Airborne LiDAR-Based Mortality Monitoring of Citrus Tree Decline under Huanglongbing

Abstract. Huanglongbing (HLB)-related decline has caused severe canopy degradation and tree mortality in citrus orchards, creating a need for timely and interpretable tree-level monitoring methods. This study evaluated airborne LiDAR-derived vertical structural information for monitoring dead citrus trees under HLB-related decline. Tree-level point clouds from two orchard fields in southwest Florida were converted into 20-bin normalized vertical return profiles, from which four interpretable descriptors were derived: vertical return centroid, upper-canopy retention, lower-profile allocation, and vertical redistribution. These descriptors were compared with the full normalized vertical profile using threshold-based classification, logistic regression, and random forest. Dead trees showed a distinct downward redistribution of LiDAR returns, with reduced upper-canopy retention and increased lower-profile allocation. Individual descriptors achieved strong discrimination performance, with AUC values of 0.987–0.995, approaching the full-profile models with AUC values of 0.996–0.999. Cross-field transfer between DKSB and BTMV provided preliminary evidence that the mortality-related vertical structural signal may be consistent across orchard fields. These results suggest that LiDAR structural descriptors can support interpretable tree-level mortality monitoring and dead-tree screening in citrus orchards.