1. Surface Characterization 4th Annual Workshop on Hyperspectral Meteorological Science of UW MURI And Beyond Donovan Steutel Paul G. Lucey University of Hawai‘i

2. Surface Emissivity Simulation Provide emissivity boundary condition for radiative transfer modeling of atmosphere Use multispectral infrared satellite data as base map Fit infrared multispectral data with hyperspectral library data Produce continuous sampled spectrum at each pixel

3. MODIS features several bands with weak atmospheric extinction appropriate for surface characterization

4. MODIS “surface” bands constrain surface compositional types

5. MODIS team generates emissivity models for cloud-purged mosaics False color MODIS image

6. Emissivity assignment Extract soil emissivities at MODIS wavelengths (assuming Kirchoff’s Law) from ASTER spectrum library (41 soils) Treat ASTER soils as lookup table Compare each MODIS spectrum to lookup table and return soil sample# for closest fit (by RMS differences) Insert full resolution ASTER spectrum at each location

8. 0%1.5%3% Difference of library match Mean-normalized comparison

9. 0%0.75%1.5% Emissivity difference after gain/offset Absolute emissivity comparison

10. MODIS Emissivity Type Map Spatial resolution: 0.05° (~5km at equator) Wavelength range: 685-2250 cm -1 Wavelength sampling: 2 cm -1 Spectral channels: 801 White dune sand stony coarse sand Coniferous vegetation Deciduous vegetation Grass Other No data

11. Contrast ratios between library and MODIS spectra with similar spectral shapes can achieve large values routinely. Is this physical?

12. DARPA research conducted by the University of Hawai‘i [Johnson et al., 1998] discovered a consistent difference in spectral contrast associated with soil disturbance: Sampled sites have lower spectral contrast than surfaces measured in the field.

13. Background soil Disturbed location Solid is ground truth data Symbols are airborne hyperspectral data1

14. LWIR Observable Quartz. Particles >75 microns show high reflectance at 9 microns. Sample containing large amounts of fine particles show low reflectance If particle sizes are less than the attenuation scale, partial coupling is enabled because of incomplete damping. This leads to higher emissivity and lower reflectance of small particles. (Gold blacks exploit this phenomenon in the visible).

15. “Dirty” Dirt (sampled) “Clean” Dirt (unsampled)

16. Particles in the size range of 1-10 microns are very abundant in natural soils leading, and their abundance is coupled to natural geologic processes. This leads to relatively large variations in spectral contrast Sampled soils show low spectral contrast relative to the surroundings, as well as lower variation in contrast. This is consistent with an enhanced abundance of small particles. Conclusion: High contrast ratios are physical, but need further investigation in individual cases Library vs. MODIS emissivity contrast

17. Model inputs may be dynamic to account for temporal variations in emissivity Seasonal emissivity products MOD11C3 Dec. 2002 monthly average, 0.05° scale Dec. 2002

18. MOD11C3 Aug. 2003 monthly average, 0.05° scale Seasonal emissivity products Model inputs may be dynamic to account for temporal variations in emissivity Aug. 2003

19. Seasonal emissivity products MOD11C3 Oct. 2003 monthly average, 0.05° scale Model inputs may be dynamic to account for temporal variations in emissivity Oct. 2003

20. Assign hyperspectral signatures to each geographically defined pixel for each season From seasonal data produce a dynamic emissivity simulation to include vegetation senescence effects –MOD11C series available daily (C1), 8-day (C2), and monthly (C3) Delivery of this product: April 2005 Planned work