HR: 14:00h
AN: B22C-02 INVITED     [PDF]
TI: Accuracy of the MODIS Snow-Cover Maps in Forest and Mixed Agriculture and Forest Land-Cover types
AU: * Hall, D K
EM: dorothy.k.hall@nasa.gov
AF: Hydrological Sciences Branch, Code 974, NASA/GSFC, Greenbelt, MD 20771 United States
AU: Riggs, G A
EM: griggs@glacier.gsfc.nasa.gov
AF: Science Systems and Applications, Inc., 10210 Greenbelt Rd., Suite 600, Lanham, MD 20706 United States
AU: Salomonson, V V
EM: vincent.v.salomonson@nasa.gov
AF: Earth Sciences Directorate, Code 900, NASA/GSFC, Greenbelt, MD 20771 United States
AU: Casey, K A
EM: kcasey@icecap.gsfc.nasa.gov
AF: Science Systems and Applications, Inc., 10210 Greenbelt Rd., Suite 600, Lanham, MD 20706 United States
AB: Snow-cover maps have been produced globally since March of 2000 from the Moderate-Resolution Imaging Spectroradiometer (MODIS) on NASA's Terra satellite. Other snow products such as fractional-snow cover and snow albedo have also been developed and will be available soon. The 500-m resolution snow maps have been validated in different land covers. It has long been known that it is difficult to map snow in dense forests, and thus the snow-mapping accuracy is lowest in forested land covers because trees, particularly dense forests, block the view of the ground where the snow accumulates. Accuracy of mapping snow cover at the global scale varies by month, mainly because the percentage of snow-covered forest varies as snow accumulates and melts during the year. Errors are greatest between the months of November and April in North America (9%) and Eurasia (9-10%) when the snow completely covers the boreal forest, and the least in the months of July through September. Using a 1-km resolution MODIS 17-class land-cover map product, developed at Boston University, we classify the Northern Hemisphere into 7 land-cover types (plus water): forest, mixed agriculture and forest, barren/sparsely vegetated, tundra, grasslands/shrublands, wetlands, and permanent snow and ice. Under a complete snow cover, snow-mapping errors are generally $<$5% in all of the classes except forest and mixed agriculture and forest, where errors can be as high as 10% in the spring and fall transition seasons. In this paper we study in detail the accuracy of the 500-m resolution MODIS Version 4 snow maps in these regions using Landsat ETM+ data used as "ground truth." For example, in the mixed agriculture and forest land cover type in south-central Nebraska and northern Kansas, comparison of a Landsat-derived snow map (used herein as "ground truth") with a MODIS snow map on 14 December 2000 showed that 96% of the pixels were correctly identified as snow (after cloud masking). We also use the MODIS vegetation continuous fields (VCF) product to determine percent tree cover for use in further validating the snow maps in the forest land-cover type. Limitations of this method of snow mapping include the presence of cloud cover.
UR: http://modis-snow-ice.gsfc.nasa.gov
DE: 1827 Glaciology (1863)
SC: Biogeosciences [B]
MN: 2003 Fall Meeting