VIIRS Aerosol Optical Depth Algorithm and Products Istvan Laszlo NOAA VIIRS Aerosol Science and Operational Users Workshop November 21-22, 2013 National Center for Weather and Climate Prediction (NCWCP) College Park, MD
Feb 23, 2016
VIIRS Aerosol Optical Depth Algorithm and Products
Istvan LaszloNOAA
VIIRS Aerosol Science and Operational Users Workshop November 21-22, 2013
National Center for Weather and Climate Prediction (NCWCP)
College Park, MD
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 2
Name Organization Major TaskKurt F. Brueske IIS/Raytheon Code testing support within IDPS
Ashley N. Griffin PRAXIS, INC/NASA JAM
Brent Holben NASA/GSFC AERONET observations for validation work
Robert Holz UW/CIMSS Product validation and science team support
Nai-Yung C. Hsu NASA/GSFC Deep-blue algorithm development
Ho-Chun Huang UMD/CICS SM algorithm development and validation
Jingfeng Huang UMD/CICS AOT Algorithm development and product validation
Edward J. Hyer NRL Product validation, assimilation activities
John M. Jackson NGAS VIIRS cal/val activities, liaison to SDR team
Shobha Kondragunta NOAA/NESDIS Co-lead
Istvan Laszlo NOAA/NESDIS Co-lead
Hongqing Liu IMSG/NOAA Visualization, algorithm development, validation
Min M. Oo UW/CIMSS Cal/Val with collocated MODIS data
Lorraine A. Remer UMBC Algorithm development, ATBD, liason to VCM team
Andrew M. Sayer NASA/GESTAR Deep-blue algorithm developmentHai Zhang IMSG/NOAA Algorithm coding, validation within IDEA
VIIRS Aerosol Cal/Val Team
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 3
Outline
• VIIRS instrument• Aerosol algorithm
– history– over-land algorithm– over-ocean algorithm
• VIIRS vs. MODIS algorithm• VIIRS aerosol products• Summary
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 4
VIIRSVisible Infrared Imaging Radiometer
Suite (VIIRS)• cross-track scanning radiometer with
~3000 km swath – full daily sampling• 7 years lifetime• 22 channels (412-12,016 nm)
– 16 of these are M bands with 0.742 x 0.776 km nadir resolution
– aerosol retrieval is from M bands• high signal-to-noise ratio (SNR):
– M1-M7: ~200-400– M8-M11: ~10-300
• 2% absolute radiometric accuracy• single look• no polarization
Band name
Wavelength (nm)
Bandwidth (nm)
Use in algorithm
M1* 412 20 LM2* 445 14 LM3* 488 19 L, TL TOM4* 555 21 TOM5* 672 20 L, O, TOM6 746 15 O
M7* 865 39 O, TLM8 1,240 27 O, TL, TO M9 1,378 15 TL
M10 1,610 59 O, TL, TOM11 2,250 47 L, O, TL,
TOM12 3,700 191 TLM13 4,050 163 noneM14 8,550 323 noneM15 10,763 989 TL, TOM16 12,016 864 TT, TO
*dual gain, L: land, O: ocean; T: internal test
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 5
VIIR
S (c
ont)
• Sensor Data Records (SDRs), converted from raw VIIRS data (RDR), are used in the VIIRS aerosol algorithm
• Processing is on a granule by granule basis– VIIRS granule typically consists of 768 x 3200 (along-track by cross-track) 0.75-
km pixels
Example VIIRS granule, 11/02/2013, 19:05 UTCRGB image
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 6
Aerosol Retrieval – Physical Basis· The satellite-observed reflectance
(ρtoa) is the sum of atmospheric (ρatm) and surface components (ρsrf) .
· The components are the result of reflection, scattering by molecules and aerosols and absorption by aerosols and gases.
· The aerosol portion of the atmospheric component (aerosol reflectance) carries information about aerosol.
· The aerosol reflectance is determined by the amount and type (size, shape and chemical composition) of aerosol.
atm
srf
srfatmtoa
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 7
VIIRS Aerosol Algorithm (1)• Separate algorithms used over land and ocean• Algorithm heritages
– over land: MODIS atmospheric correction– over ocean: MODIS aerosol retrieval
• The VIIRS aerosol algorithm is similar but NOT identical to the above MODIS algorithms
• Many years of development work:– Initial science version is by Raytheon– Updates and modifications by NGAS– Current Cal/Val Team is to maintain, evaluate and improve the
algorithm
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 8
VIIRS Aerosol Algorithm (2)• AOT and aerosol model is
simultaneously retrieved using reflected solar radiation in multiple VIIRS bands and ancillary data.
• Optimal solution is searched for that best matches theoretical and observed reflectances.– iteration through increasing
values of AOT and candidate aerosol models
• Must account for all important radiative processes– molecular scattering, aerosol
scattering and absorption, gas absorption, and surface reflection.
• Approach in the vector RT Second Simulation of the Satellite Signal in the Solar Spectrum (6S-V1.1) [Kotchenova and Vermote, 2007] is adopted.
terms) Lambertian-(non
surfAR
surfvAARsAAR
OHOH
ROHOHAA
OogAtoa STT
UTg
PUTg
TgTg
1),(),(
)(
)()2()(
)(22
22
3
• ρR+A, TR+A, SR+A are pre-calculated by 6S and stored in LUT; Tgs are parameterized.
(1)
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 9
Ove
r Lan
d Re
trie
val
• Atmospheric correction of reflectances [Vermote and Kotchenova, 2008]– AOT and aerosol model are by-products of surface
reflectance retrieval• Basis: aerosols change the ratios of spectral
reflectances (spectral contrast) from those of the surface values.
• AOT and aerosol model are the ones that provide the best match between ratios of surface reflectances retrieved in multiple channels and their expected values.– Expected ratios are derived empirically by
atmospherically correcting VIIRS TOA M-band reflectances using AERONET AOT (99 sites, ~60,000 matchups).
• Eq. (1) is solved for ρsurf assuming Lambertian reflection.
• 5 aerosol models [Dubovik et al. 2002]:– dust – smoke (high and low absorption)– urban (clean & polluted)– bimodal lognormal size distribution, function of
AOT, spherical particles
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 10
Ove
r Lan
d Re
trie
val (
2)
0.5 1.0 1.5 2.0 2.50.0
0.1
0.2
0.3
0.4
0.5
0.6
Ref
lect
ance
Wavelength (m)
=0.0 =0.1 =0.2 =0.5 =1.0 =2.0
TOA reflectance vs. AOT ()
M1M2M3
M5
M11
0.0 0.5 1.0 1.5 2.00.2
0.4
0.6
0.8
1.0
1.2
1.4
surfa
ce+a
eros
ol re
flect
ance
ratio
AOT
M1/M5 M2/M5 M3/M5 M11/M5
0.5 1.0 1.5 2.0 2.50.0
0.1
0.2
0.3
0.4
0.5
0.6
TOA
refle
ctan
ce
Wavelength (m)
s+a (=0.4) s+a+m s+a+m+g
M1M2M3
M5
M11
• Surface + aerosol reflectance changes with increasing AOT.– Surface: green vegetation– Atmosphere:
• aerosol: urban clean• Rayleigh: no• Gas absorption: no
– SZA=0°, VZA=30°, RAZ=20°• Normalized spectral reflectance
(“spectral shape”) also changes with increasing AOT.
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 11
Over Land Retrieval (3)• AOT is retrieved by marching
through AOTs in LUT until the retrieved M3 surface reflectance is close to the one expected from the retrieved M5 value.
0.0 0.1 0.2 0.3 0.40.03
0.04
0.05
0.06
0.07
Sur
face
refle
ctan
ce
AOT @550 nm
M5 retrieved M3 retrieved M3 expected
0.038
0.36
0.5 1.0 1.5 2.0 2.5
0.0
0.1
0.2
0.3
0.4
M1M2
M3 M4
M5
M6
M7M8
M9
M10
M11
Obs
erve
d re
flect
ance
Wavelength (m)
Location: UMBC (76.71W, 39.26N); Date: 5/17/2013
Aerosol model: urban-polluted
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 12
0.0 0.1 0.2 0.3 0.4 0.5 0.6
-0.004
0.000
0.004
0.008
0.012
0.016
0.020
model 4
(ret
rieve
d-ex
pect
ed) M
3 su
rface
refle
ctan
ce
AOT @550 nm
model 0 model 1 model 2 model 3 model 4
0.5 1.0 1.5 2.0 2.50.02
0.04
0.06
0.08
0.10
0.12
M1M2M3
M5
M11
2
expecteds
retrievedsresidual
Sur
face
refle
ctan
ce,
Wavelength (m)
retrieved expected
model 4
M1
0.56
0.36
0.26
0.320.36
1.05
E-0
2
1.43
E-0
4
1.29
E-0
4
1.56
E-0
4
1.26
E-0
4
0 1 2 3 40.0
0.1
0.2
0.3
0.4
0.5
0.6
AOT @550 nm residual
model
AO
T @
550
nm
10-5
10-4
10-3
10-2
Res
idua
l
0.5 1.0 1.5 2.0 2.50.0
0.1
0.2
0.3
0.4
0.5
0.6 M1
M2
M3
M4
M5M6
M7
M8 M10 M11
Location: UMBC (76.71W, 39.26N); Date: 5/17/2013
AO
T @
550n
m
Wavelength (m)Ove
r Lan
d Re
trie
val (
4)
0=dust; 1=smoke-high abs.; 2=smoke-low abs.;3=urban-clean; 4=urban polluted
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 13
Ove
r Oce
an R
etrie
val
• Close adaptation of the MODIS approach [Tanré et al., 1997]– search for AOT and aerosol model that
most closely reproduces the VIIRS-measured TOA reflectance in multiple bands.
– wind-dependent (speed and direction) ocean surface reflectance is calculated analytically.
• Accounts for water-leaving radiance (Lambertian, fixed pigment concentration), whitecap (Lambertian, wind-speed dependent) and specular reflection (dependent on wind speed and direction).
– Combines 5 fine mode and 4 coarse mode models with 0.01 increments in fine mode fraction (2020 models)
– TOA reflectances in selected M bands are calculated from Eq. 1 and compared to observed ones to retrieve AOT and aerosol model simultaneously.
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 14
Over Ocean Retrieval• AOT for a given aerosol model is
from matching calculated and observed M7 TOA reflectances.
• Retrieved AOT is used to calculate TOA reflectances in other channels.
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
TOA
refle
ctan
ce
AOT@550nm
M7M7
calculated observed calculated @ retrieved AOT550
M5
M6
M8M10M11
0.5 1.0 1.5 2.0 2.5
0.00
0.05
0.10
0.15
0.20
0.25
0.30M1
M2
M3
M4
M5M6
M7M8
M9M10 M11
Obs
erve
d re
flect
ance
Wavelength (m)
Location: Arica (70.31W, 18.47S); Date: 5/4/2013
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 15Ove
r Oce
an R
etrie
val (
2)
0.5 1.0 1.5 2.0 2.5
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45M1
M2
M3
M4
M5M6
M7M8 M10 M11
Location: Arica (70.31W, 18.47S); Date: 5/4/2013
AO
T @
550n
m
Wavelength (m)
200 400 600 800 1000120014001600180020000
2x10-4
4x10-4
6x10-4
8x10-4
1x10-3
1x10-3
residual = 2.3x10-5
AOT @550nm = 0.259fine mode model = 0coarse mode model = 2fine mode fraction = 0.6
Res
idua
l
Model
0 20 40 60 80 1000
2x10-4
4x10-4
6x10-4
8x10-4
1x10-3
1x10-3
Res
idua
l
Fine mode fraction (x100)
fine mode model index = 0coarse mode model index = 2
residual = 2.3x10-5
AOT @550nm = 0.259fine mode fraction = 0.6
0.5 1.0 1.5 2.0 2.50.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
M5
M6
M7
M8
M10 M11
2
obsTOA
calTOAresidual
fine mode model index = 0coarse mode model index = 2fine mode fraction (x100) = 60
TOA
refle
ctan
ce
Wavelength (m)
calculated observed
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 16
Pixe
l Sel
ectio
n &
Qua
lity
Flag
s Condition Quality Flag Applies to Detected byNot
ProducedExcluded Degraded Land Ocean VCM Internal
TestsInvalid SDR data X X X X
Cloud Contamination X X X XSun Glint X X X X XSnow/Ice X X X X X
Fire X X X XBright Surface X X X
Coastal or Inland Water X XTurbid Water X X X
Ephemeral Water X X XSolZA ≥ 80° X X X X
Out of Spec Range X X X XCloud Adjacency X X X X
Cloud Shadow X X X XCirrus X X X X X
Soil Dominated X X X65° ≤ SolZA < 80° X X X X
Large Retrieval Residual X X X X
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 17
Planned Enhancements
• Replace current fixed surface reflectance relationships with NDVI-dependent relationships.
• Extend AOT reporting range from 2 to 5.• Update ocean aerosol models with those from MODIS
algorithm.• Improve cloud/heavy-aerosol discrimination, snow/ice
detection; add spatial variability internal test.• Add Deep-Blue module (Hsu & Sayer) to extend
retrievals over bright surfaces.
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 18
VIIR
S vs
. MO
DIS
VIIRS MODISAlgorithm (general)
Main source of data screening External VCM Internal testsAggregation on Outputs InputsResidual calculated as Absolute difference Relative difference
Over-ocean algorithm
Channel used 0.67, 0.74, 0.86, 1.24, 1.61, 2.25 µm
0.55, 0.66, 0.86, 1.24, 1.61, 2.12 µm
Surface reflection Non-lambertian, function of wind speed and direction
Lambertian, independent on wind (will change in C6)
Aerosol model Combination of fine and coarse modes
Combination of fine and coarse modes
Match to TOA reflectances TOA reflectancesRetrieval over inland water No Yes
Over-land algorithmChannel used 0.41, 0.44, 0.48, 0.67, 2.25 µm 0.47, 0.66, 2.12 µmAerosol model Select one from five pre-defined
modelsMix two assigned fine and coarse mode dominated models
Spectral surface reflectance Constant ratios of 0.41, 0.44, 0.48, 2.25 µm over 0.67 µm(will depend on NDVI)
Linear relationship between 0.66 and 2.12 µm as a function of NDVI and scattering angle; constant linear relationship between 0.47 and 0.66 µm
Match to Surface reflectances TOA reflectances
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 19
VIIR
S vs
. MO
DIS
(con
t)VIIRS MODIS
ProductsNominal spatial resolution 0.75 km (IP)
6 km (EDR)10 km (C5)3 km (C6)
Granule size 86 seconds 5 minutesAOT range [0, 2] (will change) [-0.05, 5]Main product land Spectral AOT Spectral AOTMain product ocean Spectral AOT
Ångström exponentSpectral AOTFine mode fraction
Orbit
Orbit altitude 824 km 690 km
Equator crossing time 13:30 UTC 13:30 UTC (Aqua)
Swath width 3000 km 2300 kmPixel resolution (nadir) 0.75 km 0.5 kmPixel resolution (swath edge) 1.5 km 2 km
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 20
Sensor and Other Inputs• VIIRS M-band SDRs (reflectances)
– M1-M12, M15, M16, and quality flags• Solar and view geometry
– zenith and azimuth angles• VIIRS Cloud Mask (VCM)
– pixel cloud flag (clear/cloudy, probably clear/cloudy), cloud shadow, land/water, snow/ice, fire, sunglint, heavy aerosol, volcanic ash
• NCEP GFS data (backup:FNMOC/NAVGEM)– Water vapor, ozone, surface pressure, winds (speed and direction)
• Navy Aerosol Analysis and Prediction System (NAAPS) aerosol data– used for filling in missing VIIRS IP retrievals
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 21
VIIRS Aerosol Products (1)• Aerosol Optical Thickness (AOT)
– for 11 wavelengths (10 M bands + 550 nm)• APSP (Aerosol Particle Size Parameter)
– Ångström Exponent derived from AOTs at M2 (445 nm) and M5 (672 nm) over land, and M7 (865 nm) and M10 (1610 nm) over ocean
– qualitative measure of particle size– over-land product is not recommended!
• Suspended Matter (SM)– classification of aerosol type (dust, smoke, sea salt, volcanic ash) and
smoke concentration– currently, derived from VIIRS Cloud Mask (volcanic ash) and aerosol model
identified by the aerosol algorithm • Only day time and over dark land and non-sunglint ocean
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 22
VIIR
S Ae
roso
l Pro
duct
s (2)
At NOAA Comprehensive Large Array-data Stewardship System (CLASS):
• Intermediate Product (IP)– 0.75-km pixel
• AOT• APSP• AMI (Aerosol Model Information)
– land: single aerosol model– ocean: indexes of fine and coarse modes and
fine mode fraction• quality flags
• Environmental Data Record (EDR)– 6 km aggregated from 8x8 IPs filtered by
quality flags• granule with 96 x 400 EDR cells• AOT• APSP• quality flags
– 0.75 km• SM
At NOAA/NESDIS/STAR– Gridded 550-nm AOT EDR
• regular equal angle grid: 0.25°x0.25° (~28x28 km)
• only high quality AOT EDR is used
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 23
Summary• Algorithm documents
– Jackson, J., H. Liu, I. Laszlo, S. Kondragunta, L. A. Remer, J. Huang, H-C. Huang, 2013: Suomi-NPP VIIRS Aerosol Algorithms and Data Products, J. Geophys. Res. doi: 10.1002/2013JD020449
– ATBD (Draft)– OAD
• Product documents– User’s Guide– Readme files
• VIIRS Aerosol Calibration and Validation website http://www.star.nesdis.noaa.gov/smcd/emb/viirs_aerosol/index.php
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 24
Summary (cont.)• Product quality:
– Liu, H., L. A. Remer, J. Huang, H-C. Huang, S. Kondragunta, I. Laszlo, M. Oo, J. M. Jackson, 2013: Preliminary Evaluation of Suomi-NPP VIIRS Aerosol Optical Thickness, J. Geophys. Res. (in review)
– Hongqing Liu: VIIRS Aerosol Products, Data Quality, and Visualization Tools (VIIRS Aerosol Science and Operational Users Workshop , November 21-22, 2013, College Park, MD)
• Link to CLASS:– http://www.class.ncdc.noaa.gov/saa/products/welcome
• Link to gridded data:– http://www.star.nesdis.noaa.gov/smcd/emb/viirs_aerosol/produc
ts_gridded.php
VIIRS Aerosol Science and Operational Users Workshop, Nov 21-22, 2013, NCWCP, College Park, MD 25
Backup
26
AOT Product Timeline
Initial instrument check out;
Tuning cloud mask parameters
Beta status Error Beta status
Provisional status
26
28 Oct 2011 2 May 2012 15 Oct 201228 Nov 2012
23 Jan 2013
Red period: Product is not available to public, or product should not be used.
Blue period: (Beta)
Product is available to public, but it should be used with caution, known problems, frequent changes.
Green period:(Provisional)
Product is available to public; users are encouraged to evaluate.
Products go trough various levels of maturity: