Signal variability and data compression considerations for petroleum forensics in two-dimensional gas chromatography

摘要

Disentangling the source-specific signature of petroleum biomarkers against the often stronger regional fingerprint has posed a fundamental challenge to oil-spill forensics in a petroleum-rich locale (e.g. Deepwater Horizon, Gulf of Mexico, April 2010). Two-dimensional gas chromatography (GC×GC) captures the joint peak distribution of hundreds of hydrocarbon biomarkers in crude petroleum through high-resolution images, which harbor a wealth of untapped information that can be exploited to achieve robust source differentiation. We harness this richness of two-dimensional gas chromatography (GC×GC) to solve this long-standing challenge using robust peak-mapping techniques. Specifically, we build upon recently proposed mapping techniques to investigate robustness of forensic pattern recognition amidst data uncertainties introduced by unknown ground truths and experimental variability.