Authors: Janet M. Crespo Cajigas, MSa, Vidia A. Gokool, Ph.D.b,c, Howard K. Holness, Ph.D.a,
Kenneth G. Furton, Ph.D.a, Lauryn E. DeGreeff, Ph.D.a

aFlorida International University, Global Forensic and Justice Center, Chemistry; bLawrence Livermore National Laboratory, Nuclear and Chemical Sciences Division; cFormerly of Florida International University, Global Forensic and Justice Center, Department of Chemistry and Biochemistry

There have been many uncertainties related to the canine detection of marijuana due to the widespread, ongoing legalization of Cannabis sativa (i.e. marijuana) throughout the United States of America (USA). This sociolegal issue has resulted in an increasing interest regarding the generalization and specification within canine detection of marijuana. More specifically, canine trainers have expressed concerns on whether canines can generalize on the odor of marijuana regardless of the origin of their training materials. This study aims to differentiate multiple marijuana samples from a variety of regions in the USA based on the volatile organic compounds (VOCs) found in their odor profile. Some of the features to be examined for each sample group are the location of origin and whether the marijuana was confiscated by law enforcement at the time of sampling. Previous work in this area has shown promise towards the potential differentiation of several marijuana varieties. In this research, a heated headspace (HS) solid phase micro-extraction (SPME) technique is implemented for the collection of VOCs originating from marijuana. The headspace samples were analyzed using full-scan, optimized methods on a gas chromatograph coupled to a mass spectrometer (GC-MS), and a variety of chemometric tools were applied to the data to enable differentiation and potential classification between sample populations. Principal component analysis in addition to multiple supervised analysis methods employed in this study have demonstrated a disparity between marijuana varieties using the VOCs extracted from their odor profile. With this information, potential biomarkers that are common between samples were preliminarily chosen to develop a targeted analysis for differentiation. The performance metrics for both approaches were evaluated and compared in this study. This work expands on the foundation already established in SPME-GC-MS research regarding the differentiation of similar-like substances of forensic and agricultural interest such as hemp and marijuana or distinct varieties of foodstuffs. Moreover, it looks to fill in the gaps in marijuana detection research and if necessary, could lead to adjustments in future canine scent detection training of marijuana to improve the generalization of their target.