| Title | SPARC: new cloud, snow, and cloud shadow detection scheme for historical 1-km AVHHR data over Canada |
| |
| Author | Khlopenkov, K V; Trishchenko, A P |
| Source | Journal of Atmospheric and Oceanic Technology vol. 24, issue 3, 2007 p. 322-343, https://doi.org/10.1175/JTECH1987.1  Open Access |
| Image |  |
| Year | 2007 |
| Alt Series | Earth Sciences Sector, Contribution Series 2005700 |
| Publisher | American Meteorological Society |
| Document | serial |
| Lang. | English |
| Media | paper; on-line; digital |
| File format | html; pdf |
| Subjects | Nature and Environment |
| Program | Reducing Canada's Vulnerability to Climate Change
|
| Released | 2007 03 01 |
| Abstract | The identification of clear-sky and cloudy pixels is a key step in the processing of satellite observations. This is equally important for surface and cloud-atmosphere applications. In this paper, we
present the SPARC (Separation of Pixels using Aggregated Rating over Canada) algorithm, a new method of pixel identification for image data from the Advanced Very High Resolution Radiometer (AVHRR) onboard the NOAA satellites. The SPARC algorithm
separates image pixels into clear-sky and cloudy categories based on a specially designed rating scheme. A mask depicting snow/ice and cloud shadows is also generated. The SPARC algorithm has been designed to work year round (day and night) over the
temperate and polar regions of North America, for current and historical AVHRR/NOAA High-Resolution Picture Transmission (HRPT) and Local Area Coverage (LAC) data with original 1-km spatial resolution. The algorithm was tested and applied to data
from the AVHRR sensors flown onboard NOAA-6 to NOAA-18. The method was employed in generating historical clear-sky composites for the 1982-2005 period at daily, 10-day and monthly time-scales at 1-km resolution for an area of 5700×4800 square km
centered over Canada. This region also covers the northern part of the United States including Alaska, as well as Greenland and the surrounding oceans.
The SPARC algorithm is designed to produce an aggregated rating that accumulates the results of
several tests. The magnitude of the rating serves as an indicator of the probability for a pixel to belong to clear-sky, partly cloudy or overcast categories. The individual tests employ the spectral properties of 5 AVHRR channels, as well as surface
the skin temperature maps from the North American Regional Reanalysis (NARR) dataset. These temperature fields are available at 32×32 square km spatial resolution and at 3 hour time intervals. Combining all test results into one final rating for each
pixel is beneficial for the generation of multi-scene clear-sky composites. The selection of the best pixel to be used in the final clear-sky product is based on the magnitude of the rating. This provides much improved results relative to other
approaches or "yes/no" decision methods.
The SPARC method has been compared to the results of supervised classification for a number of AVHRR scenes representing various seasons (snow-free summer, winter with snow/ice coverage and transition
seasons). The results show an overall agreement between the automated (SPARC) and the supervised classification at the level of 80 % to 91 %. |
| GEOSCAN ID | 221765 |
|
|