MOF-enhanced colorimetric sensor array for early detection of citrus infestation by Bactrocera dorsalis

Early detection of Bactrocera dorsalis ( B. dorsalis ) infestation in citrus is challenging due to the asymptomatic nature of initial larval stages. Colorimetric sensor arrays (CSAs) are promising for non-destructive detection of B. dorsalis infestations in citrus, yet they still face limitations in reliably identifying the key volatile organic compounds (VOCs) released by infested citrus. In this study, we developed a MOF-enhanced colorimetric sensor array integrating pH-sensitive dyes into ZIF-7 nanoparticles for non-invasive detection. The array detected VOCs associated with early infestation, including D -limonene, decanal, and linalool, identified via GC-MS. The dye/ZIF-7 composite CSA exhibited enhanced sensitivity compared to the pure dye sensor array. Furthermore, across the 4-day test period, infested samples registered a progressive rise in total Euclidean distance (ED) of color change, reaching a 171.7-unit increase over healthy controls on day 4. To enhance the accuracy and efficiency of infestation detection, we implemented a machine-learning pipeline using Dung Beetle Optimizer (DBO) for feature selection and Tree-structured Parzen Estimator (TPE) for hyperparameter tuning. Consequently, the XGBoost classifier achieved a peak test accuracy of 97.15 %, enabling reliable detection during the early stage of infestation. Overall, this study offers a non-destructive and cost-effective approach for batch-mode early detection of B. dorsalis infestation in citrus, advancing the potential for enhanced postharvest monitoring and pest management strategies.