Temporal Analysis of Electronic Nose (E-nose) Detection of Stink Bug-Injured Cotton Bolls Under Laboratory and Field Conditions

Management decisions for stink bugs (Pentatomidae) in B.t. cotton are often complicated by time-consuming and destructive sampling methods, and there is a need for more effective and reliable detection tools.  Stink bug feeding results in the induction of volatile organic compounds (VOCs) from cotton bolls, and the blend produced is different from that of healthy undamaged bolls.  Electronic nose (E-nose) technology has been shown to be a promising tool for the detection of stink bug injury.  In this study we examined the temporal variation in E-nose detection of stink bug-injured bolls over a four-day exposure period.  In laboratory and field trials, stink bugs were caged on bolls (10-12 days post anthesis) and allowed to feed ad libitum.  In 24-hour intervals stink bugs were removed from bolls, and E-nose was trained to differentiate VOCs from injured and healthy bolls.  Canonical projection plots and principal components analysis of E-nose sensor data showed distinct separation between stink bug-injured and healthy bolls after two-days of exposure under laboratory and field conditions.  Cross validation of sensor data indicated that E-nose was 100% accurate at differentiating injured and healthy bolls in both laboratory and field trials.  However, when E-nose training sets were used to randomly classify known samples of injured and healthy bolls, results were slightly less accurate (90% and 85% correct identifications in lab and field trials respectively).  These data indicate that E-nose is a promising technology for rapid, non-destructive detection of stink bug injury to cotton bolls.  With further modifications of sensors and/or sampling procedures, this technology may result in more reliable tools for management decisions.