Peak-cognizant Signal Processing of Raw Instrument Signals to Quantify Environmental Weathering of Contaminants from the Deepwater Horizon Spill

In this work, we present peak-cognizant quantification of environmental weathering of crude oil from the from the Deepwater Horizon oil spill. The key idea is to autonomously extract peak information from raw gas chromatography-mass spectrometry (GC-MS) signals from crude oil samples, and represent the relative weathering of different peaks in a graph-based quantitative computational framework. We also present results from pre-processing the raw signals with baseline correction and signal normalization. Retention time alignment is performed by first aligning the source oil by determining the retention time drift between prominent peaks within the signals and applying the calculated drift to the weathered oil samples. Peak finding, validation, and grouping of the five weathered oil samples to a source oil sample allows compound associations to be discovered. We present preliminary results as graphical visualizations allowing for rapid and precise interpretation of weathering compounds within polycyclic aromatic hydrocarbons (PAH). Results presented were generated with oil samples showing different degrees of weathering collected from the Deepwater Horizon spill.