Simulating imaging spectrometer data: 3D forest modeling based on LiDAR and in situ data

Abstract

Remote sensing offers the potential to study forest ecosystems by providing spatially and temporally distributed information on key biophysical and biochemical variables. The estimation of biochemical constituents of leaves from remotely sensed data is of high interest revealing insight on photosynthetic processes, plant health, plant functional types, and species composition. However, upscaling leaf level observations to canopy level is not a trivial task, in particular due to the inherent structural complexity of forests. A common solution for scaling spectral information is the use of physically-based radiative transfer models. We parameterize the Discrete Anisotropic Radiative Transfer (DART) model based on airborne and in situ measurements. At-sensor radiances were simulated and compared with measurements of the Airborne Prism Experiment (APEX) imaging spectrometer. The study was performed on the Laegern site (47°28"43.0 N, 8°21"53.2 E, Switzerland), a temperate mixed forest characterized by steep slopes, a heterogeneous spectral background, and a high species diversity. Particularly the accurate 3D modeling of the complex canopy architecture is crucial to understand the interaction of photons with the vegetation canopy and its background. Two turbid medium based forest reconstruction approaches were developed and compared; namely based on a voxel grid and based on individual tree detection. Our study shows that the voxel grid based reconstruction yields better results. When using a pixel-wise comparison with the imaging spectrometer data, the voxel grid approach performed better (R² = 0.48, 780 nm) than the individual tree approach (R² = 0.34, 780 nm). Spatial patterns as compared to APEX data were similar, whereas absolute radiance values differed slightly, depending on wavelength. We provide a successful representation of a 3D radiative regime of a temperate mixed forest, suitable to simulate most spectral and spatial features of imaging spectrometer data. Limitations of the approach include the high spectral variability of leaf optical properties between and within species, which will be further addressed. The results also reveal the need of more accurate parameterizations of small-scale structures, such as needle clumping at shoot level as well as leaf angle.