Using MODIS and VIIRS for Global Snow Fraction Retrieval

The original fractional snow cover algorithms developed by the author for MODIS (jointly with Dr. Salomonson) and VIIRS are based on different approaches.
The MODIS algorithm employs the Normalized Difference Snow Index (NDSI) - a spectral band ratio that takes advantage of the fact that snow reflectance is high in the visible wavelengths (0.4-0.7 µm) and low in the short-wave infrared region (1-4 ìm). It has been demonstrated that the NDSI has enough of a “signal” to obtain subpixel estimates of snow cover using the statistical linear relationships between the NDSI from MODIS observations and the fraction of snow cover, though the approach to establishing optimal relationship was not trivial.
The VIIRS algorithm uses reflectances in nine VIIRS moderate resolution bands to retrieve snow fraction. The algorithm is a modification of Multiple Endmember Spectral Mixture Analysis (MESMA). Spectral mixture analysis “unmixes” the mixed reflectances, determining the fractions of snow and non-snow spectral endmembers that produce the mixed pixels spectral signature. The developed approach is based on the assumption that the non-snow endmember spectrum for each pixel is operationally provided by the VIIRS albedo algorithm. Snow endmembers are modeled on the basis of 6S and DISORD radiotransfer simulations taking into account specific conditions characteristic to each pixel under consideration.
Fractional snow retrieval was used to analyze changes in Earth's surface state for different scales from day-to-day changes to interannual variability, to consider snow processes and land surface hydrology.
The performance of both fractional snow cover algorithms is tested against ground truth obtained from Landat-7 ETM+ scenes covering a substantial range of snow cover conditions. The validation of the algorithms is quantified by such measures as mean absolute error, mean-square-root error, and coefficient of correlation.
Proposed MODIS algorithm is currently routinely employed in the processing of MODIS data (collection 5) to obtain daily, global and regional snow fraction estimates.
Fractional snow cover is planned to be a standard VIIRS level 2 product provided globally with uncertainty of 0.1 for the measurement range of snow cover from 0 to 1.
The approach developed for VIIRS retrieval is based on the hypothesis that the best quality of retrieval is obtained when the assumption about variation of endmembers and algorithm parameters from pixel to pixel is taken into consideration.
The results from developed MODIS approach could be also improved if the „universal“ relationship "tuned" for a specific area. Improved fractional snow cover retrieval can be achieved by an algorithm that takes into account the variability of snow and non-snow spectral signatures.
It has been demonstrated that the use of scene-specific snow and non-snow endmembers improves estimates of snow fraction retrievals. The results of the study indicate a way to improve performance of algorithms identifying the endmembers from observations.
It could be concluded that the development of scene-based algorithms suppressing sensitivity to illumination and non-snow cover conditions is a very powerful way to significantly improve the accuracy of fractional snow cover derivation.