OMI BSDF Validation Using Antarctic and
Greenland Ice
Glen Jaross and Jeremy Warner
Science Systems and Applications, Inc.
Lanham, Maryland, USA
Outline
• Justification for using ice surfaces
• The technique, including necessary external information
• Error budget – where do we focus attention?
• Results for OMI, TOMS, MODIS, and SCIAMACHY
Time Dependence of radiometric calibration
TOMS Nimbus 7
380 nm
Greenland
Antarctica
Seasonal Cycle:
• Neglecting terrain height variations
• Surface reflectance non-uniformity
TOMS Earth Probe
360 nm
OMI (Aura)
360 nm
Validation of Absolute Radiometry
1.
Develop a 2 steradian directional reflectance (BRDF) model for the
Antarctic surface; independent of wavelength.
2.
Combine BRDF with surface measurements of total hemispheric
reflectance measurements; wavelength-dependent
3.
Create a look-up table of sun-normalized Top-of-the-Atmosphere (TOA)
radiances for all satellite observing conditions using a radiative transfer
model
4.
Process sensor sun-normalized radiance data from a region of Antarctica
chosen for uniformity and low surface slope
5.
Compute ratio between each measurement and table entries; average
results
Surface properties based upon reflectance
measurements by Warren et al.
Spectral Albedo
Measuremnts at South
Pole, 1986
 = 600 nm
Sol. ZA = 80
Warren et al. Reflectance
anisotropy derived from
1986-1992 data
BRDF probably the
same:
300-800 nm
BRDF derived from
parameterization of measured
reflectance anisotropy
Error Budget
Source
100  Rel. std. uncertainty
Statistical
Systematic
Ground measurements
0.3
1
Satellite measurements
< 0.01

Atmospheric effects
0.3
0.2
Surface BRDF
0.5
2
Total
0.7
2.2
Surface BRDF model represents
single largest error source
Surface BRDF model vs. Solar Zenith Angle
SolZA=40
SolZA=60
SolZA=50
SolZA=85
BRDF is most important at longer wavelengths
BRDF plays bigger
role as
Simulated Nadir-scene albedos
diffuse / direct ratio
decreases
Solar Zenith Angle = 75
Column Ozone = 325 DU
Lambertian
Non-Lambertian
Non-Lambertian /
Lambertian Radiance Ratio
OMI Results
OMI L1b Data:
7 Dec – 4 Jan, 2004
• Perfect model would yield flat
SolZA dependence
• Perfect calibration would yield
values = 1 at all wavelengths
 Plot suggests probable radiative
transfer errors
– surface BRDF model
– treatment of atmosphere
 We believe that results obtained
below SolZA = 70 fall within our
2.2% uncertainty estimate
OMI Full spectral range ice radiance results
83 < SolZA < 86
62 < SolZA < 68
Flat spectral
result gives us
confidence that
result is
resonable
Spectral
dependence is
not realistic –
consistent with
BRDF error
Apparent error
increases at
long  as
predicted
Shadowing Errors
From
Radarsat-1
Large scale structures (snow dunes)
not captured by ground
characterizations
Simple linear shadow model for testing errors
Tune barrier height and
separation to yield flattest
SolZA dependence in data
Shadow study using MODIS / Aqua
Comparison to RTM, without correction
Comparison to RTM, with shadow correction
Consistent with ~2%
uncertainty estimate
Comparison between MODIS, OMI, TOMS and
model radiances
MODIS / Aqua
TOMS / EP
RTM handles
ozone poorly at
 < 330 nm
OMI / Aura
O2O2
Absorption
RTM does not
include Ring
Effect or O2-O2
abs.
X-track and wavelength error surfaces
UV2
VIS
Noise in UV2 masks X-track dependence
X-track errors at selected wavelengths
TOMS / Earth Probe
OMI / Aura
MODIS / Aqua
Shared wavelengths do not
behave the same for different
sensors
Comparison of OMI to Predicted Albedos (470 nm)
Longer
wavelengths 
larger errors due to
BRDF
Why should we
trust the solid
curve ?
X-track dependence also supports smaller solar zenith angles
Evaluation of Aerosol Index wavelengths
OMTO3 method
utilizes land
surface
reflectances;
cannot evaluate
absolute
Methods agree
that there is
little wavelength
dependence in
the BSDF error
Comparison of ice and land X-track dependence (360 nm)
Ice minus OMTO3
OMTO3
Ice
OPFv8 IRR error
Solar Irradiance for Ice
taken near reference
azimuth
dIrrad error in OMTO3
results can account for
much of the difference
UV2 – VIS overlap region looks good
Common
structure
suggests it’s
geophysical
Noise in
longwave UV2
probably from
irradiance
UV2 – VIS overlap has slight X-track dependence
Structure most
likely from UV
irradiance
Summary
Model calculations of TOA radiances over Antarctica are good to
approximately 2% at low solar zenith angles (i.e. near Dec. 21)

Radiometric characteristics of nadir-viewing sensors can be validated from
~330 nm to ~750 nm

Wavelength-to-wavelength radiometry is better than 2%, but not useful for
absorption spectroscopy


We derive the following OMI Day 1 BSDF errors
Nadir: -2.5% (330 <  < 500 nm)
Positions 1 & 60 : -4.0% (approximately)
Future Work :

Re-derive ice and land X-track errors using GDPS 9.15

Confirm shadowing error
Spares
Preliminary SCIAMACHY Results
SCIAMACHY Level 1b ( v5.04 )
18 – 24 Dec., 2004
Provided by R. van Hees, SRON
}
Ozone
Absorption
ignored
Comparison with
RTM over Sahara
(from G. Tilstra,
KNMI)
OMI radiances compared directly to MODIS / Aqua band 3
MODIS
OMI
Same time and
geographic
location
MODIS has broad
bandwidth
(459 <  < 479 nm)
which includes O2O2 absorption