Space Grant Research at Augustana
The funds which support Space Day '97 come from a NASA Space Grant. This same grant also funds research fellows at Augustana College. Individual projects include the following:
  • Dr. Larry Tieszen
  • Bruce Wylie "Spectral Unmixing"
  • Dr. Daniel Swets
  • Dr. Gil Blankenspoor "Breeding Habitat Suitability in the Common Merganser (Mergus merganser) as Determined by Remote Sensing"
  • Joel Vander Kooi "Sampling Biophysical Parameters Using Radiometric Instrumentation at The Nature Conservancy's Tallgrass Prairie Preserve, Oklahoma: Summer, 1996"
  • Michael Lehmkuhl "Spectral Unmixing as a Means of Estimating Subpixel Biomass"
  • Jon Rogness

    • Back to the Agenda.
    Back to the Main page.
     

    Spectral Unmixing

    Background

    Spectral observations often contain parts of several spectral components. A spectral signature obtained from a grassland with a hand held radiometer with a field of view of 0.5 m2 could easily contain shadow, bare ground, and foliage spectral signatures mixed together. This same concept applies to satellite-based observations with pixel sizes ranging from 1 km to 1 m.

    Spectral unmixing estimates the combinations of spectrally pure signatures (called end members) which would produce a spectral signature similar that of the "mixed" pixel. It then estimates the percentage of that "mixed" pixel which would be made up of the various spectral end members.

    Applications

    Spectral unmixing has many applications. Spectral mixing components have been used to estimate biophysical parameters (Hall et al. 1995, van Leeuwen and Huete 1996, Su et al. 1995) and classification (Ustin et al. 1996). It also is a useful tool for scaling applications (Puyou-Lascassies et al. 1994)

    Study

    We have demonstrated that biophysical parameters for grasslands can be predicte from remotely sensed vegetation indices. These relationships appear to be fairly robust across varying levels of standing dead vegetation (Wylie et al. 1996). We wondered if spectral unmixing components would provide more accurate estimates of biophysical parameters than vegetation indices. Would the spectral unmixing estimates be more robust across varying levels of standin dead vegetation and sites?

    We used a nine band hand held radiometer (Cropscan) to collect spectral reflectance data for a series of 0.5 m2 quadrats. These data were collected from two sites, the Nature Conservancy bison preserves in the Nebraska Sand hills (Niobrara Valley Preserve) and southern Flint Hills tall grass ( Tall Grass Preserve). After reflectance data was collected, the quadrat were sampled for leaf area index (LAI), fPAR, and biomass. LAI estimates where obtained with a LAI 2000 Plant Canopy Analyzer and fPAR obtained with CEPTOMETERS. Biomass was clipped at ground level and sorted into live, standing dead, and litter components.

    Spectral end members were obtained using principal component techniques (Bateson and Curtiss 1996) and spectral angle mapper unmixing techniques used.

    Results at the Niobrara site indicated equivalent abilities to estimate grassland biomass with spectral unmixing components and NDVI. However, when the Tall Grass and Niobrara data sets were combined end member selection became more difficult.

    References

  • A. Bateson and B. Curtiss. 1996. A method for manual endmember selection and spectral unmixing. Remote Sens. Environ. 55:229-243.
  • Hall, F.G, Y.E. Shimabukuro, and K.F.Huemmrich. 1995. Remote sensing of forest biophysical structure using mixture decomposition and geometric reflectance models. 1995. Ecological Applications 5:993-1013.
  • Van Leeuwen, W.J.D. and A.R. Huete. 1996. Effects of standing litter on the biophysical interpretation of plant canopies with spectral indices. Remote Sens. Environ. 55:123-138.
  • Puyou-Lascassies, P., G. Flouzat, M. Gay, and C. Vignolles. 1994. Validation of the use of multiple linear regression as a tool for unmixing coarse spatial resolution images. Remote Sens. Environ. 49:155-166.
  • Ustin, S.L., Q.J. Hart, L.Duran, and G. Sheer. 1996. Vegetation mapping on hardwood rangelands in California. Int. J. Remote Sensing 17:3015-3036.
  • Wylie, B.K., D. DeJong, L.L. Tieszen, and M.E. Biondini. 1996. Grassland canopy parameters and their relationships to remotely sensed vegetation indices in the Nebraska sand hills. Geocarto Int. 11: 39-52.

    Back to the top.
    Back to the Agenda.
    Back to the Main page.
     

    Breeding Habitat Suitability in the Common Merganser (Mergus merganser) as Determined by Remote Sensing

    In recent years, the Common Merganser has been identified as a principal host for adults forms of the parasitic blood flukes that cause human cercarial dermatitis (swimmer's itch)(Loken et al., 1995; Blankespoor and Reimink, 1991). The usual approach to controlling swimmer's itch has been to kill snail hosts of the intermediate stage of the parasite with massive applications of copper sulfate. It is now generally recognized that this approach to control is unacceptable because of the considerable damage copper sulfate does to aquatic ecosystems. In several states, such use of copper sulfate is now illegal. Off and on, in recent years, I have been part of a research effort to find alternative control strategies. These control strategies have focused on the Common Merganser (Blankespoor and Reimink, 1991) and also, to some extent, on another, less important, host of the adult forms of the parasite, the Mallard (Anas platyrhynchos)(Reimink et al., 1995; Blankespoor et al., 1994).

    One goal of this effort, related to Common Mergansers, is to learn what makes a lake suitable for breeding. We know that lakes are not equally suitable because some lakes in a region support breeding Common Mergansers while others do not. Distribution of breeding Common Mergansers is possibly influenced by one or more of the following: availability of fish prey, availability of nest sites, distribution of snail hosts, size of lake, clarity of water, water depth, water temperature, type of lake substrate, amount of shoal, and degree of lakeshore development.

    The basic objective of my project will be to use and interpret satellite imagery to identify differences in two categories of lakes -- those that do and do not support breeding Common Mergansers. Once a suite of characteristics that make a lake either suitable or not suitable for Common Merganser breeding has been identified, it ought to be possible to use imagery to predict breeding suitability for lakes all across the breeding range. A final step would be to test these predictions by actually surveying lakes.

    It is anticipated that the results of this study will enhance our ability to design effective control strategies for the swimmer's itch parasite. In addition, they will make a contribution to what is known about the natural history of the Common Merganser.

    References

  • Blankespoor, H. D., and R. L. Reimink. 1991. The control of swimmer's itch in Michigan: past, present, future. Michigan Academician 24:7-23.
  • Blankespoor, G. W., B. Loken, H. D. Blankespoor, and R. L. Reimink. 1994. Breeding populations of Mallards, Anas platyrhynchos, on four urbanized lakes in Michigan. Canadian Field-Naturalist 108:292-296.
  • Loken, B. R., C. Spencer, and W. O. Granath, Jr. 1995. Prevalence and transmission of cercariae causing schistosome dermatitis in Flathead Lake, Montana. J. Parasitol. 81:646-649.
  • Reimink, R. L., J. A. Degoede, and H. D. Blankespoor. 1995. Efficacy of praziquantel in natural populations of mallards infected with avian schistosomes. J. Parasitol. 81:1027-1029.

    Back to the top.
    Back to the Agenda.
    Back to the Main page.
     

    Sampling Biophysical Parameters Using Radiometric Instrumentation at The Nature Conservancy's Tallgrass Prairie Preserve, Oklahoma: Summer, 1996

    Quantification of biomass, leaf area index, fraction of Photosynthetically Absorbed Radiation (fPAR), and chlorophyll from both detailed study sites and regional estimates are essential for understanding global change and land surface interactions with the atmosphere. Monitoring of these biophysical parameters allows calibration and validation of global change, atmospheric, and biogeochemical models. Regional estimation of these parameters can best be achieved through the use of remote sensing. Previous work conducted at The Nature Conservancy's Niobrara Valley Preserve in Nebraska and the Tallgrass Prairie Preserve in Oklahoma combined with similar data from Grasslands National Park in Canada allowed the establishment of predictive relationships that estimate biophysical parameters from hand held radiometer reflectances (Cropscan). These biophysical parameters are then estimated for a systematic grid of Cropscan reading representing a 30 m by 30 m area (readings taken every six m) and correlated to SPOT or TM satellite data allowing estimation of these parameters over a larger area.

    The research effort at the Tallgrass Prairie Preserve in the summer of 1996 attempted to monitor Cropscan reflectances of twenty-one 30 m by 30 m systematic grids continuously (at least once per two weeks) through the first half of the summer, allowing comparison to any SPOT or TM satellite data collected during this time period that was cloud free. This was accomplished with data collected on the ground early in the spring (3/31/96-4/2/96) and continuously from 5/20/96 through 7/28/96. Relatively cloud free satellite imagery was obtained for the Tallgrass Prairie Preserve. An on site shadow band radiometer and sun photometer will allow atmospheric correction of the satellite data sets.

    Biophysical parameter data and Cropscan reflectance data were collected to assess the robustness of biophysical parameters - vegetation index relationships derived in 1995 and 1994. At monthly intervals, 18 quadrats (0.5 m2) were sampled for biophysical parameters (including fPAR and LAI at various height increments through the canopy and the height of canopy closure). Six quadrats were sampled in each of three treatment areas: burned, ungrazed, and traditional (burned and cattle-grazed). These data will be use to assess the robustness of current relationships across season and across treatments. In addition, parameters collected such as canopy height will be investigated as parameters that may improve predictability of biophysical parameters and possibly reduce saturation effects observed in some of the current relationships. Biomass from two quadrats per treatment was clipped and sorted into green and litter components. Cropscan and fPAR reflectance data were also collected after the quadrats were clipped. These data will be used to develop canopy reflectance models that will provide another estimation of biophysical parameters from reflectance data.

    Reflectance data was also collected from soil, shadow, and sunlit canopy. These will be used as spectral end members to assess the utility of spectral unmixing to estimate biophysical parameters.

    To estimate biophysical parameters associated with trees and shrubs, 8 grids of 30 m by 30 m were selected representing varying levels of woody vegetation. This would allow the development of relationships between satellite reflectance data and shrub biophysical parameters and subsequent extrapolation to larger areas. The tree grids were sampled for fPAR and LAI at 6 m intervals for two heights: below the herbaceous layer and above the herbaceous but below the shrub layer. These grids were sampled early in the spring (4/1) before leaf emergence and in July after full leaf extension to allow estimation of foliar LAI and fPAR. In addition, these grids were sampled for tree density by species, height, and breast height diameter. Existing algorithms will allow estimation of foliar biomass at each grid.

    Back to the top.
    Back to the Agenda.
    Back to the Main page.
     

    Spectral Unmixing as a Means of Estimating Subpixel Biomass

    Spectral mixing occurs as a result of spatially mixed materials (e.g. bright and dark soils, vegetation, shadows). The materials whose spectral signatures are linearly independent and can explain the spectral signature of the pixel as a whole are called endmembers. Spectral unmixing is a process whereby the relative percentages of per pixel endmember abundances can be determined. Different combinations of candidate endmembers were chosen to unmix based on the plots of the endmember signatures in the projection from a principle components analysis (PCA). The endmembers were those materials whose spectral signatures represented extrema, enclosing other data points. The combinations of these endmembers were used by the Spectral Image Processing System (SIPS) software (CSES, CIRES, University of Colorado, Boulder) which performed the actual unmixing algorithm. The unmix function produced per pixel endmember abundance values. The abundance values were extracted from the SIPS output and input to a stepwise linear regression to determine the accuracy of the prediction of biophysical parameters such as leaf area index (LAI), fraction of photosynthetically absorbed radiation (fPAR), and total green biomass. The regressions will also be run against chlorophyll and other biophysical parameters.

    The data set was recorded with a nine band Cropscan ground radiometer in the Niobrara Valley Preserve in Nebraska and the Tall Grass Preserve in Oklahoma. The bands used were centered at 460, 507, 559, 613, 661, 706, 760, 813, and 1650 microns. For each ground plot, the biophysical parameters mentioned above were recorded.

    Spectrally derived vegetation indices (e.g. NDVI) are often used as predictors of biophysical parameters, but are sometimes scale variant (i.e. non linear). We will investigate the feasibility of using spectral unmixing components to estimate and scale biophysical parameters.

    Back to the top.
    Back to the Agenda.
    Back to the Main page.