Tree growth rates are an important component of forest productivity. Commonly, measurements of stand productivity come from diameter measurements of inventoried plots at two or more points in time. While useful, plot-level measurements may not represent the variation that occurs on the larger landscape because of spatial variability in conditions and forest composition. Airborne hyperspectral remote sensing (AHRS) is a powerful tool for estimating forest productivity because it has high spectral and spatial resolution capable of detecting canopy traits of individual tree crowns continuously over large areas. However, it is unclear if tree growth rates can be accurately monitored using AHRS. Our objective was to determine if tree growth rates of multiple species could be accurately predicted from a single acquisition of AHRS data. Our central hypothesis is that tree growth can be estimated across species with canopy reflectance because tree canopies have converged on a set of properties, detectable from hyperspectral data, that optimize growth and survival given the constraints and stressors of the environment.
Results/Conclusions
We measured individual tree diameter growth from tree cores of canopy trees eastern Alabama (n=151). For the same trees, we extracted the canopy reflectance from hyperspectral aerial images, which were collected by the NEON Airborne Observation Platform. We performed a partial least squares regression on mean canopy reflectance and mean annual basal area increment (BAI). For the hardwood species, mean BAI for 5 years had a mean of 14 sq cm/yr and ranged from 5 - 30 sq cm/yr with Shagbark Hickory and Sweetgum the slowest and fastest growing species, respectively. Mean canopy reflectance alone explained 54% of the variation in growth rates across species and sites, with a RMSE of 4.5 sq cm/yr. We found patterns between reflectance and growth primarily in the visible to red-edge and shortwave infrared regions of the spectrum. This work is important because it suggests that a single aerial hyperspectral image can be used to estimate differences in tree productivity among individual trees. It not only provides information about stand productivity, but could also be used to target management efforts for specific trees and stands.