Predicting lidar measured forest vertical structure from multi-angle spectral data

A capability to remotely measure the vertical and spatial distribution of forest structure is required for more accurate modeling of energy, carbon, water, and climate over regional, continental, and global scales. We examined the potential of using a multi-angle spectral sensor to predict forest vertical structure as measured by an airborne lidar system. Data were acquired from AirMISR (Airborne Multi-Angle Imaging Spectrometer) and airborne LVIS (Laser Vegetation Imaging Sensor) for a 7000 ha study site near Howland Maine, consisting of small plantations, multi-generation clearings and large natural forest stands. The LVIS data set provided a relatively direct measure of forest vertical structure at a fine scale (20 m diameter footprints). Multivariate linear regression and neural network models were developed to predict the LVIS forest energy height measures from 28 AirMISR multi-angle spectral radiance values. The best model accurately predicted the maximum canopy height (as measured from LVIS) using AirMISR data (rmse = 0.92 m, R2 = 0.89). The models developed in this study achieved high accuracies over a study site with an elaborate patchwork of forest communities with exceptional diversity in forest structure. We conclude that models using MISR-like data are capable of accurately predicting the vertical structure of forest canopies.