May 2003
SSEC Retrieval Workshop
Overview of Atmospheric Radiative Transfer
Sergio De Souza-Machado,
L. Larrabee Strow,
Scott Hannon, Howard Motteler
YLAND BA
L
U M B C
MO
UN
RE COUNT
Y
IVERSITY O
F
AR
TI
M
Workshop for Soundings from
High Spectral Resolution observations
SSEC - UWisconsin
May 6-8, 2003
1966
SSM, UMBC
1
May 2003
SSEC Retrieval Workshop
Overview
• Introduction
• Line by line Infrared Spectroscopy
• Radiative transfer : Clear, cloudy, NLTE
• Fast model (spectroscopy + radiative transfer)
• Validating the Fast model (clear sky)
• Sources of unknown bias?
• Cirrus observations
• Duststorm observations
• Volcanic eruption observations
• Conclusions
SSM, UMBC
2
May 2003
SSEC Retrieval Workshop
Introduction
• New generation instruments for remote sensing are high resolution, low
noise
• Require accurate forward models for radiative transfer
• Both spectroscopy and radiative transfer algorithm must be accurate
• Complications in IR spectroscopy include water vapor, CO2 lineshapes
deviate from Lorentz in the wings, where temperature/amount retrievals
require most accuracy!
• Complications in radiative transfer include scattering effects due to
clouds and aerosol, surface emissivity, reflectivity uncertainties
• Instruments download tons of data daily, so need accurate, fast models
SSM, UMBC
3
May 2003
SSEC Retrieval Workshop
Radiance vs Wavenumber : Important Gases, Window Regions
300
↔
↔
WINDOW
290
280
CFCs
WINDOW
SO2
BT(K)
270
260
↔
250
CO
O3
240
230
220
210
500
SSM, UMBC
↔
CO2
1000
↔
H2O
1500
2000
Wavenumber cm−1
↔
CO2
2500
3000
4
May 2003
SSEC Retrieval Workshop
Absorption in atmosphere : Infrared Molecular Spectroscopy
• Custom “line-by-line” code UMBC-LBL
• Line parameters obtained from standard databases, such as HITRAN
• For most gas molecules, use voigt lineshape. Quite fast (unless you have
hundreds of lines, such as HNO3)
• CO2 : special case, needs linemixing lineshape
• H2O : special case, needs continuum lineshape
SSM, UMBC
5
May 2003
SSEC Retrieval Workshop
Absorption in atmosphere : Scatterers
• Atmosphere has clouds, aerosols etc
• These particles can strongly absorb and/or scatter radiation
• Modelling their scattering parameters can be straightforward (assume
spherical particles, simple size distribution) or painful (irregular
particles, multimode size distributions etc)
• Need accurate knowledge of refractive index
• Some literature, or OPAC database, has low resolution info (5 cm-1)
• In the 10-12 um window region the absorption of a cloud strongly affects
the TOA radiance. This gives valueable clues to type of particles in cloud,
as absorption depends on refractive index
• In the 3.8 um window region scattering of cirrus particles can give clues
about crystal habit
• Thin cirrus/ aerosol events seem to make it thru the clear sky filter and
could impact the retrieval algorithm
SSM, UMBC
6
May 2003
SSEC Retrieval Workshop
Fast Models
• “line-by-line” codes are accurate, but SLOW
• ‘kCARTA uses compressed look up database of optical depths to
compute ’monochromatic LBL” radiances
• SARTA is a Fast Model based on kCARTA
• Fast Model 0.5 secs/profile; kCARTA 10 mins/profile (1 GHz machine)
SSM, UMBC
7
May 2003
SSEC Retrieval Workshop
Histogram of AIRS-RTA fitting errors.
1000
# of Channels
800
600
400
200
0
SSM, UMBC
0.02
0.04
0.06
0.08
0.1
RMS Fitting Error in K
0.12
0.14
0.16
8
May 2003
SSEC Retrieval Workshop
Radiative Transfer
• At steady state, the 1D Schwartzchild Equation says
µ
dI(ν, θ)
= −I(ν, θ) + J(ν)
ke dz
• µ = cos(θ), dz is the vertical coordinate
• ke is the total extinction (due to gases, clouds etc)
• ke dz = dτ is the optical depth
• I(ν, θ) is the radiance intensity
• J is the source function
• Clear Sky, in LTE : J = B(ν, T ), Easy!
0
• Clear Sky, in NLTE : ke = r1 ke ; J = r2 B(ν, T )
• Cloudy Sky, in LTE : integro-differential equation, Hard!
SSM, UMBC
9
May 2003
SSEC Retrieval Workshop
Effect on Non-LTE on AIRS instrument
B(T) in K
280
260
240
O−C in K
10
5
0
2220
SSM, UMBC
2240
2260
2280 2300 2320 2340
Wavenumber (cm−1)
2360
2380
2400
10
May 2003
SSEC Retrieval Workshop
Solution of Radiative transfer Equation III : Cloudy Sky
• For Cloudy Sky, solution is much more complicated!
• Solution by specialised codes, such as DISORT,RTSPEC,kTWOSTREAM
• Need to worry about scattering parameters : particle size distribution,
particle shape, cloud vertical extent
• Depending on complexity of solution, code can be quite slow!
• DISORT well tested, solar beam scattering, multiple streams, slow!
• RTSPEC well tested, two streams, no solar beam, fast!
• kTWOSTREAM tested against above codes, two streams, solar beam
scattering, quite fast!
• We wrote kTWOSTREAM so we included it in Fast Model!
• Fast Model 0.75 secs/profile; LBL 15 mins/profile (1 GHz machine)
• Plan to validate kT woStr eam with LIDAR Data from Chesapeake
Lighthouse (Dr. Ray Hoff) – ice crystal habits (Dr. Anthony Baran will
provide scattering parameters for eg aggregates )
SSM, UMBC
11
May 2003
SSEC Retrieval Workshop
Clear sky : Requirements for small biases
Bias = (mean) obs - (mean) calcs
• Spectroscopy for AIRS-RTA is correct
kCARTA has P/R linemixing for CO2, updated H2O continuum
• SRFs used to simulate the AIRS-RTA radiances are accurate
SRFs well characterized post-launch
• Profile data fields used for simulations must be statistically accurate
ECMWF data fields are best global NWP fields
We are also starting to use ARM-CART radiosondes
• AIRS radiances are radiometrically correct
Analysis of sea surface radiances by Aumann et al indicate radiometric
accuracy of at least 0.5K
SSM, UMBC
12
May 2003
SSEC Retrieval Workshop
Clear Sky : Comparing AIRS observations to SARTA
simulations
• AIRS-RTA developed for Clear Sky
• 1) Select nitetime views over ocean (sea surface emissivity well
characterised)
• 2) Restrict latitudes to ± 60 degrees, use all scan angles
• 3) find FOVS where adjacent BTs are within 0.25 K of center FOV.
Test done in window channels of 10, 3.8 micron region.
Test channels used = [900,961],[2611,2616] cm−1
Throws partly coudy, highly variable water vapor scenes away
• 4) SST from channels in 10, 3.8 microns identical withn about 0.4 K
Test lessens dependancy on AIRS radiometric calibration
• 5) ECMWF model temperatures, observed SST agree within 4K.
Test discriminates against low cloud decks
• These tests “throw” away lots of data, yet we typically analyse about 400
FOVs per granule!!
SSM, UMBC
13
May 2003
SSEC Retrieval Workshop
Clear Test Channels
290
280
B(T) in K
270
260
250
240
230
220
800
SSM, UMBC
1000 1200 1400 1600 1800 2000 2200 2400 2600
−1
Wavenumber (cm )
14
May 2003
SSEC Retrieval Workshop
AIRS brightness temperature biases for clear, ocean, night
scenes for October 2002 with ECMWF model fields
B(T) in K
280
260
240
220
O − C in K
Absolute Bias
Secant Bias
1
0
−1
1000
SSM, UMBC
1500
2000
−1
Wavenumber (cm )
2500
15
May 2003
SSEC Retrieval Workshop
Mean monthly bias between AIRS and ECMWF computed
brightness temperatures as a function of SST.
B(T) in K
280
260
240
220
O − C in K
0.5
280K
290K
300K
305K
0
−0.5
−1
−1.5
800
SSM, UMBC
1000 1200 1400 1600 1800 2000 2200 2400 2600
Wavenumber (cm−1)
16
May 2003
SSEC Retrieval Workshop
Scatter plot of bias at 2616 cm-1
SSM, UMBC
17
May 2003
SSEC Retrieval Workshop
Oct 2002 bias vs Validated bias
Top plot : Mean spectra
Bot plot : Validated data, adjusting continuum coeffs, fixed gas predictors
300
B(T) in K
280
260
240
220
200
Orig RTA
RTA mod by ARM
O − C in K
1
0
−1
700
SSM, UMBC
800
900
1000 1100 1200 1300 1400 1500 1600
Wavenumber (cm−1)
18
May 2003
SSEC Retrieval Workshop
Scattering particles in Atmosphere : Refractive Index
1.2
Water
Ice
Volcano
Sea Salt
Desert Dust
1
Imag(n)
0.8
0.6
0.4
0.2
0
500
SSM, UMBC
1000
1500
2000
Wavenumber (cm−1)
2500
3000
19
May 2003
SSEC Retrieval Workshop
Detecting Cirrus Clouds with AIRS
• Refractive index of ice means that typically, BT(980) ≥ BT(820)
• Crystal habit information can be recovered from analysing the 10-12, and
8 um windows and 3.8 um windows as well
• Have detected and analysed cirrus clouds using scattering parameters for
spheres and aggregates (Dr. Anthony Baran)
SSM, UMBC
20
May 2003
SSEC Retrieval Workshop
Detecting Cirrus Clouds with AIRS (contd)
• Currently assume cf r ac = 1, do CO2 - slicing for cloud top pressure (C.
Barnet will do this work for us!!!)
• Assume only ONE cloud in FOV
• Use ECMWF fields for water vapor, temperature, ozone, surface params
• Use about 200 channels, spaced in the 10-12, 8 and 3.8 um windows
• Do a least squares fit using SART A CLOUDY
• Ice aggregates (Dr. Anthony Baran) fits AIRS spectra in 10-12 um and 3.8
um windows, better than ice spheres.
• Currently working on a nonlinear retrieval algorithm (seems to work for
large cirrus signatures, problems with small signatures)
SSM, UMBC
21
May 2003
SSEC Retrieval Workshop
Average monthly difference between the 961 and 820 cm−1
channel biases
SSM, UMBC
22
May 2003
SSEC Retrieval Workshop
Water vapor under control : Mean monthly bias between AIRS
observed brightness temperatures and those computed from
ECMWF model fields for the 2616 cm−1 channel.
SSM, UMBC
23
May 2003
SSEC Retrieval Workshop
Cirrus Spectral Signature
granule = 1 day = −1 iwp,dme = 7.5 15
obs
clear
cloudy
260
B(T) in K
250
240
230
220
210
800
SSM, UMBC
1000 1200 1400 1600 1800 2000 2200 2400 2600
−1
Wavenumber (cm )
24
May 2003
SSEC Retrieval Workshop
Mean Cloud forcing (daytime) : Aggregates versus Spheres
10
5
0
−5
−10
−15
−20
−25
Clear
Spheres
Aggregates
−30
−35
500
SSM, UMBC
1000
1500
2000
2500
3000
25
May 2003
SSEC Retrieval Workshop
Distribution of IWP, DME : Aggregates versus Spheres
20
Eff Diam (um)
sph
agg
10
0
0
20
40
60
80
sph
agg
50
Ice Path (g/m2)
0
0
30
20
40
60
80
sph
agg
20
Optical Depth
10
0
0
SSM, UMBC
1
2
3
4
5
6
26
May 2003
SSEC Retrieval Workshop
Detecting aerosols with AIRS
• Refractive index of silicates means that typically, BT(980) ≤ BT(820)
• Uncertainty in refractive index
• Have detected and analysed various desert storms with sand blown over
water : Atlantic (W.Africa Sahara), Eastern Mediterranean, Gobi Desert
dust to China Sea
• Assume only ONE cloud in FOV
• Use ECMWF fields for water vapor, temperature, ozone, surface params
• Use about 100 channels, spaced in the 10-12, 8 um windows
SSM, UMBC
27
May 2003
SSEC Retrieval Workshop
MODIS image for October 19, 2002 over E. Mediterranean
SSM, UMBC
28
May 2003
SSEC Retrieval Workshop
Using AIRS to detect dust : 960 - 1216 cm-1 BT diffs
SSM, UMBC
29
May 2003
SSEC Retrieval Workshop
Gobi, Cyprus Dust spectra : MILD cases that passed clear sky
test
300
B(T) in K
280
260
240
220
Cyprus
China
O − C in K
1
0
−1
−2
−3
−4
700
800
900
1000
1100
1200
Wavenumber (cm−1)
SSM, UMBC
30
May 2003
SSEC Retrieval Workshop
Gobi Dust spectra : SEVERE cases that failed clear sky test
290
B(T) in K
280
270
260
250
240
O − C in K
0
−2
−4
−6
−8
800
SSM, UMBC
900
1000
1100
−1
Wavenumber (cm )
1200
31
May 2003
SSEC Retrieval Workshop
Detecting volcanic eruptions with AIRS
• Mt. Etna erupted in late October, observed by many satellite sensors
• AIRS has sensitivity to SO2 , inside the strong water band.
• AIRS can detect eruption plume in the window region.
• Assumed plume was composed of basalt.
• Andesite and SO2 signatures well separated in wavelength
• Made preliminary estimates of SO2 column, plume particle size, and
plume particle density, assuming plume is about 8 km high
SSM, UMBC
32
May 2003
SSEC Retrieval Workshop
MODIS Image of Plume
SSM, UMBC
33
May 2003
SSEC Retrieval Workshop
Aerosol Plume
SSM, UMBC
34
May 2003
SSEC Retrieval Workshop
SO2 Plume
SSM, UMBC
35
May 2003
SSEC Retrieval Workshop
Oct 28, 2002; Granule 123; Profile 1502
300
Obs
Clear
Cloud
B(T) in K
280
260
240
220
0
O − C in K
O − C in K
5
−5
−10
Clear
SO2 * 4660
−15
−20
800
1000
1200
1400
1600
1800
2000
2200
2400
2600
Wavenumber (cm−1)
SSM, UMBC
36
May 2003
SSM, UMBC
SSEC Retrieval Workshop
37
May 2003
SSM, UMBC
SSEC Retrieval Workshop
38
May 2003
SSEC Retrieval Workshop
Dobson
SSM, UMBC
39
May 2003
SSEC Retrieval Workshop
Conclusion
• Infrared Spectroscopy
• Radiative Transfer : Clear and Cloudy
• Observations using the AIRS instrument
• Clear sky RTA is in good shape
• Is the Clear Sky Filter letting in thin cirrus effects?
• Cloudy sky RTA seems to be working quite well too!
SSM, UMBC
40