August 1, 2014

Earth Radiation Budget & Cloud Products

The original instrumental capabilities of Parasol (multidirectionality, multipolarization, and multispectrality) offer opportunity for deriving cloud parameters at global scale.
Parasol Earth Radiation & Cloud products provide the following set of parameters:

  • cloud cover
  • cloud phase
  • cloud optical thickness and albedo
  • SW albedo
  • 2 cloud pressures: from oxygen absorption and from Rayleigh scattering
  • Water vapor integrated content

Daily products and monthly syntheses are produced at 20 km resolution (after cloud detection performed at full resolution, 6 km, and for every direction).

The interesting and unique feature about Parasol retrievals for clouds is its ability to drive the inversion process by a proper discrimination of cloud phase and cloud microphysical model thanks to the directional polarization measurements.
Moreover, the consistency check of the retrieved parameters direction by direction is an indicator of the performance of the retrieval algorithms.

Cloud detection and Cloud Cover

The first stage of the "ERB & clouds" line is the recognition of cloud-contaminated pixels. This step is crucial since it controls further processing and it has a major impact on determining others products.
The cloud detection algorithm is based on a series of sequential threshold tests applied to each individual pixel (6 km) and for every viewing direction. Five tests aim at detecting clouds and three more tests are added in order to identify the clear pixels. If a PARASOL pixel fails all of the tests, it remains unclassified for a given viewing direction. If the pixel remains undetermined, it is then relabelled as clear or cloudy depending on the classification of the neighbouring pixels and the spatial variability of the reflectance at 865nm. Afterwards, when all of the elementary pixels are identified as cloud-free or cloudy the cloud cover is computed at the super-pixel scale (3 x 3 pixels, 20km resolution), direction by direction.
Clearly, the multidirectional capability of Parasol is very useful for discriminating between clear and cloudy pixels. As an example, over ocean a simple reflectance threshold test can always be applied since a Parasol pixel can be observed with angular configuration outside the glint region.

Cloud cover
Also available for each viewing direction
Cloud cover confidence

On the other hand, as illustrated on the figure here-above, a cloud cover confidence index is defined from the result of the cloud detection algorithm. 1 is for high confidence and 0 is for low confidence. This parameter takes advantage of the multidirectional capability of Parasol. For instance light blue lines west of the Parasol orbits are mainly due to the presence of the sunglint in these angular regions.

Comparison to MODIS has been performed at level 2 and at level 3 at global scale.

Example of comparison at level 2 spatial resolution

Besides the overall consistency of the two products work is in progress to understand/explain the differences (see for instance over land in the middle part of the figure)


Distribution of (Polder-MODIS) cloud cover differences for September 24, 2005 (percentage of cases in 0.05 bins)

mean dif. = -0.05
rms dif. = 0.27

Over ocean:
mean dif. = -0.06
rms dif. = 0.27
Over land:
mean dif. = -0.01
rms dif. = 0.25
Comparison of Polder (blue line) and MODIS (red line) cloud cover distributions (a) over ocean and (b) over land for September 24, 2005 . The Y-axis correspond to the percentage of cases in 0.1 bins.

Cloud thermodynamic phase

Parasol cloud phase is unique and based on polarization signatures. The method has been refined to better account for complex mixed cases. A phase index (clear, ice, liquid, mixed, undetermined) with confidence level is now provided. The ability do detect of strong aerosol events over underlying liquid clouds has been proved.

The retrieval of cloud thermodynamic index allows for constructing monthly syntheses of liquid or ice cloud frequency which are very useful for comparison to modelling at global scale.


September 2005


Liquid cloud fractionIce cloud fraction

Cloud optical thickness - Spectral albedoes - Shortwave albedo

To derive the albedo from bidirectional reflectance observations, or equivalently, the hemispherical flux from radiance observations, several approaches are possible. Ours is based on radiative transfer modelling. In a first step, the narrowband albedoes are derived from bidirectional reflectances using a radiative transfer model. These retrievals are performed at 490, 670, and 865nm. Spectral cloud optical thicknesses as well as narrowband albedoes are derived by using a look-up table (LUT) technique. LUTs are calculated by using a plane-parallel radiative transfer model applied to two cloud types (ice and liquid water) depending on the cloud thermodynamical phase index derived from Parasol polarized reflectances.

Cloud optical thickness
Spectral albedo
670nm over land
865nm over ocean

First comparisons to the MODIS equivalent cloud parameters are very encouraging

Cloud optical thickness distributions

Relative bias is -11% (-19%) over ocean (over land)

In a second step of the analysis, all the three narrowband albedoes are used to estimate the broadband shortwave albedo.

Example of monthly mean shortwave albedo

September 2005

In a very near-future the so-derived SW albedo will be compared to the CERES estimation which is derived using a totally different technique based on statistical Angular Distribution Models.

Cloud pressure

Parasol provides 2 pressures. The Rayleigh pressure uses polarized measurements at 490 nm, whereas oxygen pressure relies on radiances at 763 and 765 nm. The former is sensitive to the cloud top, the latter to cloud middle.

Cloud pressure derived from the polarized component of the Rayleigh scattering
Cloud pressure derived from absorption by O² in the A-band

The 2 Parasol pressures have been compared to the so-called MODIS collection 4 day pressures which combine "CO²-slicing method" and "standard thermal IR method". The general trend of the comparisons is that high clouds are seen higher by MODIS than by PARASOL (35 hPa) whereas low clouds are seen lower (55 hPa).

Rayleigh pressure versus MODIS day pressure over oceanRayleigh pressure versus MODIS day pressure over ocean, 2005/09/24

The Parasol pressures have also been compared to Lidar (SIRTA/STRAT) measurements.

Rayleigh pressure versus Lidar cloud top pressureOxygen pressure versus mean cloud pressureRayleigh pressure versus Lidar cloud top pressure (left) and Oxygen pressure versus mean cloud pressure (right)

For cloud optical thickness greater than 4 (red dots):

  • the mean difference between the lidar cloud top pressure and the Rayleigh pressure is +60 hPa (rms dif. = 190 hPa)
  • the mean difference between the lidar mean cloud pressure and the Oxygen pressure is -30 hPa (rms dif. = 30 hPa)

Whereas ground measurements provide very few matchup situations, Calipso and Cloudsat profiles are expected to soon help resolve the discreapancies and assess precisely the performance of the various sensors' pressure estimates.

Water Vapor

Parasol retrieves the water vapor integrated content over land (and over the ocean in glitter conditions) by the differential absorption method applied to the 910 and 865 nm channels. The algorithm is tuned over ocean wrt AMSR (Polder used SSMI). The performance was evaluated for Polder 2 against radiosonde measurements and showed no bias and a rms difference of 3.4 kg/m2.


For more information about the algorithms refer to Polder background.
To access the user services please go to ICARE

Cloud properties retrieval from synergy between Polder3/Parasol and MODIS/Aqua, Preliminary results, AMS 2006, Riedi, Cécile Oudard, Jean-Marc Nicolas, Laurent Labonnote, Frédéric Parol, Steven Platnick et al

View of Nabi Typhon by Parasol (AVI format)