CLOUD MASK AND QUALITY CONTROL FOR SST WITHIN THE ADVANCED CLEAR SKY PROCESSOR FOR OCEANS (ACSPO)

1
CLOUD MASK AND QUALITY CONTROL FOR SST WITHIN THE
ADVANCED CLEAR SKY PROCESSOR FOR OCEANS (ACSPO)
A. Ignatov1 (GOVERNMENT PRINCIPAL INVESTIGATOR),
B. Petrenko2 1NOAA/NESDIS/STAR, 2IM Systems
Group, Inc.

• 1. Requirements
• The Advanced Clear-Sky Processor for Oceans
  (ACSPO) is developed at NESDIS to generate clear
  sky brightness temperatures (BT), aerosol and Sea
  Surface Temperature (SST) from AVHRR-like
  measurements at a pixel resolution.
• The purpose of the ACSPO Quality Control (QC) is
  to screen out the pixels, not usable for
  clear-sky products while preserving as many
  useful pixels as possible. Since the major part
  of pixel contaminations is caused by clouds, QC
  can be treated as an efficient cloud masking
  algorithm. QC extensively uses online clear-sky
  radiative transfer simulations and real-time NWP
  information.
• In addition to QC, the future ACSPO versions will
  include the Cloud Mask (CM) module, which will
  screen out clouds using only static ancillary
  data. This will enhance robustness and stability
  of the cloud masking process in case if the NWP
  information is unavailable 1.
• 2. The Features of ACSPO
• The ACSPO QC has some unique features, which make
  it different from the majority of existing cloud
  masking algorithms 4.
• 2.1. Emphasis on On-line Clear-Sky RTM
  simulations
• Instead of exploiting cloud emission and
  reflection properties, ACSPO QC makes an emphasis
  on more accurate online clear-sky simulations.
  ACSPO incorporates the Community Radiative
  Transfer Model, which simulates clear-sky BT from
  0.25o High Resolution-Blended SST (OISST) 2)
  and 6-hour 1o NCEP GFS upper air fields 3.
  Anomalies DBTObserved BT- CRTM BT and
  DSSTRetrieved BT OISST are used as input for
  QC.
• 2.2. Accounting for Biases in Observed BT and
  Retrieved SST

2.3. Detecting Ambient Cloudiness ASPO
QC includes two sequential tests, which use DSST
as a quality predictor. The Static SST test
initially separates all ocean pixels into
clear-sky and cloudy clusters by detecting
unrealistically cold DSST values. The Adaptive
SST test refines the initial clasterization based
on statistics of DSST over clear-sky and
cloudy pixels in the neighborhood of every
clear-sky pixel. This approach allows to avoid
excessively strict restrictions on realistic
DSST values and at the same time to reject
ambient clouds, typically surrounding cloudy
systems. 2.4. Advanced Detection of Subpixel
Clouds Existing cloud masking
algorithms typically include spatial uniformity
tests, which detect subpixel cloudiness by
elevated spatial variability of observed BT. The
typical drawback of this kind of tests is that
intensive thermal fronts in the ocean can be
misclassified as clouds. In the ACSPO QC the
Uniformity test has the following peculiarities
- it applies to retrieved SST rather than
to observed BT this allows direct addressing
cloud contaminations in the retrieved variable.
- The retrieved SST field is passed through
the 2D median filter and the test applies to the
difference retrieved SST median (retrieved
SST). This improves discrimination between
thermal fronts and random subpixel clouds.
- In the daytime detection of subpixel clouds
further improves with the BT/Reflectance
Cross-Correlation test. This test enhances
detection of subpixel clouds detecting
correlation of small negative SST variations with
positive variations in Ch2 reflectance

• 3. Applications
• 3.1. AVHRR

Fig. 2. Composite map of DSST from nighttime
METOP A measurements on August 1, 2009 over the
Gulf of Mexico. a) after Static SST test b) after
Static and Adaptive SST tests
ACSPO
Fig. 4. The MetOp-A FRAC images of the Gulf of
Mexico obtained from ACSPO and OSI SAF SST
product.
The ACSPO is currently used in OSDPD for
operational processing of AVHRR data from the
platforms NOAA-16, -17, -18,-19 and MetOp-A.
Comparison with other world-class cloud mask
products (CLAVRx 5 OSI SAF 6) shows similar
or better performance of the ACSPO cloud masking
algorithm.
OSI SAF
Fig. 3. SST field east off the South America as
observed with MetOp-A AVHRR on August 1, 2008
(night) and processed with CLAVRx 5 BT
uniformity tests (a) and with ASPO SST Uniformity
test (b)
Fig. 5. The example of the full disk SST and
cloud distributions generated with ACSPO from
MSG2 SEVIRI measurements 1230 UTC, June 6 2008.
The ACSPO will be also used for processing data
of the Advanced Baseline Imager (ABI) onboard the
GOES-R satellite. The GOES-R SST and QC
algorithms are modified from AVHRR-ACSPO using
MSG2 SEVIRI as a proxy 7.
Fig. 1. Histograms of DSST over clear pixels
for 4 platforms, carrying AVHRR sensors, August
1-7, 2008
Another potential ACSPO application is the
Visible Infrared Imager Radiometer Suite (VIIRS)
onboard the National Polar-orbiting Operational
Environmental Satellite System (NPOESS)