As the majority of the cloud analysis methods based on satellite observations the cloud detection scheme is the first step of the POLDER "ERB & clouds" processing line. It takes advantage of the original capabilities of the POLDER instrument (spectral polarization and multiviewing) in order to recognize cloud-contaminated pixels. Obviously this step is crucial since it controls further processing and has a major impact on determining other cloud products.
Basically the cloud detection scheme is similar to an usual threshold method. It uses six threshold tests based on pressure, reflectance, polarized reflectance and spectral variability. The physical principles of the method are extensively developed in Parol et al., 1999 and Sèze et al., 1999.

A short description is given here.

A more complete description can be downloaded from the "ERB, WV & clouds" algorithm documents for POLDER-2. If there are further questions regarding the cloud detection scheme, please contact Pr. F. Parol (at LOA)

See below the main parameters related to cloud detection that appears in the level-2 POLDER "ERB, WV & clouds" product.

CC	Cloud Cover, defined as the number of cloudy pixels divided by the total number of pixels. Click to view image Note that this parameter is also furnished for each viewing direction
CC±	Uncertainty on cloud cover related to re-classification, i.e. number of pixels firstly classified as indeterminate and finally classified as clear (+) /cloudy (-) divided by the total number of pixels
QCC	Quality index related to cloud detection. It ranges from 0 (very bad) to 1 (very good). Click to view image

See below the main parameters related to cloud detection that appears in the level-3 POLDER "ERB, WV & clouds" product.

M(CC)	Monthly mean of the cloud cover. Click to view image
TD(CC)	Temporal standard Deviation of the cloud cover
M(CC±)	Monthly mean of the uncertainty on cloud cover related to re-classification

---

© CNES Copyrights 2004