Accelerant Detection at the Scene: Canines, PID, Sampling

Canine accelerant detection (CAD) teams have operated in fire investigation since the early 1980s. The Bureau of Alcohol, Tobacco, Firearms and Explosives (ATF) in the United States developed the first formal CAD programme in 1986 and remains the reference standard for training and certification worldwide. The programme pairs a dog trained to alert on the vapour signatures of ignitable liquids with a handler who is also a certified fire investigator. As of 2024, the ATF National Canine Centre in Front Royal, Virginia, has certified approximately 62 accelerant detection canine teams deployed across state fire marshal offices, local fire departments, and federal agencies.

In the United Kingdom, the National Fire Chiefs Council (NFCC) and the Metropolitan Police Service both operate CAD teams; the Met's Directorate of Forensic Services has deployed accelerant detection dogs since the late 1990s. Scotland's Fire Investigation Unit and the Scottish Police Authority maintain a joint canine programme. In India, the Central Industrial Security Force (CISF) and several state police forensic science laboratories operate canine units trained primarily for explosives; a smaller number of dogs are dual-trained for accelerant vapours, and CFSL Hyderabad has evaluated dual-use training protocols.

The training method is operant conditioning, specifically a variant of clicker training. The dog is taught to associate the odour profile of a hydrocarbon mixture with a food reward, then progressively trained on a menu of ignitable liquid vapours: gasolines of different grades and weathering states, kerosene, diesel, lamp oil, charcoal lighter fluid, and industrial solvents. The ASTM E2323 standard, first published in 2003 and updated most recently in 2019, specifies the minimum competency requirements for an accelerant detection canine, including the vapour concentrations at which the dog must reliably alert and the distractor substances the dog must not alert on (motor oil, charred wood, rubber, asphalt). Training decoys use sterile gauze pads doped with known concentrations of accelerant vapour held in screw-top glass containers, a method standardised by the NFPA 921 Technical Committee on Fire Investigations.

The biological basis for canine detection is well characterised. Dogs process olfactory information through approximately 300 million receptor cells in their olfactory epithelium, compared with approximately 6 million in humans, and their olfactory cortex represents roughly 12 percent of total brain volume versus 0.01 percent in humans. More practically, the olfactory threshold for many petroleum hydrocarbons in trained dogs has been measured by controlled trials (including the work of Jezierski et al., published in Forensic Science International in 2012) at concentrations in the range of 1 to 10 parts per billion, well below the detection threshold of most electronic instruments in field use.

Independent peer-reviewed research on CAD reliability has raised a concern that fire investigators and courts must understand: handler cueing. The study most often cited is by Lisa Lit and colleagues, published in Animal Cognition in 2011, which found that dogs alerted significantly more often in locations where their handlers believed accelerant was present, even when no accelerant had been placed there by the researchers. The study design used professional detection dogs and handlers, and the handlers were misled about the location of target scents. The conclusion was not that dogs are unreliable detectors, but that handler expectation can be communicated to the dog through subtle postural and behavioural cues that the handler does not consciously intend.

The forensic implication is procedural. In high-stakes fire investigations, a double-blind CAD search protocol is the defensible standard: the handler should not know in advance which areas the investigator suspects as origin zones or pour-pattern locations. In practice, a double-blind search is not always achievable at an active fire scene, but the principle of minimising handler cueing should inform how the investigator directs the team through the structure.

False positives also arise from genuine odour confusion. The ASTM E2323 training list covers the most common accelerants, but not every substrate that produces a hydrocarbon vapour signature at elevated temperatures. Asphalt flooring, rubber-backed carpet, electrical cable insulation, certain adhesives, and engine oil can produce vapour signatures that a dog trained primarily on petroleum distillates may alert on. This is the scene-side manifestation of the same pyrolysis interferent problem that challenges laboratory analysts during GC-MS background subtraction. This is an inherent property of a detection system operating on vapour chemistry rather than molecular-level specificity. The electronic confirmation step, and more definitively the laboratory GC-MS step, resolves the ambiguity.

False negatives are the less-discussed risk. Heavy suppression water, long time between ignition and canine search, intense heat that volatilised all residue, or heavy-pour burial under debris can all produce zero canine response to genuine accelerant. A negative canine result does not exclude accelerant use. NFPA 921 (2024 edition, section 18.5.1) is explicit: an absence of canine alert should not be interpreted as absence of ignitable liquid.

Two classes of electronic instrument are in regular fire investigation use: photoionisation detectors (PIDs) and combustible gas indicators (CGIs). They operate on different physical principles, have different sensitivity profiles, and should be understood as complementary rather than interchangeable.

A photoionisation detector uses an ultraviolet lamp, typically at 10.6 electronvolts (eV), to ionise organic vapour molecules in a sampled air stream. When a molecule absorbs a photon of sufficient energy, it loses an electron and becomes a positively charged ion; the instrument measures the resulting current between two electrodes. PID response is compound-dependent: molecules with ionisation potential below the lamp energy (benzene, toluene, ethylbenzene, xylenes, many aliphatic hydrocarbons) produce a strong signal, while compounds with higher ionisation potentials or high polarity may not ionise at the lamp energy used. The MiniRAE 3000 (Honeywell Analytics), the ppbRAE 3000 (RAE Systems), and the Tiger (Ion Science) are the most widely used PIDs in fire investigation; all three can detect aromatic and aliphatic hydrocarbons at the parts-per-billion level with correction factors applied for specific compound classes.

A combustible gas indicator (CGI), also called a catalytic bead sensor or pellistor detector, measures the heat of combustion of a vapour-air mixture across a heated platinum-wire bead catalyst. The change in bead resistance is proportional to the vapour concentration and is read against a calibration curve, typically expressed as percent of lower explosive limit (% LEL). CGIs are robust, inexpensive, and effective at concentrations near the LEL of common fuels (1 to 4 percent vapour in air for most petroleum products), but they are not sufficiently sensitive at the sub-LEL concentrations typical in fire debris that has been suppressed and cooled. Most fire investigation protocols treat the CGI as a safety and gross-concentration screening tool rather than a detection tool.

In practice, an experienced fire investigator uses the PID as a semi-quantitative guide to prioritise sampling locations. A reading that spikes when the probe tip is held near a floor joint, a carpet seam, or low-lying debris is not confirmatory evidence of accelerant, but it directs where to collect. Readings should be documented (instrument type, serial number, calibration date, reading value, and probe location) because the PID survey record may be reviewed in court and any unexplained gap in documentation creates an evidentiary vulnerability.

Temperature matters for PID use. At fire scene temperatures (floor surfaces may be 30 to 60°C even hours after suppression), vapour pressure of residual accelerant is elevated and PID response will be higher than at ambient temperature. The investigator should note the surface temperature when recording PID readings, both for interpretation and to flag when suppression water may be volatilising residues from flooring that are not accelerant-origin but simply mobilised by heat.

The collection of fire debris samples for laboratory analysis is governed by ASTM E1492, the standard practice for receiving, documenting, storing, and retrieving evidence in the forensic science laboratory, and by NFPA 921 chapter 18, which covers fire debris sampling at the scene. The two most common container types are new, unused metal paint cans (1-quart or 1-gallon, with friction-fit lids) and nylon bags (specifically Kapak or equivalent heat-sealable nylon).

Metal paint cans are the traditional standard and remain the default for most fire investigation agencies in the US, UK, and Australia. They are airtight when properly sealed, chemically inert to petroleum hydrocarbons at scene temperatures, and can accommodate irregular solid debris (charred wood, carpet fragments, soil). The critical discipline is that cans must be new and unused; a recycled paint can with dried paint residue may contain hydrocarbon solvents that will appear in the GC-MS analysis as apparent evidence of an accelerant. In the UK, the same new-can requirement is carried forward in the Forensic Science Regulator's Codes of Practice and Conduct, which explicitly cover physical fire debris evidence handling.

Nylon bags are increasingly used for large or awkward samples, for bulk soil collection, or when paint cans are unavailable. The nylon polymer is permeable to water vapour but largely impermeable to petroleum hydrocarbons at the concentrations found in fire debris. Bag sealing must be heat-sealed or double-knotted with nylon-compatible closures; rubber-banded closures allow vapour escape. Glass jars with PTFE-lined lids are an alternative endorsed by the SWGFEX (Scientific Working Group for Fire and Explosion Investigation) best practice guide for liquid samples where the volume is small enough to avoid headspace loss.

1. Mark the collection location
   Before collecting any debris, photograph the undisturbed area and record its map coordinates within the scene diagram. Note the surface type (concrete, hardwood subfloor, carpet with pad, soil), its distance from the suspected origin, and any pour-pattern indicator that directed your attention to this area.
2. Select a new, unused paint can
   Inspect the can for interior coating (bare metal interior only), no previous use, and an undamaged friction lid that seats fully. Record the can manufacturer and lot number in the scene log.
3. Collect debris to fill the can one-half to two-thirds full
   Use a clean, new trowel or stainless steel spatula (never a tool used for a previous sample in the same investigation without solvent cleaning and air-drying). Fill the can with charred substrate, soil, carpet, or flooring fragment. Avoid overfilling: headspace is essential for vapour to accumulate during passive concentration.
4. Collect a comparison sample
   From unburned material of the same substrate type in the same room or an adjacent unaffected area, collect a separate comparison sample in an identically prepared new can. Label it clearly as a comparison (control) sample. This sample provides the substrate background against which GC-MS data will be interpreted.
5. Seal immediately and label
   Seat the friction lid firmly by hand and use a rubber mallet if needed. Apply a tamper-evident label across the lid joint. Record the sample number, case number, collection date, time, location descriptor, collector name, and weather conditions on the can label and in the scene log.
6. Complete the chain-of-custody form
   Begin the chain-of-custody record at the point of collection, not at the point of delivery to the laboratory. Every person who subsequently handles the sealed can must sign and date the record before the seal is broken.

The comparison sample requirement warrants specific emphasis. NFPA 921, SWGFEX best practice, and the UK Forensic Science Regulator guidance all identify comparison sampling as a mandatory discipline, not an optional enhancement. Laboratory analysts using the background subtraction methodology (covered in Topic 3 of this module) cannot produce a defensible conclusion without knowing what the unburned substrate contributes to the chromatogram. Without comparison samples, the analyst must either make assumptions about substrate background or qualify the conclusion to the point of marginal usefulness.

Chain of custody in fire debris investigation has two distinct vulnerabilities: pre-collection contamination (introducing hydrocarbon from the investigator's own clothing, equipment, or vehicle) and post-collection contamination or vapour loss (allowing the sealed evidence to degrade before laboratory analysis). Both vulnerabilities have generated successful evidentiary challenges in multiple jurisdictions.

Pre-collection contamination is more common than most investigators acknowledge. A fire investigator who refuelled their vehicle on the way to the scene, whose overboots have walked through a puddle of spilled fuel in the fire apparatus bay, whose examination gloves are the same ones used to handle a different sample earlier, or whose sampling tools were cleaned with a petroleum-based solvent without adequate airing can introduce hydrocarbon residues into the collected sample. The standard precaution is dedicated scene footwear (new or solvent-washed polyethylene overboots), clean latex or nitrile examination gloves changed between samples, and sampling tools that are either new and disposable or subject to a documented solvent-wash and air-dry protocol between samples.

Post-collection vapour loss is the second major vulnerability. A sealed paint can with a poorly seated lid will lose low-boiling components from the headspace between collection and laboratory analysis. At scene temperatures, the rate of loss depends on the tightness of the lid seal and the ambient temperature during storage. Samples stored in a vehicle boot on a warm day (35 to 40°C) may lose a meaningful proportion of the lighter fractions within 24 hours. Laboratory analysis should ideally occur within 72 hours of collection. The ASTM E1492 standard provides refrigerated storage guidance (4°C slows vapour loss substantially), and the SWGFEX best practice guide recommends laboratory receipt within 30 days with refrigerated storage, noting that heavily loaded samples (saturated debris) may be stable longer than lightly loaded ones.

In high-profile arson prosecutions, defence experts have successfully challenged fire debris evidence on contamination grounds in all three major jurisdictions. In the US, the Innocence Project has documented multiple post-conviction cases where fire debris evidence originally presented as confirmatory of arson was later reinterpreted after chain-of-custody review. In the UK, a 2004 arson retrial centred on chain-of-custody documentation for fire debris samples; this is a documented example of evidence handling being contested at retrial. In Australia, the 2009 Black Saturday royal commission reviewed fire investigation protocols for the entire Victoria state service following concerns about consistency of fire debris sampling procedure.

The practical standard is that every action from the moment the investigator identifies a candidate sampling location to the moment the laboratory analyst opens the can must be documented, time-stamped, and traceable. Documentation is not a bureaucratic exercise; it constitutes the evidentiary record for the evidence itself.

The confirmatory principle in fire debris investigation is the single most important conceptual bridge between scene-side detection and laboratory analysis. Neither a canine alert nor an electronic instrument reading is, in isolation, admissible evidence of accelerant use. Both are screening tools that generate hypotheses for laboratory testing. The laboratory result, reported under ASTM E1618 criteria and interpreted against the comparison sample, is the evidence.

This distinction has practical implications for how fire investigators communicate at scene and in reports. A CAD team alert should be documented as "the accelerant detection canine alerted in the area of the south corner of the living room, adjacent to the baseboard, at approximately 0940 hours." It should not be documented as "accelerant was detected by the CAD team in the south corner." The first formulation accurately describes what happened. The second draws a conclusion that only the laboratory can support, and if the laboratory later finds no ignitable liquid, the investigator has created an internal inconsistency in the case file that an experienced defence expert will exploit.

The same discipline applies to PID readings. "The photoionisation detector registered a peak reading of 47 ppm (isobutylene equivalents) at probe position 3, adjacent to the carpet seam at the base of the south wall" is a scene observation. "The PID detected accelerant at position 3" is a conclusion. An investigator who understands what each instrument actually measures will write the first type of note; the second formulation is difficult to defend under cross-examination if the laboratory finds only pyrolysis products from the substrate.

International accreditation frameworks have aligned on this structure. The FBI's Handbook of Forensic Services (current edition), the UK Forensic Science Regulator's Codes of Practice and Conduct, and the accreditation requirements of Standards Australia (AS 5388 series for forensic analysis) all treat the laboratory report as the threshold evidence and the scene screening tools as investigative aids. The SWGFEX best practice document, developed jointly by fire investigators, forensic chemists, and laboratory directors across the US, explicitly states (section 4.2, 2016 edition) that positive canine alerts must be confirmed by laboratory analysis before being cited in a report as evidence of an ignitable liquid.