Advancing retrievals of surface reflectance and vegetation indices over forest ecosystems by combining imaging spectroscopy, digital object models, and 3D canopy modelling

Abstract

Imaging spectroscopy based methods offer unique capabilities for retrieving narrow-band vegetation indices which can be empirically related to functional traits of plants. However, in areas with complex topography, illumination effects affect the retrieval of such indices from high spatial resolution airborne or satellite data. Irradiance components at the pixel level are determined by atmospheric composition, as well as instantaneous illumination-surface-sensor geometries. An accurate pixel-wise description of direct and diffuse irradiance components is necessary to perform atmospheric corrections, finally resulting in improved surface reflectances and hence products. We assess three atmospheric correction strategies, differing in their approaches to simulate instantaneous as well as pixel-wise abundances of diffuse and direct irradiance. We use physically-based approaches in combination with either digital elevation models (DEM), fine resolution digital object models (DOM), or 3D modelling output from the Discrete Anisotropic Radiative Transfer (DART) model. The such obtained top-of-canopy reflectances at the Laegern test-site in Switzerland, are used to assess retrieval improvement for a set of indices (Normalized Difference Vegetation Index (NDVI), Photochemical Reflectance Index (PRI), as well as chlorophyll and carotenoid indices). We demonstrate that both, the DOM and the DART based approach, improve the retrievals for flat cast-shadows by ≤71% compared to using a DEM. In dense forest areas, improvements are less significant. Remaining key issues are related to overestimating surface reflectance under extreme illumination conditions.