and Technology.
  • Cooper, L. Y. (2002b). Smoke and heat venting. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-219–3-242). National Fire Protection Association; Society of Fire Protection Engineers.
  • Gann, R. G., Hamins, A., McGrattan, K. B., Mulholland, G. W., Nelson, H. E., Ohlemiller, T. J., Pitts, W. M., & Prasad, K. R. (2005). Reconstruction of the fires in the World Trade Center towers (NIST NCSTAR 1-5). National Institute of Standards and Technology.
  • Gayle, F. W., Fields, R. J., Luecke, W. E., Banovic, S. W., Foecke, T., McCowan, C. N., Siewert, T. A., & McColskey, J. D. (2005). Mechanical and metallurgical analysis of structural steel (NIST NCSTAR 1-3). National Institute of Standards and Technology.
  • Gray, B. (2002). Spontaneous combustion and self-heating. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-211–2-228). National Fire Protection Association; Society of Fire Protection Engineers.
  • Hall, J. R., Jr., & Ahrens, M. J. (2002). Data for engineering analysis. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-65–5-78). National Fire Protection Association; Society of Fire Protection Engineers.
  • Hopkins, M., Schoenrock, J., & Budnick, E. (2005). Fire suppression systems (NIST NCSTAR 1-4B). National Institute of Standards and Technology.
  • Iqbal, N., & Salley, M. H. (2004). Fire Dynamics Tools (FDTs): Quantitative fire hazard analysis methods for the U.S. Nuclear Regulatory Commission Fire Protection Inspection Program (NUREG-1805). U.S. Nuclear Regulatory Commission, Office of Nuclear Reactor Regulation.
  • Kuligowski, E. D., Evans, D. D., & Peacock, R. D. (2005). Post-construction fires prior to September 11, 2001 (NIST NCSTAR 1-4A). National Institute of Standards and Technology.
  • McGrattan, K., Hostikka, S., Floyd, J., McDermott, R., Vanella, M., Mueller, E., & Paul, C. (2026). Fire Dynamics Simulator technical reference guide: Vol. 1. Mathematical model (6th ed., NIST Special Publication 1018-1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.SP.1018
  • National Institute of Standards and Technology. (2005). Final report on the collapse of the World Trade Center towers (NIST NCSTAR 1).
  • National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). International fire engineering guidelines (2005 ed.). Australian Building Codes Board.
  • Notarianni, K. A. (2002). Uncertainty. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-40–5-64). National Fire Protection Association; Society of Fire Protection Engineers.
  • Ohlemiller, T. J., Mulholland, G. W., Maranghides, A., Filliben, J. J., & Gann, R. G. (2005). Fire tests of single office workstations (NIST NCSTAR 1-5C). National Institute of Standards and Technology.
  • Peacock, R. D., McGrattan, K. B., Forney, G. P., & Reneke, P. A. (2026). CFAST – Consolidated Fire and Smoke Transport: Vol. 1. Technical reference guide (Version 7, NIST Technical Note 1889v1). National Institute of Standards and Technology. https://doi.org/10.6028/NIST.TN.1889v1
  • Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-342–3-366). National Fire Protection Association; Society of Fire Protection Engineers.
  • Purser, D. A. (2002). Toxicity assessment of combustion products. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-83–2-171). National Fire Protection Association; Society of Fire Protection Engineers.
  • Van Rossum, E. (2027). Introduction to Fire Safety Engineering. Fire Safety Academy.
Retrieved from "https://fire-safety.academy/w/index.php?title=Fire_investigation&oldid=2804"