DART radiative transfer modelling for sloping landscapes

Topography is one of the key factors that impact remotely sensed data and their interpretation. Indeed, combined with the viewing geometry and neighbour effects, it strongly affects the direct, diffuse and multi-scattered scene irradiance, which in turn impacts the radiative budget and remote sensing signals of the landscapes. The increased availability of digital elevation models (DEM) and the advancement of 3D radiative transfer (RT) models allow us to better address these topographic effects. DART (Discrete Anisotropic Radiative Transfer) is one of the most accurate and comprehensive 3D RT models that simulate remote sensing observations of natural and urban landscapes with topography and atmosphere. It simulates environmental effects (i.e., impact of adjacent landscape on the observed landscape) using a so-called infinite slope mode that infinitely duplicates the observed landscape while ensuring the continuity of slope and altitude at the DEM edges. Up to DART version 5.7.4, this mode was slightly inaccurate and computer intensive, depending on the topography. This paper presents an innovative modelling strategy that greatly improves it in terms of accuracy, image quality and computer efficiency. For that, a fictive auxiliary oblique plane, adapted to the landscape topography, is introduced for managing the scene illumination, the Earth-Atmosphere coupling and the storage of the radiation that exits the scene before being projected onto the sensor plane. Improvements and validations are illustrated both visually and quantitatively by DART images, radiometric products and radiative budget. For example, the observed reflectance of a Lambertian slope is equal to the expected analytical value. In addition, the solar plane reflectance of a forest on a mountain slope (experimental scene) has an average error of about 0.01% relative to the reflectance of the same forest stand in the reference scene (i.e., nine duplications of the experimental scene). This new modelling is already integrated in the official DART version (https://dart.omp.eu).