Quantifying uncertainties in compositional mapping with imaging spectroscopy

Abstract:

Estimating surface and atmospheric attributes with fine spectral structures, requires both precise and accurate retrievals of surface reflectance. We present an expansion of the physical model to account for adjacency effects, sub-pixel multiple scattering, and shadows. Using the OE framework, we demonstrate how it is possible to evaluate different physical models while adjusting the surrounding statistical assumptions to achieve a controlled environment. Our new formulation would advance from estimating the hemispherical-directional reflectance factor (HDRF) to a new quantity of reflectance more intrinsic to the materials within the pixel. Preliminary results show promise in multiple challenging environments, including vegetation and shade. This approach is readily adapted into current open-source implementations of OE, and has the potential to increase the capacity of imaging spectroscopy in traditionally challenging scientific applications.