Mapping three-dimensional variation in leaf mass per area with imaging spectroscopy and lidar in a temperate broadleaf forest

Imaging spectroscopy is a valuable tool for mapping canopy foliar traits in forested ecosystems at landscape and larger scales. Most efforts to date have involved two-dimensional mapping of traits, typically representing top-of-canopy conditions. However, traits and their associated biological functions vary through the canopy vertical profile, such that incorporating information about vertical patterns may improve modeling of ecosystem processes like primary productivity. In 2016 and 2017, we collected extensive field data in forests in Domain 5 (Great Lakes) of the National Ecological Observatory Network (NEON) to characterize the vertical variation in leaf mass per area (LMA), an important foliar trait related to plant growth and defense. Fieldwork was coincident with NEON Airborne Observation Platform (AOP) overflights which collected imaging spectroscopy and lidar data. Using imaging spectroscopy to map top-of-canopy LMA and lidar to model vertical gradients of transmittance, we developed a method to map three-dimensional patterns in LMA in temperate broadleaf forests. Partial least squares regression (PLSR) was used to estimate top-of-canopy LMA (R-2: 0.57, RMSE 10.8 g m(-2)), which, along with lidar-derived metrics of light transmittance and height, was used in a multilevel regression to model within-canopy LMA (R-2: 0.78, RMSE 8.3 g m(-2)). The coupled models accurately estimated LMA throughout the canopy without taking into account species composition (R-2 = 0.82, RMSE: 8.5 g m(-2)).