HR: 1330h
AN: C12A-0865    [PDF]
TI: Using Kernel Methods to Detect Clouds, Snow, Ice and other Geophysical Processes.
AU: Srivastava, A
EM: ashok@email.arc.nasa.gov
AF: Ashok Srivastava, Research Institute for Advanced Computer Science NASA Ames Research Center M/S T35B-1, Bldg T35B, Moffett Field, CA 94035 United States
AU: * Stroeve, J C
EM: stroeve@kryos.colorado.edu
AF: Julienne Stroeve, University of Colorado National Snow and Ice Data Center Campus Box 449, Boulder, CO 80309 United States
AU: Oza, N
EM: oza@email.arc.nasa.gov
AF: Ashok Srivastava, Research Institute for Advanced Computer Science NASA Ames Research Center M/S T35B-1, Bldg T35B, Moffett Field, CA 94035 United States
AB: The detection of clouds within a satellite image is essential for accurately retrieving many surface geophysical parameters from optical and thermal imagery. Even a small percentage of cloud cover within a radiometer pixel can adversely affect the determination of surface variables such as albedo and temperature. Unfortunately, cloud detection over and snow- and ice-covered surfaces remains difficult due to the small spectral contrast in the visible and thermal infrared between clouds and snow/ice. This study examines the use of kernel methods to distinguish between snow and clouds over the Greenland ice sheet using MODIS imagery. Besides being able to classify clouds, the method is also able to pick up regions of different snow types (i.e. grain size variations) over the Greenland ice sheet. Thus, the results presented here lend credibility to the idea that a kernel- based method could reveal geophysical processes and discriminate bewteen such processes. Here we present the results for several images acquired during the melt season in 2002.
DE: 1863 Snow and ice (1827)
DE: 3360 Remote sensing
DE: 9820 Techniques applicable in three or more fields
SC: Cryosphere [C]
MN: 2003 Fall Meeting