Retrieving structural and chemical properties of individual tree crowns in a highly diverse tropical forest with 3D radiative transfer modeling and imaging spectroscopy

Author(s)

Ferreira, Matheus Pinheiro ^[1] ; Feret, Jean-Baptiste ^{[2, 3]} ; Grau, Eloi ^{[2, 3]} ; Gastellu-Etchegorry, Jean-Philippe ^[4] ; Shimabukuro, Yosio Edemir ^[1] ; de Souza Filho, Carlos Roberto ^[5]

Affiliation

^[1] Natl Inst Space Res, Remote Sensing Div, Av Astronautas 1758, BR-12227010 Sao Jose Dos Campos, SP - Brazil

^[2] Univ Montpellier, CNRS, CIRAD, TETIS, Irstea, AgroParisTech, Montpellier - France

^[3] Maison Teledetect, 500 Rue Jean Francois Breton, F-34000 Montpellier - France

^[4] Univ Toulouse, CNRS, IRD, Ctr Etud Spatiales BIOsphere CESBIO, UPS, CNES, F-31401 Toulouse 9 - France

^[5] Univ Estadual Campinas, Geosci Inst, R Joao Pandia Calogeras 51, BR-13083870 Campinas, SP - Brazil

• Document type: Journal article
• Source: REMOTE SENSING OF ENVIRONMENT; v. 211, p. 276-291, JUN 15 2018.
• Web of Science Citations: 6

Abstract

Spatial and temporal information on the structural and chemical properties of tropical forest canopies are key to understanding ecosystem processes. However, such information is usually limited to field studies performed at the plot level (similar to 1 ha). The combination of imaging spectroscopy with physically based radiative transfer (RT) models holds great promise for generalizing and extrapolating insights from plot-based studies to whole landscapes. Here, we tested the capacity of a simplified 3D RT approach to retrieve the structural and chemical traits of individual tree crowns (ITCs) from a highly diverse tropical forest. We first produced two datasets called measured and simulated. The measured dataset was composed of ITC reflectance extracted from sunlit imaging spectroscopy pixels. The simulated dataset was produced using a look-up-table approach and the discrete anisotropic radiative transfer (DART) model. We then compared the simulated and measured reflectances of ITCs in terms of shape difference by computing the spectral angle. The results showed small disagreements between the simulated and measured reflectances. Such differences impacted neither the spectral variability nor the spectral regions recognized as useful for species discrimination, showing that the spectral angle was a suitable measure of spectral similarity. Simulation robustness was assessed by comparing model parameters obtained by inversion to imaging spectroscopy vegetation indices and the proportion of non-photosynthetic vegetation (NPV), green photosynthetic vegetation (GV) and shade estimated within ITCs. DART canopy structural parameters were related to NPV (R-2 = 0.71), GV (R-2 = 0.78) and shade (R-2 = 0.55). DART canopy foliar parameters such as chlorophyll and carotenoids were related to the ratio of TCARI/OSAVI (R-2 = 0.80) indices and the simple ratio between reflectances at 515 nm and 570 nm (R-515/R-570) (R-2 = 0.54), respectively. Species-related differences in NPV, GV and shade were explained by variations in crown architectural characteristics. The simulation framework employed in this study can be applied to retrieve structural and chemical traits of ITCs from other areas in which high-resolution imaging spectroscopy data are available. (AU)

• FAPESP's process: 15/50484-0 - Functional diversity of intact and regenerating Amazon, Atlantic, and Cerrado systems using hyperspectral imagery
• Grantee: Fabien Hubert Wagner
• Support type: Research Grants - Young Investigators Grants
• FAPESP's process: 16/24977-1 - Measuring and mapping Atlantic Forest canopy foliar traits with hyperspectral images and radiative transfer modeling
• Grantee: Matheus Pinheiro Ferreira
• Support type: Scholarships in Brazil - Post-Doctorate