Submitted to: Journal of Applied Remote Sensing (JARS)
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 15, 2007
Publication Date: January 19, 2007
Citation: Hunt, E.R., Jr., Daughtry, C.S.T., Kim, M.S., Parker Williams, A.E. 2007 Combining canopy reflectance models and spectral angle mapper to detect invasive weeds by remote sensing. Journal of Applied Remote Sensing. 1:1-17. Interpretive Summary: Research to develop methods for the remote sensing invasive weeds can be summarized as trial and error. One of the problems is there are large amounts of variation in the spectral reflectance from vegetation canopies because of differences in the growth (measured by leaf area index or LAI), ground cover, soil background reflectance and other variables. Canopy reflectance models, such as the Scattering by Arbitrarily Inclined Leaves (SAIL) model, can predict canopy reflectances. We developed a Microsoft Windows graphical user interface to the SAIL model to facilitate its use. The next problem is to quantify the similarity between two reflectance spectra, the Spectral Angle Mapper uses vector algebra to define a spectral angle, where the larger the spectral angle, the two spectra are more different. So by combining the spectral angle mapper with SAIL model output, we can predict the conditions for detection of a particular species by remote sensing. This method was tested for the detection of the invasive weed, leafy spurge, at Devils Tower National Monument using images from Airborne Visible InfraRed Imaging Spectrometer (AVIRIS), Landsat 7 Enhanced Thematic Mapper Plus (Landsat 7 ETM+), and System Pour d’Observation de la Terre 4 (SPOT 4). Detection accuracy was highest using the visible and near-infrared bands of AVIRIS and lowest using all four bands of SPOT 4, which follows the predictions. Therefore, if a collection of reflectance and transmittance spectra for various invasive weeds were developed, called a spectral library, we would reduce the amount of trial and error and speed up the development of remote sensing methods for invasive weeds.
Technical Abstract: One of the goals of remote sensing is to use a spectral library to detect the presence of invasive weeds and other species in remotely-sensed imagery; however, variation in leaf area index (LAI), cover, and soil background make detection of single species amongst other vegetation difficult. Variation in canopy reflectance may be simulated using the Scattering by Arbitrarily Inclined Leaves (SAIL) model. The resulting simulations are then used to calculate spectral angles (used in the Spectral Angle Mapper classification technique), and the spectral angles indicate the conditions under which invasive species may be detected by remote sensing. We tested this idea with leafy spurge (Euphorbia esula L.), which is a distinctive invasive weed with yellow-green flower bracts. Model results show that leafy spurge may be detected with hyperspectral imagery when LAI is greater than 1.0 and flower cover is greater than 20%. Detection of leafy spurge with multispectral imagery with broad visible and near-infrared bands was predicted to be marginal. Airborne Visible InfraRed Imaging Spectrometer (AVIRIS), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), and System Pour d’Observation de la Terre (SPOT) 4 data were acquired around Devils Tower National Monument in northeastern Wyoming as part of The Ecological Area-wide Management of Leafy Spurge project. Classification accuracy for the AVIRIS imagery with 49 contiguous bands in the visible and near infrared was 82%, whereas accuracies for Landsat 7 ETM+ and SPOT 4 data were better than chance, in agreement with the predictions. Use of the Spectral Angle Mapper in classifying imagery shows that differences in spectral angles of about 3' (0.05 radians) can be used to determine whether an invasive species is detectable.