MODIS Retrieval of Cloud Optical

1
MODIS Retrieval of Cloud Optical Microphysical
Properties
Michael D. King NASA Goddard Space Flight Center

• Optical thickness, particle size (effective
  radius), and water path
• 1 km spatial resolution, daytime only, liquid
  water ice clouds
• Solar reflectance technique, VIS through MWIR
• Water nonabsorbing bands 0.65, 0.86, 1.24 µm
• Water absorbing bands 1.6, 2.1, 3.7 µm
• Land, ocean, and snow/sea ice surfaces
• Land surface 0.65 µm
• Ocean surface 0.86 µm
• Snow/ice surfaces 1.24 µm
• MODIS 1st satellite sensor with all useful SWIR,
  MWIR bands

2
Terra
Launched December 18, 1999
MODIS
MOPITT
ASTER
MISR
CERES
3
MODIS Scan Swath
4
MODerate-resolution Imaging Spectroradiometer
(MODIS)

• NASA, Terra Aqua
• launches 1999, 2002
• 705 km polar orbits, descending (1030 a.m.)
  ascending (130 p.m.)
• Sensor Characteristics
• 36 spectral bands ranging from 0.41 to 14.385 µm
• cross-track scan mirror with 2330 km swath width
• Spatial resolutions
• 250 m (bands 1 - 2)
• 500 m (bands 3 - 7)
• 1000 m (bands 8 - 36)
• 2 reflectance calibration accuracy
• onboard solar diffuser solar diffuser stability
  monitor

5
Shortwave Properties of Clouds
MODIS Atmosphere Bands
1.0
tc(0.75 µm) 16
re 4 µm
0.8
re 8 µm
re 12 µm
re 16 µm
0.6
re 20 µm
Spherical Albedo
re 20 µm vapor
0.4
0.2
0.0
0.5
1.5
1.0
2.0
2.5
3.0
3.5
4.0
Wavelength (µm)
6
Infrared Properties of Clouds
Wavelength (µm)
12
16
10
5
4
3
6
8
340
320
300
Brightness Temperature (K)
280
260
240
1500
1000
2000
2500
3000
Wavenumber (cm-1)
7
Reflection Function of Clouds as a Function of
Cloud Optical Thickness at 0.65 µm
8
Definition of Effective Radius
The effective radius re is defined by re
where r particle radius n(r) particle
size distribution
9
Infrared Properties of Clear Skies Cirrus Clouds
10
MODIS Reveals Atmospheric Moisture Details As
Never Seen Before
MODIS Water Vapor (1 km)
GOES-8 Water Vapor (4 x 8 km)
11
MODIS Channels

• Principal Channels

12
Global Level-1B Composite Image
May 28, 2001
R 0.65 µm G 0.56 µm B 0.47 µm
13
Global Level-1B Composite Image
May 28, 2001
R 0.65 µm G 0.56 µm B 0.47 µm
example data granule coverage (5 min)
14
MODIS Cloud Products

• Pixel-level (level-2) products
• Cloud mask for distinguishing clear sky from
  clouds (288 _at_ 47.4 MB)
• Cloud radiative and microphysical properties (144
  _at_ 69.6 MB 144 _at_ 14.1 MB)
• Cloud top pressure, temperature, and effective
  emissivity
• Cloud optical thickness, thermodynamic phase, and
  effective radius
• Thin cirrus reflectance in the visible
• New algorithms from greater spectral coverage,
  heritage algorithms at higher spatial resolution,
  products include QA (processing, assessment info,
  weighted statistics)
• Gridded time-averaged (level-3) atmosphere
  product
• Daily, 8-day, and monthly products (474.8, 883.2,
  883.2 MB)
• 1 1 equal angle grid
• Mean, standard deviation, marginal probability
  density function, joint probability density
  functions
• Quicklook imagery available at MODIS atmosphere
  web site
• modis-atmos.gsfc.nasa.gov

15
Retrieval of tc and re

• The reflection function of a nonabsorbing band
  (e.g., 0.86 µm) is primarily a function of
  optical thickness
• The reflection function of a near-infrared
  absorbing band (e.g., 2.14 µm) is primarily a
  function of effective radius
• clouds with small drops (or ice crystals) reflect
  more than those with large particles
• For optically thick clouds, there is a near
  orthogonality in the retrieval of tc and re using
  a visible and near-infrared band

Liquid Water Clouds - ocean surface
16
Retrieval of tc and re

• The reflection function of a nonabsorbing band
  (e.g., 0.86 µm) is primarily a function of
  optical thickness
• The reflection function of a near-infrared
  absorbing band (e.g., 2.14 µm) is primarily a
  function of effective radius
• clouds with small drops (or ice crystals) reflect
  more than those with large particles
• For optically thick clouds, there is a near
  orthogonality in the retrieval of tc and re using
  a visible and near-infrared band

Ice Clouds - ocean surface
17
Cloud Optical Microphysical Properties
Retrieval Example
Liquid Water Clouds - ocean surface
Liquid Water Clouds - ice surface
18
Cloud Optical Microphysical Properties

• Critical input
• Cloud mask
• to retrieve or not to retrieve?
• Cloud thermodynamic phase
• use liquid water or ice libraries?
• Surface albedo
• for land, ancillary information regarding
  snow/ice extent (NISE data set)
• Atmospheric correction
• requires cloud top pressure, ancillary
  information regarding atmospheric moisture
  temperature (e.g., NCEP, DAO, other MODIS
  products)
• 3.7 µm emission (band contains both solar and
  emissive signal)
• need cloud top temperature, ancillary for surface
  temperature (e.g., from NCEP, DAO, ...)

19
MODIS Cloud Mask(W. P. Menzel, S. A. Ackerman,
R. A. Frey)

• MODIS cloud mask uses multispectral imagery to
  indicate whether the scene is clear, cloudy, or
  affected by shadows
• Cloud mask is input to rest of atmosphere, land,
  and ocean algorithms
• Mask is generated at 250 m and 1 km resolutions
• Mask uses 17 spectral bands ranging from
  0.55-13.93 µm (including new 1.38 µm band)
• 11 different spectral tests are performed, with
  different tests being conducted over each of 5
  different domains (land, ocean, coast, snow, and
  desert)
• temporal consistency test is run over the ocean
  and at night over the desert
• spatial variability is run over the oceans
• Algorithm based on radiance thresholds in the
  infrared, and reflectance and reflectance ratio
  thresholds in the visible and near-infrared
• Cloud mask consists of 48 bits of information for
  each pixel, including results of individual tests
  and the processing path used
• bits 1 2 give combined results (confident
  clear, probably clear, probably cloudy, cloudy)

20
Level-2 Cloud Mask Images
April 25, 2001
21
MODIS Cloud Thermodynamic Phase(M. D. King, S.
Platnick, B. A. Baum, S. A. Ackerman, et al.
NASA GSFC, NASA LaRC, U. Wisconsin/CIMSS)

• Bispectral IR test (BT8.5-BT11, BT11 thresholds)
• Uses water/ice emissivity differences in 8.5 µm
  band (BT8.5-BT11 positive and large for ice
  clouds, small and negative for water clouds)
• 5 km resolution (currently)
• Solar test (e.g., R1.6/R0.86 ratio test, in
  development)
• Decision tree approach ecosystem-dependent
  assessment of individual cloud mask test results,
  current technique in production
• Validation with MODIS Airborne Simulator
  instrument flown on high altitude NASA ER-2 (can
  resolve water/ice spectral signatures in 1.6,
  2.1, 3.7 µm spectral bands)

22
Decision Tree for Cloud Retrievals
Results of individual cloud mask tests Cloud
mask ecosystem map
Decision Tree
When to retrieve Estimate of cloud phase
23
Cloud Phase Decision Tree Processing Path
No
cloud mask determined?
Yes
No
daytime?
Yes
Yes
shadow?
No
Determine Ecosystem Type
24
Cloud Phase Decision Tree - Ocean Processing
Example
Ocean
probably clear (11)
probably cloudy (01)
confident clear (10)
cloudy (00)
branch removed
cloud mask probability
No
Yes
heavy aerosol?
thin cirrus? R1.38 test
Stop
Stop
being evaluated
Yes
No
Yes
Ice Cloud
sunglint prob. cloudy?
No
T11-T3.9 R1.38 tests (lt-8K, lt0.035)
No
No
No
R1.38 test (gt 0.035)
T14 test (lt 240K)
Undetermined (water cloud)
Yes
Yes
Yes
Ice Cloud
Ice Cloud
Water Cloud
indicates MAS thresholds used in Arctic
25
MODIS Decision Tree Results
26
MODIS Top Properties(W. P. Menzel, R. Frey, K.
Strabala, L. Gumley, et al. NOAA NESDIS, U.
Wisc./CIMSS)

• Cloud top pressure, temperature, effective
  emissivity
• Retrieved for every 5 x 5 box of 1 km FOVs, when
  at least 5 FOVs are cloudy, day night
• CO2 Slicing technique (5 bands, 12.0-14.2 µm)
• ratio of cloud forcing in 2 nearby bands
• retrieve pc Tc from temperature profile
• most accurate for high and mid-level clouds
• Previously applied to HIRS (NOAA POES, 20 km)
• MODIS 1st satellite sensor capable of CO2 slicing
  at high spatial resolution

27
Weighting Functions for CO2 Slicing
10

• CO2 slicing method
• ratio of cloud forcing at two near-by wavelengths
• assumes the emissivity at each wavelength is
  same, and cancels out in ratio of two bands
• The more absorbing the band, the more sensitive
  it is to high clouds
• technique the most accurate for high and middle
  clouds
• MODIS is the first sensor to have CO2 slicing
  bands at high spatial resolution (1 km)
• technique has been applied to HIRS data for 20
  years
• retrieved for every 5 x 5 box of 1 km FOVs, when
  at least 5 FOVs are cloudy, day night

Central Wavelength (µm) 12.020 13.335 13.635 13.93
5 14.235
Channel 32 33 34 35 36
100
Pressure (hPa)
35
36
34
32
33
1000
0.0
0.2
0.4
0.6
0.8
1.2
1.0
Weighting Function dt(n,p)/d ln p
28
Cloud top pressure
1000
850
700
550
Cloud Top Pressure (hPa)
400
250
100
29
Cloud top temperature
320
295
273
Cloud Top Temperature (K)
250
225
250
180
30
Atmospheric Correction

• Cloud library calculations give cloud-top
  quantities (no atmosphere)
• atmosphere included during retrieval
• Rayleigh scattering
• iterative approach applied to 0.65 µm band only
  (used over land)
• important for thin clouds and for any clouds with
  large solar/view angle combinations
• Atmospheric absorption
• Well-mixed gases a function of pc, water vapor
  absorption a function of profile both a weak
  function of temperature
• Assume above-cloud column water vapor amount the
  primary parameter, vapor profile of minor
  consequence
• Library calculations made at a variety of pc,
  above-cloud column water amounts (scaled from
  various water vapor and temperature profiles),
  geometries
• using MODTRAN 4.0 with scripts for 2-way
  transmittance calculations
• requires cloud top pressure, and ancillary
  information regarding atmospheric moisture
  (currently using NCEP)

31
Two-way Atmospheric Path Transmittance (1/µ
1/µ0)

• pc 900 hPa
• w 2.0 g-cm-2 above-cloud precipitable water
• µ0 0.8

0.86, 1.24 µm
1.6 µm
0.67 µm
2.1 µm
3.7 µm (1-way µ path)
Absorption transmittance
3.7 µm
cosine of viewing zenith angle (µ)
32
Ecosystem Map(A. H. Strahler, C. B. Schaaf, et
al. Boston University)

• MOD12 (IGBP ecosystem classification) USGS
  water tundra

33
Surface Albedo Surface albedo ecosystem
MOD43 (Strahler, Schaaf et al.) aggregation
34

• Albedo Movies
• Loops through bands 0.65, 0.86,
• 1.24, 1.64, 2.1, and 3.7 µm
• Loops through seasonal equinox
• and solstice, progressing from
• Julian days 91, 173, 293, 356

• Ecosystem Color Scheme
• Pink Crops
• Green Trees
• Yellows Barren/Deserts
• Blues Savannas

35
Cloud Optical Thickness in the ArcticProvisional
Production Code (edition 3)
June 2, 2001
tc
20
15
10
5
0
36
Cloud Optical Thickness in the ArcticProvisional
Production Code (new correction)
June 2, 2001
tc
20
15
10
5
0
37
Cloud Effective Radius in the ArcticProvisional
Production Code (edition 3)
June 2, 2001
re(µm)
40
34
28
22
16
10
4
38
Cloud Effective Radius in the ArcticProvisional
Production Code (new correction)
June 2, 2001
re(µm)
40
34
28
22
16
10
4
39
Level-2 Global Cloud Images
October 1, 2001
40
SAFARI 2000 Core Sites
Mongu
Okavango Delta
Etosha Pan
Maun
Sua Pan
Swakopmund
Tshane
Skukuza
Namib Desert
Inhaca Island
Drakensberg escarpment
41
ER-2, C-130 ground tracks
MODIS true color 11 Sept. 2000, 0940 UTC
ER-2
Validation region
C-130 (red in-cloud portions)
42
UK C-130 in situ droplet radius, liquid water
content 11 Sept. 2000, 0941-0953 UTC (S. Osborne,
Met Office)
43
Previous SAFARI 2000 Namibian Sc studies

California central valley fog
AVHRR
California Sc
MAS
Arctic stratus
(Jan. April 1989, 2 AM scenes, )
Namibian Sc
AVHRR
(Sep. 1999, 3 PM scenes)
(Oct. 1995, SATE-2 validation)
ATSR-2
0
5
10
15
20
25
cloud droplet effective radius (µm)
44
Comparison of Visible Optical Thickness(G. G.
Mace, S. Bensen, K. Sassen University of Utah)
Retrieved Optical Thickness
MOD06 Optical Thickness
45
Gridded Level-3 Joint Atmosphere Products(M. D.
King, S. Platnick, P. A. Hubanks, et al. NASA
GSFC, UMBC)

• Daily, 8-day, and monthly products (474.8, 883.2,
  883.2 MB)
• 1 1 equal angle grid
• Mean, standard deviation, marginal probability
  density function, joint probability density
  functions

46
Cloud Optical Thickness (M. D. King, S.
Platnick, M. Gray, E. Moody, et al. NASA GSFC,
UMBC)
Level-3 Monthly April 2001
tc
20
16
12
8
4
0
47
Cloud Effective Particle Radius (M. D. King, S.
Platnick, M. Gray, E. Moody, et al. NASA GSFC,
UMBC)
Level-3 Monthly April 2001
re(µm)
40
34
28
22
16
10
4
48
MODIS L3 aggregation from 6x 6 grid off
Namibian coastliquid water clouds
L3 product bin sizes (liquid water clouds)
49
MODIS L3 aggregation from 6x 10 grid off
California coastliquid water clouds
May 28, 2001
L3 product bin sizes (liquid water)
50
Cloud Top Pressure(W. P. Menzel, R. Frey, K.
Strabala, L. Gumley, et al. NOAA NESDIS, U.
Wisconsin/CIMSS)
Level-3 Monthly April 2001
pc (hPa)
1000
900
800
700
600
500
400
300
51
Precipitable Water over Land Sunglint(B. C.
Gao, et al. Naval Research Laboratory)
Level-3 Monthly April 2001
q (cm)
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
52
Ship Tracks in NE Pacific

• Ship Tracks occur in marine stratocumulus regions
  of the globe
• California, Azores, Namibia, and Peru
• Conditions for formation
• High humidity
• Small air-sea temperature difference
• Low wind speed
• Boundary layer between 300 and 750 m deep
• Enhanced reflectance of clouds at 3.7 µm
• Larger number of small droplets arising from
  particulate emission from ships

53
Cloud Physics

• When water vapor condenses in the atmosphere, it
  always does so on some solid nucleus
• The dust particles in the air form the nuclei on
  which it condenses
• If there was no dust in the air there would be no
  fogs, no clouds, no mists, and probably no rain
• John Aitken (1880) On dust, fogs, and
  clouds

54
Ship Tracks in NE Pacific

• MODIS
• 2.1 µm

October 2, 2000
55
Ship Track Formation
N 40 cm-3 W 0.30 g m-3 re 11.2 µm
N 100 cm-3 W 0.75 g m-3 re 10.5 µm
56
Distribution of Ship Tracks in NE Pacific during
June 1994

• Ship Tracks identified from AVHRR imagery
• 1362 identified during June 1994 off the US west
  coast
• The head of each track is denoted by a dot
• The greatest concentration of tracks occurs along
  the great-circle shipping lanes
• Ship Track Characteristics
• 296 233 km long
• 7.3 6 hours old
• 9 5 km wide
• 16 8 km from the head of the ship track and 25
  15 minutes after ship passed
• Boundary layer between 300 and 750 m deep
• No tracks formed in boundary layers gt 800 m deep

57
Ship Track off West Coast of Chile

• Photo taken by shuttle astronauts during mission
  STS-65
• July 20, 1994

58
Image of a Ship and Corresponding Ship Track in
NE Pacific

• June 13, 1994
• RC-10 Camera

250 m
1400 m
59
Ship Tracks in NE Pacific June 30, 1994

• AVHRR imagery at 3.7 µm
• Wind from NNW at 13.5 ms-1
• Ship motion vectors

60
Ship Track Characteristics
61
Ship Track Occurrence

• Ship tracks form preferentially in boundary
  layers shallower than about 800 m depth

Boundary Layer Depth (m)
Number of Cases and Mean Tracks/Case
62
Global Occurrence Statistics
49 tracks
572 tracks
809 tracks
5537 tracks
Analysis Pending
470 tracks
805 tracks
63
Summary

• Particles emitted by ships increase concentration
  of cloud condensation nuclei (CCN) in the air
• Increased CCN increase concentration of cloud
  droplets and reduce average size of the droplets
• Increased concentration and smaller particles
  reduce production of drizzle (100 µm radius)
  droplets in clouds
• Liquid water content increases because loss of
  drizzle particles is suppressed
• Clouds are optically thicker and brighter along
  ship track

64
Indirect Aerosol Effects

• The increase in CCN of industrial origin might
  explain why the Northern Hemisphere has warmed
  less than the Southern Hemisphere over the last
  50 years
• Even if some compensation in surface warming
  occurs because of changes in sulfate and
  greenhouse gases, it is not clear whether that
  compensation will continue in the future
• The sulfate effect would tend to act only
  regionally, whilst the greenhouse forcing is
  global
• IPCC Scientific Assessment of Climate Change
  (1990)