Uncertainty Propagation Analysis of the airborne hyperspectral data processing chain

The processing of airborne hyperspectral imagery introduces uncertainty. In order to quantify the uncertainty in the resulting hyperspectral imagery, a concept for Uncertainty Propagation Analysis (UPA) was developed and applied. The UPA entails the Monte Carlo stochastic simulation of uncertain components of the Processing and Archiving Facility (PAF), resulting in a chain of Monte Carlo analyses. First, the Probability Distribution Functions (PDF) of the uncertain model inputs have to be derived, from which numerous model inputs are simulated. By running the PAF using these sampled model inputs, a range of possible model outcomes or simulated realities is created. The simulation results of the final processing step provide valuable information for deriving quality layers. We applied an UPA of the boresight angles and a DEM to the VITO-PAF. Given the user requirement of pixel to sub-pixel accuracy with respect to the geo-location, results show that UPA is a powerful technique for the production of quality layers informing the user about the spatial-dependent total uncertainty and the contribution of uncertain model parameter in this total uncertainty.