Mapping tropical biodiversity using spectroscopic imagery : characterization of structural and chemical diversity with 3-D radiative transfer modeling

Abstract:

The accelerating loss of biodiversity is a major environmental trend. Tropical ecosystems are particularly threatened due to climate change, invasive species, farming and natural resources exploitation. Recent advances in remote sensing of biodiversity confirmed the potential of high spatial resolution spectroscopic imagery for species identification and biodiversity mapping. Such information bridges the scale-gap between small-scale, highly detailed field studies and large-scale, low-resolution satellite observations. In order to produce fine-scale resolution maps of canopy alpha-diversity and beta-diversity of the Peruvian Amazonian forest, we designed, applied and validated a method based on spectral variation hypothesis to CAO AToMS (Carnegie Airborne Observatory Airborne Taxonomic Mapping System) images, acquired from 2011 to 2013. There is a need to understand on a quantitative basis the physical processes leading to this spectral variability. This spectral variability mainly depends on canopy chemistry, structure, and sensor’s characteristics. 3D radiative transfer modeling provides a powerful framework for the study of the relative influence of each of these factors in dense and complex canopies. We simulated series of spectroscopic images with the 3D radiative model DART, with variability gradients in terms of leaf chemistry, individual tree structure, spatial and spectral resolution, and applied methods for biodiversity mapping. This sensitivity study allowed us to determine the relative influence of these factors on the radiometric signal acquired by different types of sensors. Such study is particularly important to define the domain of validity of our approach, to refine requirements for the instrumental specifications, and to help preparing hyperspectral spatial missions to be launched at the horizon 2015-2025 (EnMAP, PRISMA, HISUI, SHALOM, HYSPIRI, HYPXIM). Simulations in preparation include topographic variations in order to estimate the robustness of image processing methods for mountainous ecosystems.