Leveraging electronic nose technology for early detection of primary stored product pest infestations

Early detection of stored product pest infestations is crucial for pest management. However, detecting primary pests, like Rhyzopertha dominica (Coleoptera: Bostrichidae) and Sitophilus zeamais (Coleoptera: Curculionidae) is often challenging because their larvae feed and develop within wheat kernels. Without early detection and treatment, primary pests contribute to the global postharvest economic loss of 1.3 billion tonnes of food. The electronic nose is a multivariate sensor array that may be able to take advantage of machine learning and pattern recognition software to identify the odors associated with specific insects, grain, and infestations, leading to earlier detection. Here, we investigated the state-of-the-art electronic nose, MSEM-160 (Sensigent, Baldwin Park, U.S.A.), and its ability to detect R. dominica and S. zeamais in wheat. Both pests were reared on hard red winter wheat tempered to 13% for eight weeks. One-hundred-gram wheat samples at different ratios of clean:infested grain were tested with supervised and unsupervised machine learning, including principal component analysis, canonical discriminatory analysis, and random forest. These were used to differentiate the pest infestation density and to identify the species infesting the wheat. We found that the electronic nose reliably recognized infestations of both species in wheat. This work has clear implications in reducing postharvest loss throughout the supply chain by detecting cryptic infestations of primary stored product pests in stored wheat. In the future, additional stored product pests and additional commodities should be evaluated.