Date of Award

Spring 1-1-2018

Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Peter Pilewskie

Second Advisor

Konrad Sebastian. Schmidt

Third Advisor

Thomas Kampe

Fourth Advisor

Owen Brian. Toon

Fifth Advisor

Xinzhao Chu


Hyperspectral instruments expand the spectral dimension of remote sensing measurements by collecting data in hundreds of contiguous wavelength channels. Spectrally resolved measurements can be used to derive science products for a diverse range of fields such as atmospheric science, geology, oceanography, ecology, climate monitoring, and agricultural science, to name a few. The spectral information collected by hyperspectral instruments enables more accurate retrievals of physical properties and detection of temporal changes. These advantages have led to an increasing number of active and planned hyperspectral instruments. This thesis describes methods for attributing hyperspectral radiation observations to physical sources.

We developed, validated and characterized improvements to a hyperspectral instrument, the Solar Spectral Irradiance Monitor (SSIM), built at the University of Colorado Boulder’s Laboratory for Atmospheric and Space Physics. Contributions include the characterization of the optics’ angular response, testing of an optics stabilizing platform and the development and testing of a spectrometer thermal control system. This instrument was then deployed on an aircraft for a field study with the National Ecological Observatory Network (NEON). SSIM measurements of upwelling and downwelling irradiance were used in conjunction with NEON’s Imaging Spectrometer to enable atmospheric correction of imagery collected below cloud layers.

We developed a numerical spectral unmixing algorithm, Informed Non-Negative Matrix Factorization (INMF), to separate contributions to hyperspectral imagery from distinct physical sources such as surface reflectance, atmospheric absorption, molecular scattering, and aerosol scattering. INMF was tailored for hyperspectral applications by introducing algorithmic constraints based on the physics of radiative transfer. INMF was tested using imagery collected by the Hyperspectral Imager for the Coastal Ocean (HICO). To validate the method INMF results were compared to model-based atmospheric correction results. We demonstrate possible applications of INMF by presenting the retrieval of two physical properties, aerosol attributed radiance and seafloor depth. The retrievals were evaluated by comparing INMF output to independent retrievals of aerosol properties from the Moderate Resolution Imaging Spectroradiometer (MODIS) and in-situ seafloor depth measurements from the U.S. Coastal Relief Model. In these comparisons INMF shows promise for retrieving both physical properties, and may be improved with physics-based constraints on the seafloor and aerosol source spectra.