Using AIRS to Assess MODIS Radiances
Dave Tobin
CIMSS / SSEC / UW-Madison
NOAA Cooperative Research Program
Second Annual Science Symposium
13 July 2005
Introduction
• Overall Goal:
– Create a long term record of well calibrated and characterized
satellite radiances
– Emphasis on high spectral resolution (AIRS, TES, IASI, CrIS, …)
but also including broadband sensors (MODIS, VIIRS, GOES,
…)
• Post-launch AIRS Radiance Cal/Val activities at
CIMSS/SSEC
–
–
–
–
Noise characterization
Early radiance comparisons with GEOs
Spatial co-registration
Spectral Radiance validation with aircraft underflights with
Scanning-HIS and NAST-I
 AIRS/MODIS comparisons
– Obs-Calc Analyses
• AIRS / MODIS Comparisons
– Review of Scanning-HIS validation of AIRS
– AIRS / MODIS Comparison Approach
• Match spectral resolutions
• Match spatial resolution and sampling and select uniform fields
of view
– Differences characterized as a function of scene temperature,
scan angle, and solar zenith angle for global data collected on 6
Sept 2002 and 18 Feb 2004
– Tobin, D. C., H. E. Revercomb, C. C. Moeller, and T. S. Pagano,
Use of AIRS high spectral resolution infrared spectra to assess
the calibration of MODIS on EOS Aqua, J. Geophys. Res.,
submitted, April 2005.
• Important for:
– Diagnosing the calibration of both sensors
– Understanding differences between AIRS products and MODIS
products
– Development of applications utilizing data from both sensors
(e.g. AIRS cloud-clearing using MODIS, synergistic use
of AIRS and MODIS for cloud property retrievals)
AIRS underflight by the
Scanning-HIS,
21 November 2002
Gulf of Mexico
Daytime
AIRS
S-HIS
AIRS / S-HIS comparison, without
accounting for viewing geometry or
spectral resolution/sampling
differences:
Tobin, D. C., H. E. Revercomb, R. O. Knuteson, F. A. Best, W. L. Smith, P. van Deslt, D. D. LaPorte, S. D. Ellington, M. W. Werner, R. G. Dedecker, R.
K. Garcia, N. N. Ciganovich, H. B. Howell, S. B. Dutcher, J. K. Taylor, K. Vinson, T. S. Pagano, S. A. Mango, Radiometric and Spectral Validation of
AIRS Observations with the Aircraft based Scanning High resolution Interferometer Sounder, J. Geophys. Res., submitted, April 2005.
A sample AIRS
brightness
temperature
spectrum overlaid
with the Aqua
MODIS Spectral
Response
Functions
36 35 34
30
25
33
29
24
32
28
23
wavenumber
31
27
22,21
20
To match the MODIS spectral resolution, the AIRS spectra are convolved with the
MODIS SRFs
RMONO  SRFMODIS – (RMONO  SRFAIRS)  SRFMODIS
Convolution Correction: factor that accounts for small gaps in AIRS spectra
when convolving AIRS radiance spectra with the MODIS SRFs.
Corrections for standard atmospheres
4.0
4.0
4.1
4.4
4.5
6.8
7.3
8.5
9.7
11.0
12.0
13.4
13.7
13.9
14.2
m
for 6 September 2002 AIRS data
In the following comparisons, the correction is represented as the mean of the six
standard atmosphere values shown above for each band. This treatment is more
accurate for bands for which the correction is small and for which the correction does not
vary largely with the profile/spectrum.
The 1 km MODIS data is
collocated with AIRS by
representing the AIRS
FOVs as slightly oversized
circular footprints, and
computing the mean
MODIS value within those
footprints for each band.
Spatially uniform scenes
are selected by requiring
the standard deviation of
the MODIS data within
each AIRS footprint to be
0.2K or less.
Example comparisons for band 22
(4.0 m) on 6 Sept 2002.
mean= -0.05 K
Little Dependence on
Scene Temperature
Little Dependence on
X-track View Angle
Little Dependence on
Solar Zenith Angle
AIRS BT (K)
AIRS minus MODIS (K)
Example comparisons for band 34
(13.7 m) on 6 Sept 2002.
AIRS
MODIS
Histograms of
brightness
temperature
differences.
6 September 2002
18 February 2004
m
14.2
13.9
13.7
(Light gray curves
are distributions
without the
convolution
corrections)
13.4
12.0
11.0
9.7
7.3
6.8
4.5
4.4
4.1
4.0
4.0
Brightness
temperature
differences as a
function of scene
temperature.
6 September 2002
18 February 2004
m
14.2
1K
13.9
13.7
13.4
12.0
11.0
9.7
7.3
6.8
4.5
4.4
4.1
4.0
4.0
Band 35 (13.9 m)
brightness temperature
differences for one orbit
of data on 6 Sept 2002
using (1) the nominal
MODIS SRF and (2) the
MODIS SRF shifted by
+0.8 cm-1.
MODIS SRF out-of-band
response also currently
being investigated.
unshifted
unshifted
shifted
shifted
unshifted
shifted
Brightness
temperature
differences as a
function of scan
angle.
6 September 2002
18 February 2004
m
14.2
1K
13.9
13.7
13.4
12.0
11.0
9.7
7.3
6.8
4.5
4.4
4.1
4.0
4.0
Band 24 (4.4 m), descending
Scan Angle Asymmetry
using non-polar clear sky swaths
6 September 2002
AIRS
MODIS
Band:
24
33
(Low Yield)
34
35
36
The Longwave CO2 band biases are making a large (positive)
impact on MODIS CO2 slicing algorithm performance
Channel Pair Representing the Best CTP Retrieval
Pink, cyan, and green are
36/35, 35/34, 34/33, respectively
c/o Rich Frey
Zonal Plus Constant Radiometric Bias
Zonal Plus Variable Radiometric Bias
Comparisons with CERES Window Channel Radiances:
Monochromatic
AIRS
CERES FM4 SRF
Convolution Errors (K):
-0.44 Tropical
-0.52 MLS
-1.07 MLW
-0.80 SAS
-1.77 SAW
-0.70 US Std.
AIRS / CERES-FM4 Window
Channel Comparisons
18 Feb 2004, Nighttime, Ocean only, 60S to 60 N
Summary
•
Comparison of EOS Aqua AIRS and MODIS infrared radiances for spatially
uniform scenes collected on 6 September 2002 and 18 February 2004 have
been presented.
•
A simple approach to account for spectral gaps in the AIRS spectra when
convolving with the MODIS SRFs has been introduced.
•
Estimates of the absolute uncertainty of the comparisons are 0.1 K or less for
the majority of the MODIS bands.
•
Mean differences between AIRS and MODIS are ~1 K or less for all bands and
many bands show agreement of 0.1 K or better. But at the same time, only band
22 (3.9 m) shows good absolute agreement and no significant dependence on
scene temperature, scan angle, or solar zenith angle.
•
Differences for MODIS bands 27 (6.8 m), 28 (7.3 m), 34 (13.7 m), 35 (13.9
m), and 36 (14.2 m) display clear and significant dependencies on scene
temperature.
•
Results for the two days are very similar with changes in mean differences of 0.1
K or less for most bands.
•
Preliminary AIRS/CERES comparisons look good; more accurate comparisons
require a more sophisticated approach to account for the AIRS spectral gaps.