OOS 81-9
Combined hyperspectral VSWIR and broadband thermal infrared analysis of vegetation-substrate mixtures in a mixed natural and anthropogenic landscape

Thursday, August 13, 2015: 4:20 PM
341, Baltimore Convention Center
Dar A. Roberts, Department of Geography, University of California at Santa Barbara, Santa Barbara, CA
Philip E. Dennison, Department of Geography, University of Utah, Salt Lake City, UT
Keely L. Roth, Land, Air & Water Resources, University of California Davis, Davis, CA
Glynn Hulley, Jet Propulsion Laboratory, Pasadena, CA
Kenneth Dudley, Geography, University of Utah
Erin Wetherley, Geography, University of California Santa Barbara, Santa Barbara, CA

Considerable ecological potential exists in the analysis of combined Visible-Short-Wave Infrared (VSWIR) imaging spectrometry data and broadband thermal infrared data, similar to that which would be acquired using the NASA’s proposed Hyperspectral Infrared Imager (HyspIRI). VSWIR imaging spectrometry has the capability of discriminating plant species and providing improved estimates of fractional surface cover in complex landscapes. Strong inverse relationships exist between green live cover and land surface temperature (LST), with this relationship varying between dominant plant species depending upon adaptations to water stress and other environmental controls on plant-temperature relationships. In this paper, we analyzed combined Airborne Visible-Infrared Imaging Spectrometer (AVIRIS) and MODIS/ASTER Airborne Simulator (MASTER) data acquired over a mixed natural-urbanized landscape in Santa Barbara, California in July 2011. We accounted for within class spectral variability using Multiple Endmember Spectral Mixture Analysis (MESMA).  We analyzed the relationship between LST from MASTER and fractional cover as it varies with species/plant functional type and substrate type (e.g. asphalt road, bare soil). We predicted subpixel LST using VSWIR fractions and compared these predictions to MASTER LST for plant-substrate mixtures.  We evaluated the difference between AVIRIS-predicted and MASTER-measured LST in mixed pixels as it varies by plant functional type and substrate.   


Significant species-level clustering was observed in a scatter plot between green vegetation fraction (GVF) and LST in natural and anthropogenic vegetation. Closed canopy, tall forest tended to have high GVF fractions and lowest LST. However, differences were observed, between closed canopy avocado  and Eucalyptus and live oak, with elevated temperatures observed in avocado, potentially resulting from higher temperatures in the underlying substrate or differences in slope and aspect for individual stands. Open stands, including citrus and blue oak had lower GVF and higher LST compared to closed stands. No significant differences were observed in the GVF-LST relationship between blue oak and citrus.  Plant species could not be discriminated in urban areas, but tall trees and short grass could be mapped at high accuracy and relationships between green cover and LST could be assessed for these two broad urban functional types. Preliminary results suggest that differences between predicted and measured temperatures offer a viable means for assessing sub-pixel elevated canopy temperatures and quantifying plant stress in mixed landscapes.