Influence of 3D Spruce Tree Representation on Accuracy of Airborne and Satellite Forest Reflectance Simulated in DART

Abstract

Advances in high-performance computer resources and exploitation of high-density terrestrial laser scanning (TLS) data allow for reconstruction of close-to-reality 3D forest scenes for use in canopy radiative transfer models. Consequently, our main objectives were (i) to reconstruct 3D representation of Norway spruce (Picea abies) trees by deriving distribution of woody and foliage elements from TLS and field structure data and (ii) to use the reconstructed 3D spruce representations for evaluation of the effects of canopy structure on forest reflectance simulated in the Discrete Anisotropic Radiative Transfer (DART) model. Data for this study were combined from two spruce research sites located in the mountainous areas of the Czech Republic. The canopy structure effects on simulated top-of-canopy reflectance were evaluated for four scenarios (10 × 10 m scenes with 10 trees), ranging from geometrically simple to highly detailed architectures. First scenario A used predefined simple tree crown shapes filled with a turbid medium with simplified trunks and branches. Other three scenarios used the reconstructed 3D spruce representations with B detailed needle shoots transformed into turbid medium, C with simplified shoots retained as facets, and D with detailed needle shoots retained as facets D. For the first time, we demonstrated the capability of the DART model to simulate reflectance of complex coniferous forest scenes up to the level of a single needle (scenario D). Simulated bidirectional reflectance factors extracted for each scenario were compared with actual airborne hyperspectral and space-borne Sentinel-2 MSI reflectance data. Scenario A yielded the largest differences from the remote sensing observations, mainly in the visible and NIR regions, whereas scenarios B, C, and D produced similar results revealing a good agreement with the remote sensing data. When judging the computational requirements for reflectance simulations in DART, scenario B can be considered as most operational spruce forest representation, because the transformation of 3D shoots in turbid medium reduces considerably the simulation time and hardware requirements.

Monitoring LAI, Chlorophylls, and Carotenoids Content of a Woodland Savanna Using Hyperspectral Imagery and 3D Radiative Transfer Modeling

“…The validity of a simplified representation of trees (simple forest representation, SFR) within the DART scene (such as the one done by Gascon et al [14]) when working with medium-resolution hyperspectral images of very sparse forests was tested in this study, based on the work presented in Widlowski et al [32]. As acceptable representation of coniferous trees within RTM is still a difficult topic [34]; this study only focused on broadleaved trees as a preliminary test case to study Mediterranean forests. Validation data were available at multiple dates, allowing for a more robust calibration of the inversion method.…”

Phenotyping Plant Responses to Biotic Stress by Chlorophyll Fluorescence Imaging

“…Nevertheless, the application of SiF and hyperspectral reflectance on plant phenotyping need of further research on the effects of environmental factors (i.e., incoming PAR, viewing solar geometry, direct-to-diffuse light ratios, air temperature, and wind), canopy heterogeneity, and architecture on the remote measurements. In this sense, advances have been made in the last year, developing 3D models of canopies at different spatial resolutions by a range of methodologies (Gastellu-Etchegorry et al, 2017; Janoutova et al, 2019; Roitsch et al, 2019). Particularly in the case of hyperspectral reflectance, the applicability of spectral indices for predicting photosynthetic activity and GPP should be explored across different canopy types and temporal and spatial scales.…”

Sensitivity Analysis of the DART Model for Forest Mensuration with Airborne Laser Scanning

“…In this latter mode of operation, DART can replicate structurally complex forests through the direct importation of 3D trees generated in external modeling software [183,184] or trees digitized from terrestrial laser scanning measurements [124,185]. Recent work by Roberts et al [186] and Janoutová et al [187] suggests that forest biometrics may be reliably inverted using simplified trees with foliage represented by turbid media. However, in the present study, we seek to minimize geometric assumptions associated with simplified crowns, therefore ALS acquisitions are simulated using trees with explicitly defined crown geometries and reflectance characteristics.…”

“…Our study therefore highlights the complexities of: (i) generating synthetic trees that are biologically realistic, and (ii) parameterizing tree modeling software such as OnyxTree, Arbaro [240] or AMAPstudio [184] based on manual field measurements. In future investigations, it may be preferable to perform radiative transfer simulations using trees digitized directly from terrestrial laser scanning measurements [99,124,185,187] or digital photogrammetry [241] since these techniques can enable the importation of biologically realistic trees.…”