Radiative Transfer Modelling using Highly Detailed 3D Models from Terrestrial Laser Scanning

Abstract:

Accurate simulations of light transmission within forests is essential for forestry remote sensing. It is of great significance for promoting the validation of remote sensing products. Radiative Transfer Models (RTMs) are physical models capable of simulating the interaction between light and forest structure. Traditional RTMs often use simplified forest structures such as describing canopy as homogeneous turbid medium or 3D archetypes, which can lead to bias in the radiative transfer simulation. RTMs using highly detailed and realistic forest structure can offer a solution to simulate light transmission within forest more realistically. One promising approach to reconstruct highly detailed forest structural models is terrestrial laser scanning (TLS). Forest structural models can be reconstructed from TLS point cloud data in different ways. Radiative transfer simulation is impacted by the detail level of forest structural models derived from TLS. The objective of this study is to evaluate this impact. To achieve this objective, we conducted radiative transfer simulations by using highly detailed 3D forest models. A one-hectare forest stand of the deciduous forest in Wytham Woods (Oxford, UK) was scanned by TLS. Two 3D forest models, a 3D explicit Quantitative Structure Model (QSM) and a voxel model, were built based on the TLS point cloud data. 559 individual trees were reconstructed for each forest model. The reconstructed 3D forest models were used as structural input for the RTMs. The Discrete Anisotropic Radiative Transfer (DART) model was used in this study. Radiometric properties were allocated according to the species of each tree. The performance of these RTMs was evaluated by efficiency and accuracy. The efficiency was evaluated by required RAM and computational time. The accuracy was evaluated by simulated reflectance value.