A Cloud Typing Approach for the VIIRS

1
A Cloud Typing Approach for the VIIRS

• Michael J. Pavolonis
• CIMSS/SSEC
• Andrew K. Heidinger NOAA/NESDIS

2
Motivation

• Analysis has shown that the VIIRS baseline
  approach for cloud typing can be improved.
• AVHRR cloud typing has demonstrated that
  detection of multi-layer cloud is possible (not
  in VIIRS baseline)
• NGST has shown an interest in using a multi-layer
  cloud flag in the generation of other cloud
  EDRs.
• NGST wants to reduce the amount of ambiguous
  cloud type from the VIIRS algorithm.

3
Cloud Types

• Liquid Water
• Supercooled Liquid Water/Mixed Phase
• Opaque Ice
• Non-opaque Ice (Cirrus)
• Overlapping Clouds

4
Needed Inputs

• REF (0.65mm)
• REF (1.38mm)
• REF (1.65mm)
• REF (3.75mm)
• EMS (3.75mm)
• BT (11mm)
• BTD (3.75mm - 11mm)
• BTD (8.5mm 11mm)
• BTD (11mm 12mm)

• Solar Zenith Angle
• Viewing Zenith Angle
• Relative Azimuth Angle
• Surface Type (water, land, desert)
• Cloud Mask

Day only Night only Both day and night
5
(No Transcript)
6
(No Transcript)
7
Cloud Overlap Detection I Split Window BT Test
Additional Criteria 1. REF (0.65mm) gt 0.30 (30)
2. BT (11mm) lt 270.0 K BTD (11mm - 12mm)
threshold is also a function of solar zenith
angle and viewing zenith angle.
8
Cloud Overlap Detection II Near-Infrared
Reflectance Test
Additional Criteria 1. REF (1.38mm) lt 0.40
(40) 2. RAT (1.65mm/0.65mm) lt 1.0 3. If (REF
(1.38mm) lt 0.08 (8)) Check BTD (11mm - 12mm) 4.
BT (11mm) lt 280.0 K REF (1.65mm) threshold is
also a function of scattering angle.
9
Overlap Detection Summary

• VIIRS algorithm if a given pixel passes EITHER
  the split window BT test or the near-infrared
  reflectance test, then cloud overlap is said to
  be present.
• The VIIRS algorithm is discussed in detail in the
  recently accepted paper Pavolonis, M. J. and A.
  K. Heidinger Daytime cloud overlap detection
  from AVHRR and VIIRS, J. Appl. Meteor.

10
(No Transcript)
11
Cirrus Detection
VIIRS algorithm 1. BTD (11mm - 12mm) gt THRES 2.
REF (3.75mm) lt THRES 3. REF (1.38mm) gt THRES
12
(No Transcript)
13
Other Tests
VIIRS IR/NIR Test BTD (8.5mm - 11mm) test is
supplemented with REF (3.75mm), REF (1.38mm), and
RAT (1.65mm/0.65mm).
14
Daytime VIIRS Algorithm (Tropical Eastern Pacific
- MODIS)
15
Daytime VIIRS Algorithm (South America - MODIS)
16
Daytime Cloud Overlap Validation

• The millimeter cloud radar merged moments product
  from the ARM SGP and TWP sites was used to
  validate each algorithm.
• 180 cloudy TERRA-MODIS scenes were used.
• Thirty minute radar time series (centered on
  Terra overpass time) were grouped into 3
  categories single layer cloud, cloud overlap
  present less than 50 of time series (but greater
  than 0), and cloud overlap present at least 50
  of time series.
• The overlap fraction from each algorithm was
  calculated as Noverlap/Ncloudy for MODIS pixels
  within 15 km of radar.

17
VIIRS 0
VIIRS 15
VIIRS 82
Figures are from http//www.met.utah.edu/mace/ho
mepages/research/eos.html
18
Validation Continued
19
(No Transcript)
20
(No Transcript)
21
Nighttime Cloud Overlap Detection
These thresholds are used for mid-latitude
conditions. Daytime signal is stronger.
22
(No Transcript)
23
Other Nighttime Tests

• Split window cirrus test is used with the added
  condition that EMS (3.75mm) must be greater than
  some threshold.
• Bi-spectral BTD (8.5mm-11mm) as a function of BT
  (11mm) test is applied to non-overlapped and
  non-cirrus cloudy pixels.

24
Nighttime VIIRS Algorithm (Eastern Pacific -
MODIS)
25
Nighttime Cloud Overlap Validation
Cloud radar data over Barrow, AK and AVHRR data
were used in this preliminary comparison.
26
Cloud Phase Validation/Comparison
VIIRS and MODIS-IR algorithms were applied to
MODIS data over the North Slope of Alaska ARM
site (NSA). A total of 67 cases were used in
this analysis. For this comparison, only data
when NSA radar indicated single layer mixed phase
cloud was used. Further phase validation/comparis
on work is ongoing.
27
Daytime Global VIIRS Cloud Type (using MODIS)
VIIRS cloud type on a 0.5 degree grid with data
from the descending node of TERRA-MODIS on April
4, 2003.
28
Daytime Global VIIRS Baseline (MODIS-IR
Algorithm) Cloud Type
MODIS cloud phase on a 0.5 degree grid with data
from the descending node of TERRA-MODIS on April
4, 2003. Regions of VIIRS-detected cloud
overlap generally correspond to ice clouds.
29
Nighttime Global VIIRS Cloud Type (using MODIS)
VIIRS cloud type on a 0.5 degree grid with data
from the ascending node of TERRA-MODIS on April
4, 2003.
30
Nighttime Global VIIRS Baseline (MODIS-IR
Algorithm) Cloud Type
MODIS cloud phase on a 0.5 degree grid with data
from the ascending node of TERRA-MODIS on April
4, 2003.
31
Daytime Global VIIRS Cloud Type (using MODIS)
VIIRS cloud type on a 0.5 degree grid with data
from the descending node of TERRA-MODIS on April
4, 2003.
32
Nighttime Global VIIRS Cloud Type (using MODIS)
VIIRS cloud type on a 0.5 degree grid with data
from the ascending node of TERRA-MODIS on April
4, 2003.
33
Summary

• The VIIRS algorithm can be readily implemented
  (no ancillary data sets are needed) and is
  computationally efficient.
• Works on single pixels.
• Globally applicable.
• Accurate classification of optically thin clouds
  can still be problematic.