Development of the secondgeneration operational algorithm

1
Development of the second-generation
operational algorithm
2
The Satellite Signal is complicated
Assumptions Aerosol physics scattering/extinction
properties Radiative Transfer (including
Rayleigh and wavelength) Surface
Properties Retrieval Technique
(not to mention cloud effects, calibration, etc)
3
Aerosol physics(e.g. assuming aerosol
scattering/extinction properties)
4
Aerosol optical models
Revised size distribution for CLAMS (Levy et al.,
2005)
Revised SSA for SAFARI (Ichoku et al., 2002)
New Aerosol models improve retrievals for CLAMS
and SAFARI Suggests need to re-evaluate models,
globally
5
Global aerosol optical models An AERONET view
(since 1994)
JJA
Non-absorbing
Moderate absorbing
Absorbing
?0 0.91,0.85,0.95
?0 0.85, 0.91, 0.95 (Fine) ?0 0.95
(Coarse/Spheroid/Dust)
Cluster by ?0 and g, seasonal, ---gt 3
fine-dominated models
6
Radiative Transfer and lookup tables (and
Rayleigh and wavelengths)

• Re-calculated central MODIS wavelength for each
  channel. (Combination of MODIS response function
  and spectral dependence of F0). Corrected assumed
  sea-level spectral Rayleigh optical depth.
• Calculated TOA reflectance and atmospheric
  transmission functions for lookup tables, for a
  variety of aerosol loadings and geometry
• Radiative Transfer code includes atmospheric
  polarization effect on TOA reflectance (up to 10
  error if neglected Levy et al., 2004)

7
Surface Reflectance Properties
8
Surface reflectance ratios (CLAMS-2001)
Atmospheric Correctionsurface MODIS - AERONET
C004
CLAMS-derived
?s0.47 / ?s2.1 0.25 ?s0.66 / ?s2.1 0.50
?s0.47 / ?s2.1 0.37 ?s0.66 / ?s2.1 0.63
? 0.08
Reduces ocean/land discontinuity
But over-corrects here
Constant ratios do not work everywhere!! (Levy et
al., 2005)
9
Surface reflectance global
C004 (Global fixed ratio of VIS to 2.1 ?m) ?s0.66
0.5(?s2.1) ?s0.47 0.25(?s2.1)
y 0.011 0.259x, R 0.75 y 0.001 0.551x,
R 0.93
y -0.008 0.508x, R 0.87
VIS vs IR
Blue vs Red
?550 lt 0.2 N 4200
C005 (parameterization of 2.1 ?m, angle, and
surface) ?s0.66 f0.66(?s2.1, ???NDVISWIR) ?s0.47
g0.47(?s0.66, ???NDVISWIR)
10
Inversion of Spectral Reflectance
11
Retrieval technique

• OLD (c004)
• No aerosol at 2.1 ?m
• 2.1 for surface only
• Independent retrievals at 0.47 and 0.66 ?m
• Interpolated to 0.55 ?m

Spectral reflectance of different aerosol

• NEW (c005)
• Aerosol information at 2.1 ?m
• Inversion of 3?channels yields ???? (fine aerosol
  weighting) and ?s (surface reflectance)
• Retrieved at 0.55 ?m

12
Retrieval The Inversion
Calculate amount ??and weighting ??of
fine-dominated model (to total ?), combined with
the surface reflectance ?s, that best matches the
observed spectral reflectance (0.47, 0.66 and
2.1?m)
0.55 ?m
coarse
combo
fine
reflectance
surface
wavelength
Stolen from L. Remer
Fine
Coarse
Radius (µm)
13
Summary of c005 improvements

• Recomputed center wavelengths and Rayleigh OD
• Lookup Table includes atmospheric polarization
  for intensity
• Aerosol models based on AERONET Dubovik
  retrievals and new geographic distribution
• Surface reflectance relationships are functions
  of vegetation index and scattering angle
• New inversion algorithm implemented
• Valid range of ? extended to -0.05
• Subpixel snow mask implemented
• Cloud mask adjusted
• Column mass concentration product corrected
  units
• New products and new Quality Assurance logic

New physics, Algorithmic details, other peoples
work
14
Results
15
Sanity Check Images ? over US East Coast
Anthony Wimmers, CIMMS
August 4, 2006
Cold front
C004
C005
0.0 0.2 0.4 0.6
0.8 1.0
C005 more continuity from land to ocean
16
Validation Out-sourcing to India
Kanpur is an industrial city located in the
Indo-Gangetic Basin, surrounded by tall Himalaya
in the north and mountains of height about
500-1000m in the south
Aqua 2002-2005
C004
C005
R0.82 68
R0.64 59
Hiren Jethva PhD student at Indian Institute for
Science