Markiet, V; Mottus, M (2020). Estimation of boreal forest floor reflectance from airborne hyperspectral data of coniferous forests. REMOTE SENSING OF ENVIRONMENT, 249, 112018.
Abstract
In open forest canopies, such as in boreal forests, forest floor can contribute significantly to the observed top of canopy reflectance. In order to retrieve the biophysical properties of the tree layer, correcting for forest floor is essential. Traditionally, the algorithms for retrieval of forest floor reflectance depend on tree layer information such as leaf area index, canopy cover, and site fertility. To overcome these circular dependencies, we propose an algorithm that can be applied only using airborne remote sensing data. We acquired airborne hyperspectral imagery over the Hyytiala forest research station (61 degrees 50 ' N, 24 degrees 17 ' E) in central Finland on July 3rd in 2015 using a hyperspectral pushbroom line scanner. The image data had a spectral resolution of 4.6 nm, and the spatial resolution was 0.6 m. We developed a linear spectral unmixing algorithm, which is based on the definition of the reflectance factor, taking into account the variation of incident irradiance inside the canopy. The weights of the mixture can be computed from tree canopy gap fractions, a tree species insensitive leaf albedo, and average tree stand reflectance. Canopy gap fractions were retrieved with empirical methods available in scientific literature. The forest floor reflectance in the near-infrared increased with site fertility in agreement with the forest floor field measurements. Moreover, we found that in near infrared, the reflectance of moderately rich and moist upland forests was significantly different from all other fertility classes. Finally, we tested the reflectance decomposition on the photochemical reflectance index (PRI) known to be heavily affected by understory reflectance and canopy structure, and the forest PRI to be decoupled from the PRI of the overand understory.
DOI:
10.1016/j.rse.2020.112018
ISSN:
0034-4257