WHAT’S LURKING IN THE SHADOWS

4D-EARTH

Potential supervisors

dr. Chris Hecker, dr. Harald van der Werff

Spatial Engineering

This topic is not adaptable to Spatial Engineering

Suggested Electives

Thermal Infrared Remote Sensing (2nd year)

Additional Remarks

Description

Laboratory hyperspectral image measurements should not have shadows , but they do! When using linear unmixing methods on hyperspectral datasets, a “shadow” endmember with a flat spectrum is included in the unmixing process to compensate for albedo differences caused by e.g. illumination differences due to topography (=”shadow”). Even in laboratory infrared imaging on cut samples, unmixing will result in a shadow endmember as well, even though there are barely any illumination differences and definitely no shadows on the images.
The distribution of the shadow endmember shows spatial patterns and is clearly not random. At the moment we do not understand what the shadow endmember represents. Suspicions are that it could be related to minerals that are not shortwave infrared active (e.g. quartz) or possibly differences in grain size or finely distributed dark minerals. We simply do not know at the moment, but knowing the answer has important implications on the accuracy of the mineral maps resulting from unmixing.

Objectives and Methodology

The objective of this research would be to find out what causes the spatial patterns in the “shadow” maps of laboratory hyperspectral images from the ITC Geoscience Laboratory. By investigating this topic you will aim to understand how the different possible causes are influencing the spectral unmixing and with it, the accuracy of the final mineral mapping result. Starting point of this work is the MSc research of Cecilia Isabel Contreras Acosta (2017). There are a number of rocks available from the epithermal gold system of Rodalquilar, Spain. The samples have been cut and imaged (VNIR and SWIR) by Contreras, and contain interesting patterns. These samples are the starting point but samples from other areas or LWIR measurements of the same samples could be included to assess and model the effect in different wavelength ranges. For validation purposes other laboratory techniques, like XRD, Microprobe, micro-XRF imaging are likely to be used to unravel the mystery of the shadow endmember!

Further reading