Archive for April, 2013

by: Jason A. Sutula


Complex objectives arise within every fire origin and cause investigation. These objectives include the standard goals of determining the origin and cause of the fire and can also include less common objectives such as determining the cause of the loss, the cause of the fire spread, and the cause of injury or death. Investigators who follow the Scientific Method while performing fire origin and cause investigations are increasingly relying on research and technological advancements to conduct their analyses and support their opinions and conclusions. One of the most prevalent means of testing hypotheses and supporting analyses is through computer fire modeling.

Computer Fire Modeling

The majority of all computer fire codes used within the fire investigation community can be divided into two categories – zone codes and computational fluid dynamics (CFD) codes. While these two categories are different in their unique abilities and applicability, each type of code can be used to analyze a variety of critical issues within a fire origin and cause investigation.

The zone code category encompasses the many compartment fire codes that represent a modeled system as two different zones. These two zones are composed of an upper gas layer and a lower gas layer. The fire within the system is represented as an energy source that allows for energy and mass transport from the lower layer to the upper layer. Zone codes yield relevant data that is averaged for both the upper layer and the lower layer. Observing detailed fluid flow effects within the system is not possible with this type of code, but bulk fluid transport from compartment to compartment within a model can be predicted.

One of the most commonly used multi-zone fire codes is FAST1, which builds on the routines in the fire code CFAST. Within the code, mass and energy transfer between the zones is accounted for by a plume, mixing at vents (e.g., through connections between compartments), radiation between layers, and heat transfer at the boundary surfaces. The prime equations in FAST are based on the application of mass and energy conservation principles to homogeneous upper and lower gas regions in multi-compartment systems. FAST has been extensively validated in the scientific community1,2,3.

The CFD code category encompasses a much more limited number of currently available codes that can be used to model many different aspects of fire phenomena. CFD codes can divide a system of interest into millions of control volumes or cells. Thus, while zone codes are limited to modeling rectangular structures, CFD codes can easily be manipulated to model much more complicated geometries including curved and round surfaces. CFD models yield data on a much greater scale than zone models. Data can be resolved to each cell used to comprise the system, and local fluid flow effects can be observed.

One of the most commonly used codes for the study of fire scenarios in fire investigation is the Fire Dynamics Simulator (FDS)4. The FDS fire code was developed at the National Institute of Standards and Technology (NIST) by Dr. Kevin McGrattan and co-workers. The code was written utilizing a large eddy simulation technique to describe fire induced flows. This technique splits fluid flow into large-scale turbulence and small-scale turbulence. The large-scale turbulence is directly computed using the basic equations of fluid motion, while the small-scale turbulence is approximated using a sub-grid model. These techniques are well suited for modeling many aspects of the consequences of fire, including fire growth and smoke movement, since these phenomena are driven by large-scale structures in the flow.

Carbon Monoxide Uptake in Fire

The most significant factor affecting the ability of occupants to escape when remote from the room of origin is carbon monoxide. Carbon monoxide is a product of combustion that is formed in very large quantities after flashover, or full-involvement, has occurred within a compartment fire5. Carbon monoxide will combine readily with hemoglobin in the blood to form carboxyhemoglobin (COHb), which in sufficient levels will cause incapacitation followed by death6. Experimentation has shown that incapacitation will typically occur between 30 and 40% COHb6, and that behavioral performance deficits can occur at as little as 20% COHb6. These deficits can affect the ability to make sound decisions when attempting to escape. Additionally, it has been noted in the scientific literature that whether a subject was at rest or doing light to heavy activity drastically increased the observed symptoms from exposure6. Subjects engaged in light activity were observed to have rapid loss of function, which would significantly delay or prevent escape6.

Well-known research has been conducted into predicting the uptake of carbon monoxide in the blood6,7. The most commonly used model is the Coburn, Forster, Kane Equation (i.e., CFK Equation). Peterson and Stewart7 presented a validation of the CFK Equation and included a working model in their 1975 paper. Their model represents a series of equations and variables that can be solved with any iterative program and takes the basic form of7:

eqns for CO

The CFK Equation used in the model analysis developed by Peterson and Stewart is robust enough to calculate the time history of the uptake and excretion of carbon monoxide in any human individual regardless of age. The most pertinent variables necessary for the prediction of carbon monoxide uptake and excretion are the weight, sex, and activity level of the individual exposed and, most importantly, a time history of the concentration of carbon monoxide to which they were exposed.


Neither the zone codes nor CFD codes have directly incorporated a sub-model for the actual prediction of carbon monoxide uptake in the blood of possible victims within a fire. Thus, a methodology must be developed to link the output data that can be produced by a zone fire model or a CFD fire model.

The first step in this methodology is data collection. Every fire scene produces its own unique set of data. This will include the geometrical configuration of the structure, building materials present, eye-witness testimony, and, most important to this particular analysis, the medical records of all victims of the fire. When examining both injured victims of a fire as well as victims who die as a result of the fire, it is very common for documentation to exist that tabulates the COHb of the victim. For victims that end up in the Emergency Room, a measurement of the various gases in the blood is routinely drawn to provide vital information to the medical staff. For the victims who die as a result of the fire, a blood sample is also drawn post-mortem to specifically look for carbon monoxide in the blood. Retaining this type of information in a fire can provide a “hard time” to an analysis of when a victim was removed from exposure to a fire or when a victim became incapacitated as a result of the fire conditions surrounding them.

The second step in the methodology is the development and testing of strong candidate fire initiation hypotheses. If knowledge of the exact initiating event is well supported or not disputed, then this step can be trivial. A fire model of the lone scenario can be generated and allowed to predict the necessary carbon monoxide concentration data. This is usually not the case in most fire investigations, so a matrix of fire scenarios that cannot be ruled out must be developed and analyzed using bounding analysis techniques.

The third step is to process the data generated by the chosen fire model. Carbon monoxide data produced by a given fire model can be used directly with the CFK Equation as presented above to build a time history of the COHb in a particular victim’s blood. The time history can then be use to analyze various factors in a fire reconstruction analysis such as the time to incapacitation of a victim, the position and location of a victim within a structure over the course of the fire, and the time available for safe egress. Figure 1 provides a flow chart of the linking methodology.

Figure 1 for CO

Figure 1 – Flow Chart of Linking Methodology

Strengths and Weaknesses of Each Fire Code

Both zone fire models and CFD fire models are capable of predicting the generation rate of carbon monoxide as a byproduct of combustion. There are differences, though, in how the two categories of codes predict these values. These differences affect the usefulness and applicability of a particular fire code in predicting the generation of carbon monoxide within a given fire scenario.

In general, fire models developed using a zone code will not be able to provide the same resolution of data within a single compartment as with a CFD code. This is balanced by the ease with which a zone code handles the leakage between compartments and the well-validated bulk flow of gases from one compartment to another. Another strength of zone fire models is their ability to more easily handle rapidly changing production rates of carbon monoxide at the transition from a localized compartment fire to a fully involved compartment fire. The transition to flashover represents the conversion of a localized fire to a fully-involved, vitiated compartment fire that is in effect a carbon monoxide pump, which will quickly produce fatal amounts of the gas in short periods of time. While it is possible to account for this with CFD fire codes such as FDS, the process is much more simplified, verified, and quicker with zone fire models.

Due to the various strengths and weakness of both zone and CFD fire models, a combination modeling analysis utilizing both can be effectively used in fire reconstruction analyses. The following case study is presented as a means to demonstrate a combination modeling effort and to demonstrate the use of the methodology outlined in this paper.


A fire occurred in a three-story brick duplex house. The front of the house faced south, and the east wall was shared with a neighboring unit. The first floor had a living room, dining room, and kitchen. The second floor had three bedrooms and one bathroom. The third floor had an attic, which had been converted into a bedroom. There was also a basement with an additional bathroom. Figure 2 depicts a cut away view of the residence. The east wall has been removed and north faces to the right in the figure.

Figure 2 for CO

Figure 2 – Case Study Residence (North Points to the Right)

The fire occurred sometime prior to 2:54 a.m. while the residents of the house were asleep. At the time of the fire, the mother (age 38) and father (age 40) and five children (ages 18, 16, 14, 12, and 1) were present in the residence. The mother and father shared the front bedroom on the second floor with the 1-year-old child. The 14-year-old child and 12-year-old child shared a middle bedroom on the second floor. The 16-year-old child had the rear bedroom on the second floor, and the 18-year-old had her bedroom on the third floor.

When firefighters arrived at the home at 2:54 a.m., the house was already heavily involved. Three of the children (ages 18, 16, 14) had been able to escape by jumping from a second story window to the rear of the structure and were transported to the hospital. The rest of the occupants did not escape from the fire and died as a result.

After the fire, an origin and cause investigation was conducted. Analysis of the fire damage led to an initial conclusion that the fire originated in the basement and was accidentally caused by an electrical malfunction. Another potential accidental cause was electrical activity on an artificial Christmas tree in the room of origin. Subsequent inspections were performed to examine other hazardous electrical conditions and determine if smoke detectors were present. Two smoke detectors were found. One was found in the debris on the basement stairs near the kitchen entrance. The damage to the detector and its battery indicated it was disconnected at the time of the fire. The second detector was located on the ceiling of the third floor. This detector had a weak flash and sound, which did not improve when a newer battery was inserted. Indications of a smoke detector installation were also found on the walls of the second floor, and a later examination uncovered a smoke detector on the second floor within debris.

Autopsies were conducted on the victims of the fire. The father was found at the foot of the stairs on the first floor. Blood sample analysis revealed a COHb level of 28.7%. His cause of death was listed as smoke inhalation with a significant contributory cause of coronary artery atherosclerosis. The mother was found in the rear bedroom on the second floor with her 1-year old child. Her COHb level was reported as 58.1%. The 1-year-old child had sustained a COHb level of 85.4%. The final victim of the fire, the 12-year-old child, was found on the floor in her second floor bedroom. She had sustained a COHb level of 67%. The cause of death in the autopsy reports for these three victims was listed as “smoke inhalation.”

As mentioned, the first step in the proposed methodology is to collect data. The above-presented background is only a short summary of the data that was collected during the investigation phase of this fire event. Most importantly, the autopsy reports and laboratory report of the victims COHb were available and retained in this particular case.

The second step in the proposed methodology is to develop and tabulate the candidate hypotheses for fire initiation. Unfortunately in this particular case, the physical evidence and witness testimony did not allow for the cause of the fire to be narrowed to one scenario. Thus, several hypotheses were formulated and included an electrical heating unit, an artificial Christmas tree, an overheated extension cord, and an electrical start in the entertainment center. These fire scenarios were all carried through into the modeling analysis via separate models built for each unique scenario.

Fire Modeling Analysis

The third and final step in the methodology is to generate carbon monoxide concentration data from the fire modeling analysis and post-process the data using the CFK Equation as presented above to predict the blood COHb levels of the fire victims. In this case study, this analysis was critical as it addressed the lack of operational smoke detectors within the home. It was hypothesized that working smoke detectors within the residence would have afforded the victims of the fire sufficient notification to safely egress the building. Thus, a modeling comparison between the time to smoke detector activation and the time to incapacitation due to carbon monoxide was analyzed.

In order to determine the activation time of the smoke detectors, the FDS model was used. A complete three-dimensional geometry of the residence was created in FDS. The computational geometry of this FDS model contained 489,888 cells. The cells were cubes approximately four inches on a side, and each room of the residence was explicitly modeled according to diagrams in evidence, photographic evidence, and measurements gathered (i.e., refer to Figure 2 for a depiction of the modeled geometry). Many runs were conducted with the FDS model geometry representing the various initiation possibilities.

A zone modeling analysis with FAST was used to predict the time history of the concentration of carbon monoxide within the residence. FAST was chosen for this aspect of the analysis based on the previously presented strengths and weaknesses of CFD and zone fire models. Figure 3 shows an example prediction of COHb for the mother who was found with her 1-year-old child in the rear bedroom near the same window where the survivors had jumped to safety.

Figure 3 for CO

Figure 3 – Example COHb Prediction from Fire Modeling Data

The time history data of COHb uptake was then tabulated across all modeled fire scenarios and presented in tabular format to complete the analysis. Table 1 presents an example of how the most pertinent ignition scenarios in this case study were examined to determine the amount of escape time that would have been available to the residents if they had a working smoke detector on the night of the fire.

    Table 1 for CO
Table 1 – Example of Predicted Available Escape Time

The results shown in Table 1 supported the hypothesis that a working smoke detector in the residence would have provide earlier warning to the residents and allowed more time for escape. The modeling analysis predicted that four minutes of escape time would have been available to the residents as opposed to the actual results of the fire, which afforded them no time at all for self-preservation.


Neither zone modeling codes nor CFD modeling codes currently provide a means for assessing the toxicity effects on fire victims of carbon monoxide exposure. A model user is able to predict the production and transport of the toxic gas throughout the model space, but further analysis must be completed in the post-processing of the model results. This paper has presented a standardized methodology for linking the prediction of carbon monoxide uptake in victims of a fire with computer fire modeling. This methodology is robust enough to calculate the time history of the uptake and excretion of carbon monoxide in any human individual regardless of age and aid in the analysis of various factors in a fire reconstruction analysis such as the time to incapacitation of a victim, the position and location of a victim within a structure over the course of the fire, and the time available for safe egress.


1Peacock, R. D., Jones, W. W., and Bukowski, R. W. “Verification of a Model of Fire and Smoke Transport”, Fire Safety Journal, Vol. 21, No.2 89-129, 1993.

2Nelson, H. E., and Deal, S., “Comparing Compartment Fires with Compartment Fire Models”, National Institute of Standards and Technology, Gaithersburg, MD, International Association for Fire Safety Science.   Fire Safety Science.  Proceedings. 3rd International Symposium, July 8-12, 1991, Edinburgh, Scotland, Elsevier Applied Science, New York, Cox, G.: Langford, B., Editors, pp. 719-728.

3Dembsey, N. A.; Pagni, P. J.; Williamson, R.B., Compartment Fire Experiments:  Comparison with Models. Worcester Polytechnic Institute, MA; California Univ., Berkeley, CA; National Institute of Standards and Technology, Gaithersburg, MD. Fires Safety Journal, Vol. 25, No.3, 187-227, 1995.

4McGrattan, Kevin, et al., “Fire Dynamics Simulator (Version 5) User’s Guide”, NIST Special Publication 1019-5, 2010.

5Gottuk, D.T. and Lattimer, B.Y., “Effect of Combustion Conditions on Species Production,” SFPE Handbook of Fire Protection Engineering, 4th edition, 2008.

6Purser, D.A., “Assessment of Hazards to Occupants from Smoke, Toxic Gases, and Heat,” SFPE Handbook of Fire Protection Engineering, 4th edition, 2008.

7Peterson, J., and Stewart, R., “Predicting the carboxyhemoglobin levels resulting from carbon monoxide exposures,” Journal of Applied Physiology, Vol. 39, No. 4, October 1975.