Tree trunk disease mapping using hyperspectral LiDAR

• Proposing a fully automated detecting method for tree trunk disease mapping using a tunable hyperspectral LiDAR with a 2-axis rotator. • Lambertian–Beckmann model is used to correct the reflectivity of the trunk at different incident angles. • Mapping the defected and non-defected areas of trunks by random forest algorithm with a high classification accuracy of up to 0.84. • A novel design of hyperspectral LiDAR system with ultra-broadband and simplest optics is created to assist forest disease mapping and monitoring. Forest diseases are a major disturbance affecting forest ecosystems due to climate change. Traditionally, most tree disease monitoring requires time-consuming and costly manual field investigation and assessment, which is prone to human mistakes and has limited spatial and temporal resolutions. Here, for the first time, we report an automated and objective method to non-invasively assess tree diseases with a hyperspectral LiDAR (HSL) scanning system. We measured tree trunks infected with wood decay diseases ( Ailanthus altissima infected by Porostereum spadiceum , Amygdalus davidiana by Phellinus tuberculosus , and Styphnolobium japonicum by Stereum peculiare ) using this HSL system. This HSL scan system employed a wavelength spectrum of 550–1000 nm with a 5 nm spectral resolution. This hyperspectral mapping system, assisted with a machine-learning algorithm, offers significant advantages in mapping the defected and non-defected areas of trunks, such as high classification accuracy of up to 0.842, 0.803, and 0.802, respectively, and full automation. A novel HSL system aimed at forest mapping was designed in accordance with the HSL performance and experimental results. Our results indicate that the hyperspectral LiDAR is an effective tool for high-performance forest growth monitoring and diseased mapping in accuracy, large area, objective detection, and high resolution (at the tree trunk and stem levels).