Development of an optoelectronic tool for the detection of Cotonet de les Valls in citrus crops

Delottococcus aberiae , or “Cotonet de les Valls,” is an invasive pest responsible for substantial yield losses in citrus crops across the Mediterranean region of Spain. This study reports developing and validating a novel portable optoelectronic nose that integrates optical and electrochemical sensors for early pest detection in the field. The system demonstrated selectivity towards D. aberiae under both controlled and field conditions. Principal Component Analysis (PCA) revealed clear separation between infested and pest-free samples, and the ability to distinguish D. aberiae from other pests such as Bemisia tabaci . Partial Least Squares Discriminant Analysis (PLS-DA) models achieved perfect classification metrics (accuracy, precision, recall = 1.000) under in vitro conditions and high accuracy in field trials, even at early infestation levels. Partial Least Squares (PLS) regression models yielded R² values above 0.95 for predicting incidence and over 0.91 for severity when using a reduced 8-sensor array, significantly simplifying the system without compromising performance. These results highlight the robustness, accuracy, and field applicability of the optoelectronic system as an effective decision-support tool in integrated pest management programs. • The device detects D. aberiae in orchards before visible symptoms appear, enabling non-invasive and more effective diagnosis than traditional methods. • By accurately identifying the presence and severity of the pest, the system significantly reduces pesticide use, aligning with European policies on responsible phytosanitary practices and promoting more sustainable and precise citrus management. • The prototype was designed to be portable, low-cost, and easy to use by agricultural technicians, with simple installation in the tree canopy. It offers an innovative alternative to traditional methods, enabling faster and more sustainable decision-making. • The system was optimized by reducing the number of sensors from 14 to 8, maintaining—and even improving—predictive accuracy (R² > 0.95 for incidence and R² > 0.91 for severity), thereby lowering costs and simplifying the device without compromising performance.