This is the Split Window Surface Temperature algorithm description derived
from the prototype implementation provided by the authors and subsequent
conversations and emails.

==============================================================================
Algorithm Authors:

    Aaron Gerace
    Rochester Institute of Technology
    email: adgpci@cis.rit.edu

    Tania Kleynhans
    Rochester Institute of Technology
    email: tkpci@rit.edu

==============================================================================
Algorithm Description - Overview:

        1) Create 2 Landsat emissivity bands based on Landsat top of
           atmosphere reflectance, ASTER emissivity and ASTER NDVI data.
 
        2) Read TIRS coefficients.

        3) Read 2 Landsat TIRS thermal bands.  Convert them to thermal radiance,
           and convert that to apparent temperature.

        4) Use the Landsat emissivity bands, apparent temperature bands, and
           coefficients to make a surface temperature band.

==============================================================================
Algorithm Description - Inputs:

    The primary sources of input to the algorithm are:

        Landsat L1T thermal bands retrieved from LPGS L1T products
        ASTER Emissivity and NDVI data
        Landsat TOA (top of atmosphere) reflectance bands

    The L1T thermal bands are used to create apparent temperature bands.

    The ASTER emissivity and NDVI data, as well as the Landsat TOA data
        from multiple bands, are used to simulate Landsat emissivity bands
        matching the Landsat scene.

==============================================================================
Algorithm Description - Detailed:

        1) Read Landsat top of atmosphere reflectance green, red, NIR, SWIR1,
           and brightness temperature bands for the scene.  Generate a 
           Landsat NDVI layer from the red and NIR bands.  Generate NDSI and
           snow layers from the green and SWIR bands.  Get projection
           information from the brightness temperature band.

        2) Retrieve ASTER emissivity and NDVI data, and warp them to match
           the Landsat scene.

        3) Calculate a simulated Landsat emissivity layer using the ASTER
           emissivity values, ASTER NDVI values, Landsat NDVI band, NDSI band,
           and snow locations.  Write the simulated Landsat emissivity layer
           to a file.  Repeat the the process using coefficients for 2 Landsat
           TIRS bands to produce 2 emissivity files.

        4) Read predefined TIRS coefficients.

        5) Read metadata for 2 Landsat thermal bands.

        6) Use the 2 Landsat L1T thermal bands and metadata coefficients to
           derive thermal radiance layers, and write band files for those
           2 layers.

        7) Use the 2 thermal radiance layers and metadata coefficients to
           compute 2 apparent temperature layers.

        8) Calculate Surface Temperature:

           epsilon = (b10_emissivity + b11_emissivity) / 2
           delta_epsilon = b10_emissivity - b11_emissivity
           b = TIRS coefficients
           b10 = band 10 apparent temperature
           b11 = band 11 apparent temperature

           surface_temperature = b[0] + (b[1] + b[2] * ((1 - epsilon) / epsilon)
                         + b[3] * (delta_epsilon / (epsilon^2)))
                         * ((b10 + b11) / 2.0) 
                         + (b[4] + b[5] * ((1 - epsilon) / epsilon)
                         + b[6] * (delta_epsilon / (epsilon^2)))
                         * ((b10 - b11) / 2.0) + b[7] * ((b10 - b11)^2)

        9) Scale the surface temperature value to include 1/10 degree precision
           in the integer product.  If intermediate bands are included in the
           final product, also scale the 2 emissivity bands and 2 thermal
           radiance bands and convert them to integer.