Spectral Correction for the Effect of Crown Shape at the Single-Tree Level: An Approach Using a Lidar-Derived Digital Surface Model for Broad-Leaved Canopy

Abstract

The spectral response of a single tree on remotely sensed images may be affected by the crown shape, yielding brighter sun-facing sides and darker sides facing away from the sun. To correct this crown-shape effect, we assumed that the roughness of the canopy surface within a single crown resembles the topography of the land surface, in the case of a closed upper-canopy layer (i.e., canopy surface) of broad-leaved deciduous trees. Previous topographic normalizations have proved to be useful in correcting the effect of rugged terrain at a large-landscape scale. However, are they also useful in correcting the rough canopy surface at the scale of a single crown? We applied conventional topographic corrections (cosine, Minnaert, and C) to a canopy surface model (CSM) of 66 trees, which was derived from airborne small-footprint Lidar data. The 12-band spectral data of each tree were acquired from the aircraft with 0.5-m spatial resolution and analyzed to validate the corrections. Among the tested methods, the C correction was the most successful in applying to the CSM, whereas the Minnaert and cosine corrections yielded errors in the reflectance of some steep canopy surfaces. We assessed the spectral feature of each tree in terms of tree vigor, by using the vegetation index employing the red-edge (700 nm) and green (541 nm) bands. Following the correction, the overall accuracy was improved from 75.5% to 84.5%. Thus, we suggest that the effect of crown shape be carefully considered and corrected in single-crown-level analyses of remotely sensed spectra.