Cloud Mask:
Description and examples

Contents:

CloudMask
  1. Goal of the Cloud Mask product
  2. Cloud Mask algorithm description
  3. List of inputs for Cloud Mask
  4. Coverage and resolution
  5. Description of Cloud Mask output
  6. Example of Cloud Mask visualisation

Goal of the Cloud Mask product

This product attempts to delineate all absolutely cloud-free pixels in a satellite scene with a high confidence. In addition, it will identify cloud free snow or ice contaminated pixels when illumination allows.

Cloud Mask algorithm description

The cm algorithm is based on a multi-spectral threshold technique applied to each pixel of the satellite scene. The thresholding algorithm, utilising all 5/6 spectral channels of the avhrr/2 or avhrr/3 sensor (or the corresponding channels from the viirs sensor), nwp short range forecast data, 1 km gis (digital elevation model and landuse) data and land emissivity maps. The scheme makes use of off-line radiative transfer simulations (6S and rttov) of cloud free atmospheres, to estimate, prior to the reception of satellite data, the optimal thresholds valid for the given satellite scene.

List of inputs for Cloud Mask

Satellite data: 0.6 microns, 0.9 microns, 11 microns, 12 microns and (1.6 or 3.7 microns), from avhrr or from viirs. From viirs also channel 8.5 microns.

Auxillary data:

Coverage and resolution

This product is available in full imager resolution and for the whole field of view.

Description of Cloud Mask outputs

The Cloud Mask output consists of:

In addition to the above, the output includes a threshold test flag, for internal use only, identifying the decisive threshold tests.

Main output

Cloud Mask Classes
Class number Class name
0 Non-processed
1 Cloud Free
2 Cloud contaminated
3 Cloud Filled
4 Snow/ice
5 Unclassified

The main output is given by the six categories listed above. Cloud free land and cloud free sea are one and the same category, so in order to make an output image like the above the user will have to apply a land/sea mask herself, or use the information available in the quality flag (bit number 0 - see below).

Quality flag

The quality flag occupies sixteen bits, and provide the user with valuable information on the conditions under which the cloud mask processing was performed, and on the quality of the thresholding. The exact outline of the quality flag is given below.

Seven bits to describe the illumination and environmental conditions under which the cloud mask was derived - numbers refer to the individual bits:

Illumination & Environmental conditions
Bit number Meaning of the bit
0 Land/Not land
1 Coast/Not coast
2 Night/Not night
3 Twilight/Not twilight
4 Sunglint/No sunglint
5 High terrain/low terrain
6 Inversion/No inversion: Low level inversion present or not.

Two bits to describe the use/availability of nwp data and the availability of the avhrr channels:

Missing data
Bit number Meaning of the bit
7 nwp data has been used/nwp data not used.
8 One or more avhrr channels missing.

Two bits to describe the quality of the thresholding - if the measurement was close to one of the thresholds of the active test the result will be assigned a low confidence:

Thresholding quality
Bit number Meaning of the bit
9 Low quality/High quality
10 Very low quality - Pixel has been reclassified after spatial smoothing
11 State before spatial smoothing was cloud contaminated
12 State before spatial smoothing was cloudy

Bit number nine is set (meaning low quality) when the value of a pixel in some feature is close to the threshold determining the output.

Bit number ten is set when an isolated pixel has been changed from cloudy to cloudfree (or vice-versa) after applying spatial smoothing. This spatial smoothing will only be applied when the T11-T37 test was the one determining the output value.

Three bits for describing the sea ice:

Sea ice data
Bit number Meaning of the bit
13 External sea-ice information used (e.g. osisaf ice maps).
14 Ice information derived from nwp model.
15 Sea-ice present according to sea ice maps (osisaf or nwp).

The sea ice schemes are used in case the ice concentration is higher then 30%.

Dust and Volcanic plume flag

The information on excesive aerosol content due to dust clouds or volcanic plumes have been excluded from the main output, due to poor separability using the spectral channel of the avhrr. Instead this information is put in a separate output flag:

Dust and Volcanic Plume
Bit number Meaning of the bit
0 Non-processed (containing no data or corrupt data)
1 Not contaminated
2 Smoke
3 Dust cloud
4 Fire
5 Unclassified (due to known separability problems)

Example of Cloud Mask visualisation

noaa19_20120514_1125_16826.mesanX.cloudmask.thumbnail.jpg (62000 bytes) noaa19_20120514_1125_16826.mesanX.cloudmask.thumbnail.jpg (60000 bytes)
unproc.gif (96 bytes) Unprocessed
cmask_land.gif (899 bytes) Cloud free land
cmask_sea.gif (897 bytes) Cloud free sea
cloudcont.gif (902 bytes) Cloud contaminated
cloudfill.gif (898 bytes) Cloud filled
snow.gif (899 bytes) Snow/Ice contaminated
unclass.gif (101 bytes) Unclassified
lowquality.gif (899 bytes) Low quality

Here are shown example images of the pps. Cloud Mask product over northern Europe. To the left it is the cloud mask only, and to the right also the cloudfree pixels with low quality, according to the processing flag bit number 9, are shown. The low quality pixels are shown in purple.

The data are from a noaa19 overpass (orbit 16826) May 14, 2012, 11:25 UTC, as received and processed at smhi, Norrköping, Sweden. The corresponding rgb images using avhrr channel 0.6µm,0.9µm,11µm and 3.7µm,11µm,12µm are shown below for comparison with the Cloud Mask.

The cloud mask images shown here have been generated with the nwcsaf/pps software. They display only part of the information available in the product. The pps Cloud products are most appropriately visualised using a dedicated Graphical User Interface (gui) tool allowing the display of all relevant information available in the hdf5 file, including flags, geographical location, etc. smhi has developed such a tool running under Unix and freely available upon request. Otherwise the hdf-viewer (hdfview) available from ncsa is recommended.

noaa19_20120514_1125_16826.mesanX.cloudmask.thumbnail.jpg (62000 bytes) noaa19_20120514_1125_16826.mesanX.cloudmask_qual.thumbnail.jpg (60000 bytes)
noaa19_20120514_1125_16826.mesanX.avhrr.ch_06_09_11.thumbnail.jpg (476000 bytes) noaa19_20120514_1125_16826.mesanX.avhrr.ch_37_11_12.thumbnail.jpg (476000 bytes)