Interpretation of Ignitable Liquid Residues in Fire Debris Analysis: Effects of Competitive Adsorption, Development of an Expert System and Assessment of the False Positive/Incorrect Assignment Rate

• NCJ Number: 248242
• Author(s): J. Graham Rankin Ph.D.; Nicholas Petraco
• Date Published: 2014
• Length: 153 pages
• Grant Number(s): 2010-DN-BX-K272
• Agencies: NIJ-Sponsored

Annotation

The study had three major goals: 1. investigate the effect of competitive adsorption of substrates typically found in fire debris on the classification by ASTM E1618 methodology, 2. develop and validate an expert system for assisting forensic analysts in classifying ignitable liquid residues in fire debris, and 3. provide statistical evaluation of error rates for experienced fire debris analysts in reviewing "case files" of gas chromatograph-mass spectrometry data for identifying and classifying ignitable liquid residues in simulated fire debris.

Abstract

The study confirmed that charred yellow pine can affect the distribution and abundance of ignitable liquid components, especially the normal hydrocarbons present in petroleum distillates. Substrates used as comparison samples may have additional incidental compounds which may be lost or reduced when charred. The expert system was developed using the open-source statistical software to be independent of vendor gas chromatography-mass spectrometry software. It was based on analyses of over 500 reference liquids from the Ignitable Liquid Reference Collection developed by the Technical Working Group for Fire and Explosions and maintained by the National Center for Forensic Science. Cross-validation of the system demonstrated correct assignment to E1618 class of 95% for most samples. The Oxygenate and Miscellaneous classes had slightly higher error rates due to the variability within these classes. Identification of these classes often requires mass spectral identification of individual components in the chromatogram. However, a series of challenge samples created by spiking commercial ignitable liquids onto charred substrates proved that the developed system needs more testing and refinement before adoption for actual case work. The statistical evaluation of experienced fire debris analysts' error rates in identifying the presence and classifying the ignitable liquid residues showed no false positives or miss-classification. False negatives or "no classification possible" were not considered errors. However, because the sample size was smaller (20 participants) than originally proposed, additional work is needed before a definitive error rate can be established. The pour pattern sampling study demonstrated that for porous materials, sampling near the center of the pour pattern had a higher recovery of ignitable liquid residues than at the edges. The "donut effect" previously observed and the way the fire is self-sustained by the substrate beyond the limit of the original pour are believed to be responsible for these results.