HR: 1340h
AN: A23A-0930    [Abstracts]
TI: Assessment of CCRS Cloud Detection Algorithm CLEAR Using AVHRR and MODIS Observations
AU: Khlopenkov, K
EM: kkhlopen@ccrs.nrcan.gc.ca
AF: Canada Centre for Remote Sensing, Natural Resources Canada, Rm 415, 588 Booth Str, Ottawa, ONT K1A 0Y7 Canada
AU: * Trishchenko, A
EM: trichtch@ccrs.nrcan.gc.ca
AF: Canada Centre for Remote Sensing, Natural Resources Canada, Rm 415, 588 Booth Str, Ottawa, ONT K1A 0Y7 Canada
AB: The cloud/clear-sky detection algorithm CLEAR (CLoud Estimation using Aggregated Rating index) has been developed at the Canada Centre for Remote Sensing (CCRS) for the processing of the historical Advanced Very High Resolution Radiometer (AVHRR) observations made from the National Oceanic Atmospheric Administration satellites NOAA-6 to NOAA-17. The algorithm employs observations from 5 AVHRR channels and temperature fields from the North America Regional Reanalysis (NARR). The unique feature of the scene identification algorithm is the incorporation of several tests to produce an aggregated effective cloudiness index. This aggregated index has better reliability than a simple sequence of separate tests and allows for a distinction between the different levels of cloudiness. Other features of the algorithm are the dynamic correction for the sun glint for water pixels, the generation of snow/ice maps for cloud-free and thin-cloud pixels, and the calculation of a cloud shadow mask. The algorithm is implemented to operate for daytime and nighttime scenes during snow-free and snow seasons, over water and land areas. An assessment of the algorithm performance was conducted using supervised cloud/snow classification employing the Maximum Likelihood Classifier (MLC) routine available as a part of the multispectral analysis package of the Geomatica system (PCI Geomatics 2003). Comparison has been done for 12 scenes that covered Canada and US mid-latitude and polar regions and represented various seasons. Comparison showed very good consistency between automated processing using CLEAR scheme and the results of supervised classification. The agreement in the cloud detection was in the range of 89 percent to 91 percent for the summer scenes and around 84 percent to 88 percent for the winter. The consistency in the snow identification varied from 86 percent in winter to 94 percent during the warm season. The CLEAR algorithm has been also tuned to MODIS channels and applied for several scenes observed from MODIS. Results derived with CLEAR algorithm show reasonable consistency with MODIS cloud mask (at the level of 70% or higher). Detection of clouds from MODIS over water in cold season and performance in cloud/snow separation over bright surface targets and areas with fractional snow coverage will be discussed in more details. This work has been supported by the Canadian Space Agency under the Government Related Initiatives Program (GRIP) and the Earth Sciences Sector of the Department of Natural Resources Canada under the Program "Reducing Canada's Vulnerability to Climate Change".
DE: 0321 Cloud/radiation interaction
DE: 0480 Remote sensing
SC: Atmospheric Sciences [A]
MN: Fall Meeting 2005