Semi-empirical Modeling of the Scene Reflectance of Snow-Covered Boreal Forest: Validation with Airborne Spectrometer and LIDAR Observations

This work aims at the development and validation of a zeroth order radiative transfer (RT) approach to describe the visible band (555nm) reflectance of conifer-dominated boreal forest for the needs of remote sensing of snow. This is accomplished by applying airborne and mast-borne spectrometer data sets together with high-resolution information on forest canopy characteristics. In case of aerial spectrometer observations, tree characteristics determined from airborne LIDAR observations are applied to quantify the effect of forest canopy on scene reflectance. The results indicate that a simple RT model is feasible to describe extinction and reflectance properties of both homogeneous and heterogeneous forest scenes (corresponding to varying scales of satellite data footprints and varying structures of forest canopies). The obtained results also justify the application of apparent forest canopy transmissivity to describe the influence of forest to reflectance, as is done e.g. in the SCAmod method for the continental scale monitoring of fractional snow cover (FSC) from optical satellite data. Additionally, the feasibility of the zeroth order RT approach is compared with the use of linear mixing model of scene reflectance. Results suggest that the non-linear RT approach describes the scene reflectance of a snow-covered boreal forest more realistically than the linear mixing model (in case when shadows on tree crowns and surface are not modeled separately, which is a relevant suggestion when considering the use of models for large scale snow mapping applications).