Scene Invariants for Imaging Spectrometer Uncertainty Quantification

Abstract:

Remote imaging spectroscopy investigations typically use Radiative Transfer Models (RTMs) to predict the measured radiance at the sensor given specific surface and atmosphere properties. Discrepancies between RTM model assumptions and physical reality can lead to subtle and systematic errors in derived measurements. We demonstrate an approach to characterize these model errors based on scene invariants - properties of the environment which are stable over time. Spatially and temporally invariant properties can disentangle different sources of retrieval variance, such as RTM model discrepancies and instrument noise. This allows investigators to estimate the aggregate uncertainty due to model discrepancies without having to explicitly identify them. The resulting distributions can be used to improve retrieval accuracy as well as the fidelity of posterior uncertainty predictions. We present the theoretical motivation for our approach, and initial results from a field dataset acquired by the AVIRIS-NG airborne imaging spectrometer.

Copyright 2020 California Institute of Technology. Government Support Acknowledged.