Chemical Source Classification in Naturally Turbulent Plumes

Abstract

The spatiotemporal dynamics of chemical plumes in natural environments imposes complex time-varying responses on chemical detectors, challenging the use of conventional chemical analysis instrumentation, which often relies upon precisely controlled sampling of analytes. Insects take a different approach to this problem, typically exposing their diversity of chemical receptors to the full extent of the space-time dynamics inherent to these environments. Here we adopt a similar approach, by exposing differentially tuned chemosensor arrays to analytes dispersed in naturally turbulent chemical plumes from a point source. We propose a novel sensor preprocessing metric for complex time-varying chemosensor responses, termed sensor variation, which after normalization generates a stable array response fingerprint representative of the analyte generating the response, and is invariant over a range of distances from the source and source volumes. When applied to chemosensor array response time series, the resultant fingerprints are demonstrated to reliably support chemical classification of a group of pure analytes advected from a point source. By comparing classification performance to the same analytes at equivalent concentrations in controlled sampling conditions, we show that chemical source classification can be achieved in turbulent chemical plumes with similar accuracy to controlled experimental conditions.