Carotenoids increase as an indicator of early stress of trees: estimations using Green Shoulder Indices from hyperspectral drone data and radiative transfer model PROSAIL

This study verifies the sensitivity of green shoulder indices calculated from hyperspectral images (490-550 nm wavelengths) to the carotenoid content of tree crowns. Five green shoulder indices were constructed and calculated from reflectance simulated by the radiative transfer model PROSAIL with varying parameters on chlorophyll, carotenoid, anthocyanin, dry matter, and leaf area index. Models to estimate pigment contents were built based on the linear relationships between pigment contents and vegetation indices. Results on simulated data showed that green shoulder indices had linear relationships with carotenoid content with R² ranging from 0.63 to 0.73 and linear relationships with carotenoid/chlorophyll ratio with R² ranging from 0.87 to 0.98. As a demonstration, models to retrieve pigment content were implemented on real-world data collected by a hyperspectral drone covering a forest infested by spruce bark beetle during early phase four times. The results showed that, with a longer time of infestation, i.e., increasing stress levels, the carotenoid content increased while chlorophyll content decreased. Using estimated carotenoid content improved the separation between healthy and infested trees compared to using estimated chlorophyll content solely, which explained the better capacity of green shoulder indices on stress detection than red-edge indices. Overall, this study exhibits carotenoid increase as a strong indicator of initial vegetation stress and highlights the potential of green shoulder indices in identifying early stress based on their sensitivity to carotenoid content.