A 1977 report by the US Justice Department listed the ‘indicators’ or ‘post-fire’ artefacts used by fire investigators in their analysis of fires (see "Fire indicators and methods of analysis" below). Worryingly, the report authors correctly pointed out that there was no scientific basis for any of these indicators – information that is critical in determining whether a fire has been started by accident or was the result of arson.
The report was the first inkling of problems in the US fire investigations service. Fire investigators in the US are generally former police and fire officers, with little or no scientific background. Training, historically, was done ‘on the job’ with an experienced mentor.
The report authors recommended publication of a handbook for fire investigators, together with a series of tests. Unfortunately, a handbook was published in 1980 without experimental data. Even more unfortunately, it was published by the highly respected US National Bureau of Standards (now the National Institute of Standards and Technology), so the indicators became firmly embedded in the culture of fire investigation.
The problem was compounded throughout the 1980s by training methods for fire investigators. They were sent on seminars during which a structure would be intentionally set on fire and then quickly put out to show them how to ‘recognise arson’. The difficulty with this type of test fire was that it did not mimic real fires, which often burn for tens of minutes. Such exercises reinforced the belief that one could tell what a ‘flammable liquid pour pattern’ looked like by observation alone.
Over the next few years, scientists investigated whether, in fact, these indicators were valid and, one by one, they were proven not to be. After a few highly publicised ‘arson’ cases were re-examined, it became clear that any fire could create patterns of damage that looked remarkably like patterns created by flammable liquids, even when none were present. A study conducted in the aftermath of the devastating Oakland California fires of 1991 concluded that many previously identified ‘indicators of arson’ could be found in structures that burned in fires known to be accidental.
Laboratory simulations also demonstrated that the temperature of a well-ventilated wood fire was the same or even higher than the temperature of a well-ventilated fire involving liquid hydrocarbons. The proposition that glass would craze in response to rapid heating was disproved. As most chemists have learned, glass will only craze when it is heated and then rapidly cooled.
In 1992, the National Fire Protection Association, (NFPA), based in Quincy Massachusetts, US, published Guide for fire and explosion investigations, debunking the indicators used by fire investigators. The document is constantly being updated and new editions are published on a three-year cycle.
When the guide was first published, the majority of fire investigators in the US rejected its advice because it was at odds with what they had ‘learned’. Eventually, however, the discipline embraced a more scientific approach to fires, and many of the ‘old school’ fire investigators retired. As the German theoretical chemist Max Planck once said, ‘science advances one funeral at a time’.
As a consequence of this work, many cases of alleged arson were revisited, and a few convictions were overturned (see "Case files" below), but the primary beneficiaries of the new knowledge were people who had been accused but not yet convicted. In time, the percentage of fires declared to have been started intentionally dropped as fire investigators became aware that their judgment was likely to be challenged.
Thirty years on
The situation today is dramatically improved but there is still much to learn about the behaviour of fires. In 2005 and 2007, the US Bureau of Alcohol, Tobacco, Firearms and Explosives – which investigates many of the high-profile fires in the US today – conducted studies that reflected poorly on fire investigators’ ability to establish where a fire is started. Unless they can pinpoint the correct origin of a fire, they are unlikely to be able to determine its cause.
In the 2005 study, two test fires were set in single-room structures and allowed to burn for two minutes after flashover. Flashover is a transition point when a ‘fire in a room’ becomes a ‘room on fire’. Fifty-three investigators were shown the two rooms and asked to write down the quadrant in which they believe the fire originated. Relying on their interpretation of the fire patterns and their belief that the lowest and deepest char indicated the origin, 50 of the 53 selected the wrong quadrant in both rooms.
In the 2007 study, three test fires were set, and allowed to burn for 30s, 70s, and three minutes beyond flashover. Again, investigators were asked to choose the quadrant where they believed the fire originated. For the 30s fire, 59 of 70 investigators correctly identified the quadrant of origin. For the 70s fire, 44 of 64 investigators identified the correct quadrant. For the three-minute fire, only 13 of the 53 identified the correct quadrant. This number is no higher than it would be if the quadrant of origin had been selected at random, or if the investigators had depended on two coin tosses.
What misleads fire investigators is ventilation-induced fire patterns, which are created only after flashover and have no relationship to the origin of the fire. Once flashover occurs, the oxygen concentration in the room drops to near zero, so burning at the origin may actually stop. The room is charged with hot combustible fuel, the products of combustion, which themselves can only burn in the presence of sufficient oxygen. Thus, the heaviest burning takes place where the fuel finds an oxygen source.
Interpreting ventilation patterns is not straightforward. The ventilation that comes in through a door may create a fire with a radiant heat flux of 150 kW/m2 on the wall opposite the door. Every fire pattern in a room needs to be examined to determine whether it can be explained in terms of ventilation, and if it can, the investigator needs to look elsewhere for the origin of the fire.
If fire investigators determine the wrong point or area of origin of a fire, the ignition source and the first fuel ignited will not be there. They might then conclude that the first fuel ignited ‘must have been’ a flammable liquid and the ignition source ‘must have been’ an open flame. There may even be a ventilation-induced ‘irregular pattern’ on the floor at the base of the fire pattern determined to have been located just above the origin. When one considers how fires can be ventilated, fire investigators need to keep in mind that usually the hot gases and flames exit the top of the opening, while cool fresh air is drawn in at the bottom. This will result in floor level burning, which prior to the 1980s was believed to indicate flammable liquids burning on the floor, because ‘heat rises’.
In a compartment fire, heat rises only until it meets an obstruction such as a ceiling, and then a hot gas layer forms. This layer becomes thicker and hotter because the fire acts as a pump for both energy and matter in the form of combustible smoke. Once the radiant heat flux from the hot gas layer reaches a value of roughly 20 kW/m2, every exposed combustible surface in the room ignites in a matter of seconds.
The understanding of this phenomenon has been critical to the improvement of the accuracy of fire investigation, but the profession still has a long way to go. The situation will only improve when the people and organisations that hire fire investigators are able to offer salaries sufficiently high to attract educated scientists.
Fire indicators and methods of analysis
Crazing of glass the formation of irregular cracks in glass was believed to be caused by rapid, intense heating, and thus an indicator of the use of a liquid accelerant.
Depth of char it was widely believed that wood burned at a fixed rate, so the depth of burnt wood was used to estimate the duration of burning.
Line of demarcation a ‘puddle-shaped’ line of demarcation on floors or rugs was believed to indicate a liquid fire accelerant. In the cross-section of wood, a sharp distinct line of demarcation was thought to indicate a rapid, intense fire.
Sagged furniture springs because the heat required for furniture springs to collapse from their own weight is ca 620°C and because of the insulating effect of the upholstery, sagged springs were believed to caused by a fire originating inside the cushions, for example, from a cigarette between the cushions, or by an external fire intensified by a fire accelerant.
Spalling the surface of the concrete chips or even explodes violently in response to heat. Brown stains around the spall were said to indicate the use of a fire accelerant.
In addition to these indicators, there were several other widely accepted means used to analyse a fire, including:
- Fire load. Knowing the energy content (as opposed to the energy release rate) of the fuel in a structure was believed to allow an investigator to calculate the damage that a ‘normal’ fire should produce in a given time. Some investigators referred to a ‘standard time–temperature curve’.
- Low burning and holes in the floor. Because heat rises, it was widely believed that burning on the floor, particularly under furniture, indicated the origin of the fire.
At about 10:15 pm on 9 March 1985, David Lee-Gavitt became aware of a fire in his home. He broke a window, cutting his arms, but was unable to save his wife and two daughters ages 3 and 11 months. When he was discharged from hospital for treatment for his cuts and burns, he was arrested and charged with three counts of murder.
Fire investigators saw what they believed to be ‘pour patterns’ on the floor in the living room, and two of 17 samples taken from the scene came back ‘positive for gasoline’. Gavitt was convicted on 18 April 1986 and sentenced to life without parole.
In May 2010, the case was reviewed and I was asked for my opinion on the chemical analysis done by Michigan State Police laboratory to identify the gasoline. The chromatograms were of poor quality and, even by 1985 standards, they were an unresolved mess. The data did not indicate the presence of gasoline.
It took almost two years, but after getting an independent chemist to review my analysis, the prosecutor agreed that the state could not prove that an arson had occurred. Gavitt was released from prison on 26 June 2012.
In February 2010, Amanda Kelly’s 11-year nightmare finally came to an end. She had just pleaded guilty to three counts of cruelty to a juvenile stemming from a 2001 fire. Her crime? She had left a 10-year old daughter to babysit with two younger siblings while she ran errands for thirty minutes. But it could have been three counts of capital murder. The state had originally wanted to impose the death penalty.
The fire occurred on 9 January 2001. Kelly, who discovered the fire when she arrived home, was immediately a suspect, but she never admitted to setting the fire. Fire investigators saw spalling on the concrete floor of her home, and concluded that this was caused by a burning flammable liquid such as gasoline. No flammable liquid was detected in the laboratory analysis of the concrete samples.
The case was brought to court by a well-known, established fire investigator, who reasoned that the fire spread too fast to be an accident, and Kelly did not have sufficient smoke on her to support her account of her attempts to rescue her children. She was indicted in May 2002 based on this testimony. She spent four years in jail awaiting trial, but in 2006 credible scientific evidence, including a valid interpretation of the fire spread, exposure of the ‘expert’s’ use of a computer model, and criticism of the destruction of evidence was put before a judge, who granted her bail, and dismissed the indictment. That had never happened before in a capital case.
Two years later when the fire investigator was asked for a new report, he said he could no longer support the arson charge. No satisfactory explanation for his delay in reporting the change of opinion has been provided.
John Lentini is an analytical chemist at Scientific Fire Analysis, Florida, US.