Combining remotely sensed estimates of structure and function to improve quantification of forest productivity

Gross primary productivity (GPP) describes the total photosynthesis (i.e., carbon fixation) for an ecosystem and is an important component of the global land carbon budget. Accurate measurements of forest carbon sequestration are crucial, as climate, soils, and management practices influence carbon uptake. Disturbances such as wildfires, diseases, insect infestations, and extreme weather events significantly impact forest health and functionality. Remote sensing is useful for monitoring global ecosystems and estimating forest vitality. High-resolution images and lidar (Light Detection and Ranging) data offer enhanced insights into forest light utilization. Combining lidar with spectral data provides a comprehensive view of forest ecosystems, integrating both structural and physiological information. This work includes three separate studies under a common objective to establish an improved understanding of forest productivity in different forest compositions using a combination of physiological function from optical imagery and morphological structure from lidar. The first study investigates the photochemical reflectance index (PRI) and how illumination affects the diurnal and vertical distributions of two managed pine stands with varying ages and row orientations, using airborne hyperspectral and lidar data. We developed a novel method to classify canopy illumination into sunlit, shaded, and mixed light areas using the hyperspectral data and a simulated panchromatic band. PRI values varied between sunlit and shaded foliage throughout the day, reflecting differences in foliage efficiency depending on light conditions and forest structure. The second study evaluates how well remotely quantified plant functional traits predict GPP across U.S. forests. Using data from NEON's airborne remote sensing and in situ flux tower measurements, the study assessed hyperspectral optical indices as physiological traits and lidar-derived products as morphological and environmental traits. By developing multiple linear regression models with separated and combined trait groups, the best prediction model combined morphological, environmental, and physiological traits with a PRESS R2 of 0.75. This underscores the importance of integrating various functional traits for accurate forest productivity predictions. The last study detects insect-induced tree mortality with separated and combined models of structure from lidar and physiology from satellite imagery. Although all models tend to overestimate tree mortality, integrating lidar data enhances predictions by 6% offering valuable structural context. A central theme of this work is that lidar-derived structural measurements were crucial in every chapter.