Improving the Calibration of Airborne Hyperspectral Sensors for Earth Observation

Abstract

Hyperspectral sensors are a widely used tool in remote sensing of the Earth's surface. Due to the versatility of the sensors, a multitude of applications profit from or use hyperspectral data, most prominently the assessment of the impact of climate change on the environment. The acceptance of hyperspectral remote sensing stems, in part, from the relatively easy access to data. Besides several existing and planned spaceborne instruments, airborne sensors are by now a commercial, if expensive, commodity. The German Aerospace Center (Deutsches Zentrum f\"ur Luft- und Raumfahrt, DLR) operates an optical laboratory for the characterization and calibration of airborne hyperspectral sensors, the calibration home base (CHB). This laboratory is needed to close the gap between the extensive characterization efforts performed for spaceborne devices, and the characterization performed by the commercial sensor's manufacturers. Currently, the latter often does not incorporate the possibility of the traceability of calibration, and may not take into consideration sensor properties that require more complex or time-consuming characterisation methods. As the airborne sensors are effectively used to develop and validate methods for spaceborne hyperspectral remote sensing, the same level of detail in characterization needs to be applied to the airborne devices, resulting in similar measurement uncertainties. One objective of this thesis is therefore to assess the impact of the characterization and calibration possible with the CHB on data from airborne hyperspectral instruments. This is shown on a pair of NEO HySpex hyperspectral instruments operated by DLR for the bathymetry of lake Starnberg as an exemplary hyperspectral remote sensing data product. The requirements for the calibration of future spaceborne hyperspectral instruments include the traceability to the système international d'unités (SI). The second part of the thesis therefore deals with different aspects of propagation of measurement uncertainty and traceability. The sensors discussed here are APEX (Airborne Prism EXperiment) and ROSIS (Reflective Optics System Imaging Spectrometer). Two software tools are discussed: First, a database for the storage of the entire APEX laboratory characterization data, with frontends allowing for the rapid analysis of characterization data as well as the generation of the data needed for the calibration. Within the database, laboratory metadata is always linked to the actual sensor data, which simplifies the tracing of calibration back to the laboratory data. The second software tool is a Monte Carlo simulation of the airborne measurement process of an inland water body. This simulation is based on a sensor model of ROSIS, but other sensor models, such as one of the HySpex instruments, can be implemented. This allows to propagate the uncertainties from the laboratory measurements to airborne data acquisition. Additionally, this simulation can be used to propagate the laboratory calibration uncertainties to final hyperspectral remote sensing products, which can support the development of hyperspectral methods.