Global Remote Sensing of Aerosol using MODIS: Algorithm and Validation

1
Global Remote Sensing of Aerosol using MODIS
Quantifying the Ambient (Undisturbed) Conditions
Yoram Kaufman, Lorraine Remer, Shana Mattoo,
Robert Levy, Rich Kleidman the MODIS Aerosol
Team at NASA / GSFC Brent Holben the AERONET
Team at NASA / GSFC
2
Spectral optical properties of aerosol
Dust
Smoke
Y. Kaufman
Big particles (e.g. Dust) reflect in IR Small
particles (smoke/pollution) do not.
3
Spectral optical properties of aerosol
Smoke / Pollution
Dust / Sea Salt
wavelength
4
Size Distribution
coarse mode
Approximately lognormal shape
dV / dlnr
fine mode
Radius , r, (?m)
5
Aerosol size and optical properties
These Aerosols have the largest impacts on
radiative budget, visibility, cloud processes and
health issues
0.001 0.01 0.1
1.0 10.0
UV BGR NIR IR
Radius, ?m
6
? Aerosol Optical Depth (extinction)
F0 F0,???0?
?
Beer-Bouguer-Lamberts extinction law L? F0,?
exp-??m where ? is wavelength, mm(?0) is
air mass (function of solar zenith angle, ?0),
LL(?0) is radiance, F0F0(?0) is solar
irradiance and ??is total optical depth.
7
Passive Remote Sensing (Scattering)
Satellite (e.g. MODIS)

• If we can measure spectral light extinction or
  scattering
• And we can separate the aerosol signal (from the
  total that includes, molecular, surface, etc)
• Then, we can retrieve aerosol optical properties
  (find a solution that fits
• the measurements)
• Which then can be related to aerosol physical
  properties (type and loading)
• We can do this from the ground (sunphotometer) or
  from space (satellite)

?
Sunphotometer
8
Aerosol Passive Remote Sensing
Satellite (e.g. MODIS)
Unit of measure Aerosol Optical Depth
AOD or ? ??depends on wavelength
(?) ??represents total column light extinction
?
Sunphotometer
9
MODIS Aerosol Important Retrieval Bands
7 bands for aerosol retrieval All in atmospheric
windows
10
MODIS observations

• MODIS observes reflectance in 7 bands at 500 m
  (or 250 m)
• These data are noisy
• These data must be
• Cloud cleared
• Decided whether over ocean or over land
• Screened for glint and other surfaces that are
  too bright
• Massaged, averaged to yield
• One set of spectral reflectance that best
  represents average conditions in 10 km x 10 km
  box
• The MODIS dark target algorithm is actually two
  separate algorithms, one over ocean and one over
  land
• More about separate algorithms later

11
Deconstruction of Clear sky Signal
Sun
MODIS
Aerosol, Molecular, Gases
Dark Surface Target (Water or land)
12
Atmospheric Path Reflectance
Gas Aerosol scattering (path radiance)
Path Radiance
13
Transmission of Surface reflectance
T???Transmission s? Atmos. backscattering ?s?
Surface Reflectance
Direct Transmission
14
Multiple Reflection
Multiple Reflection
15
Adjacency Effect, etc
Indirect Transmission (adjacency effect)
16
Complicated Clear sky TOA Signal

• Contributions from
• Gas absorption (O3, CO2, etc)
• H2O absorption
• Rayleigh (molecular) scattering
• Aerosol scattering and absorption
• Surface reflection
• Atmosphere / Surface interaction
• Contamination from neighboring pixels (clouds,
  etc)

Gas Aerosol scattering (path radiance)
Direct Transmission
Indirect Transmission (adjacency effect)
Multiple Reflection
17
What is a dark target algorithmand why do we use
it?

• We have a mixed signal of aerosol and surface
• If we remove the surface signal (our noise) we
  are left with aerosol (our signal)
• Choosing dark targets helps to minimize our error
  from the surface signal

18
Over Ocean
Water Leaving Reflectance
0.10 0.08 0.06 0.04 0.02 0.00

• We can choose the wavelengths that give little or
• no surface signal.
• The 0.55 channel is assumed to have a water
  leaving radiance of 0.005
• Longer wavelengths have
• no signal at all.
• We account for the effects
• In the 0.55 channel in our
• Radiative transfer calculations

Chlorophyll Concentration
20.0 10.0 2.0 1.0 0.2 0.1 0.02
Reflectance

• 466 553 644
  
• Wavelength

---gt DARK ? 553nm
19
Aerosol models over oceanAre represented by
theoretical modes
ATBD
20
MODIS Ocean Aerosol Retrieval
Solution
21
aerosol retrieval combined land/ocean

• Retrieved
• ? (0.55 ?m)
• ?? (0.55 ?m)
• ? (fitting error)
• ?s (over land)
• Derived
• ???
• ? Column Mass
• Diagnostics

Glint Mask
22
How is the land retrieval different than ocean?

• Scenes are screened to remove
• Water
• Snow, Ice, Melting snow

• Aerosol models are fixed and determined only by
• Location
• Season

Only 3 wavelengths are used in the
retrieval 0.44, 0.66, and 2.1
23
We Require Dark targets over land
LAND
Typical Vegetation Reflectance
MODIS Bands
Vegetation reflectance ?s????
0.55?s??? ?s???? 0.50?s???? Depends on
?????????????, ???? and ?
---gt DARK ? 644 nm
24
A Story of Two Land Pixels
Surface Error 10
True Reflectance
Measured Reflectance
Error in AOT
Pixel 1 0.2
0.22 .02 0.2
Dark Target Pixel 2 0.02
0.022 .002 0.02
25
AERONET based Aerosol models (land)
Each model has two or more modes. Size of
modes depend on ?
26
MODIS Land Aerosol Retrieval
Note multi-modal model