Minutes of the fourth International Radiative Transfer Workshop,
July 2002
Monday
Tuesday
Wednesday
Thursday
Whole Week
*******************************************************************
********************** RT WORKSHOP JULY 2002 **********************
*******************************************************************
-----===== MONDAY 2002-07-08 =====-----
1. Welcome Prof. K. Kuenzi
2. PE:
- structure of the workshop
- think about the working groups
- hand out of the agenda and the draft of the intercomparison
paper
- change of agenda:
barbecue on Tuesday
- Ute will give an additional talk about polarization in the UV/vis
on Thursday
- introduction of the workshop members
PE, Chalmers Univ. of Tech., Goeteborg
KK, IUP/sat Bremen - head of ife-microwave groups
DF, IUP Bern - RT, spec. param. catalog data base
SR, IUP/sat Bremen - ARTS scattering
CM, IUP/sat Bremen - ARTS scattering
YK, CRL, Tokyo - SMILES Team
ED, OdB, Bordeaux - Moliere/Odin
JL, OdB, Bordeaux - Moliere/Odin
UR, Swedish Inst. of Space Phys., Kiruna, ground based radiometers
HH, IUP/RAM Bremen - ground based radiometers
AH, IUP/sat Bremen - HIRS data analysis with ARTS
IW, IUP/RAM Bremen - ground based radiometers
MK, IUP/sat Bremen - ARTS application for AMSU-A/B
JM IUP/sat Bremen - polarization and scattering
CV, IUP/sat Bremen - SMILES team
JM, DLR/Berlin - RT models of IR and vis
FS, DLR/Wessling - IR RT models and different other applications
UB, DLR/Berlin - POLDER/ADEOS-I data analysis, RT models in vis/UV
NK, IUP/sat Bremen - spectroscopy, Zeeman effect
HB, Inst of Astronomy, Sao Paulo, Brasil - ground based microwave radiometers
GK, FZK/Karlsruhe - ground based radiometer
CE, IUP/sat Bremen - ARTS scattering
TK, IUP/sat Bremen - ARTS development (water vapor abs. models)
VJ, IUP/sat Bremen - ARTS application for AMSU-A/B
OL, IUP/sat Bremen - sys. administration
3. short presentations of RT models
a) (CE) introduction into scattering in ARTS-1.1
- 1D to 3D RT model
- polarization included in RT equation (Stokes vector)
- spherical geometry included (limb sounding)
- scattering is considered in boxes where clouds are present
b) (UB) RT equation in the vis region
- Fourier series of the azimuth direction to separate this
dimension from the others in the phase function
- matrix operator method
- the ocean surface layer scattering is included into the RT model.
- spherical particles with Mie theory is included
- plane parallel model with 1D atmosphere
c) (FS) MIRART developments
- efficient line-by-line RT model
- switch from pure F77 to F77 with script language Python
- weighting function with automatic differentiation
(fast and more accurate than finite differences)
"more accurate" ---> exact
(at least automatic differentiation Jacobians are as "exact" as the
original code)
d) Spectroscopic Database (DF)
- BEAMCAT -
- merge different databases
- simulation of spectra
- JPL and HITRAN
- different catalogs has good database in different ranges, so
merge data from different sources
- merging problem - lost information, the better way is to merge
data without losing any information
- META CATALOG - contains information that identifies the same
transitions in different catalogs
- SQL database - fast sorting and searching - output format can
be easily changed - new parameters can be added at anytime -
easy web access
- Demonstration of BEAMCAT
- meta catalog has to be made once by identifying the same lines
in all the catalogs(comparing quantum numbers)
- Robust line matching scheme
- web-site to access BEAMCAT data:
http://www.iapmw.unibe.ch/research/projects/BEAMCAT
- info from all catalogs, format independent, easy web
interface, easy t update
- the meta catalog is in the form of a matrix. No limit for
input source. User can update data base or write private
tables and so on. Partition functions also will be added.374
rules to merge JPL and HITRAN. In JPL some places for quantum
numbers are empty and some times letters instead of numbers are used
and this is not documented anywhere. All the entries in
BEAMCAT has a time stamp so that the new additions can be easily identified.
e) Water Vapor Continuum (TKS)
- different models: sources, theory..
- model validation: ARM, lab data
- documentation of CKD model is available in GENLEN2??
- comparison of CKD model, accurate
- SHEBA AERI/LBLRTM radiance to show the effect of different
CKD versions.
- Use of CKD model: ACECHEM, HIRS, climatology UTHMOS
- other continua bands: N2 collision broadening
- Liebe and Rosenkranz may not be valid in FIR region
- MonoRTM V2.01 is Rosenkranz in millimeter range and the normal
CKDv2.4 model in the FIR range
- CKD is the standard model in the FIR range. It contains only
the continuum part and not the lines.
- This model is implemented in ARTS and now ARTS is able to do
calculations in the infrared region
f) Formation of working group
1. Scattering from the Microwave up to the vis/UV
Participants: Claudia, Sreerekha, Christian, Ute, Jana,
Jana, Franz, Nikolay, Joachim
2. Model vs. Data - Comparison and Validation
Participants: Thomas, Mashrab, Viju, Arash, Tim, Susanne
3. JEM/SMILES
Participants: Carmen, Yasuko, Eric,
4. ARTS User Group
Participants: Patrick, Oliver, Ingo, Dietrich, Hermann,
Uwe, Gerhard, Helmut
-----===== TUESDAY 2002-07-09 =====-----
a) Intercomparison paper
The model named as MAES is in named in the plots as CRL and ss
as FIP
1. case 0
- Nikolay is giving description of the cases
- Moliere uses Kuntz implementation of the Voigt function.
- DLR is not using the partition function but some software
called ATMOS so the line strength can be different.
Franz is giving a short description of line strength
calculation by ATMOS.
* comment from Franz:
+ Kuntz is simply an optimized version of the Humlicek algorithm to
compute the Voigt profile / complex error function (Voigt=Re(cerf))
In general I use Humlicek as well, just with a different optimization.
for most of the case0 and 1 calculations submitted to Bremen, however,
I used the wofz code of Poppe&Wijers which (according to the authors)
has an accuracy of 14 significant digits (Humlicek: 10^(-4)).
+ My line strength conversion is done according to the scheme
described in Norton&Rinsland:
@article{Norton91,
AUTHOR = {R.H. Norton and C.P. Rinsland},
TITLE = {ATMOS data processing and science analysis methods},
JOURNAL = AO,
VOLUME = 30,
PAGES = 389,
YEAR = 1991}
ATMOS is a FT spectrometer flown several times on the space shuttle.
Atmospheric Trace Molecule Spectroscopy Experiment
http://remus.jpl.nasa.gov
2. case-1A
- Similar results as case0 except for FIP.
- Karlsruhe also uses ATMOS scheme for partition function
3. case-1B
- MPM-93 has lot of discrepancies.
4. case-2
- Karlsruhe uses O2-Liebe98 model, so it can be the reason for
discrepancy
- Moliere uses MPM-93 for oxygen and nitrogen.
- SMILES simulator is showing large discrepancy at higher alt.
5. case-3 Down
- DLR and Moliere participate
- coarse pressure grid and fine pressure grid:
differences are some Kelvin (see paper)
6. case-3 Limb
- In comparison with ARTS and MIRART ARTS is missing the
peak(FS).
- Deadline of the paper: August(to finalize the models and cases)
and September(to do the final iteration of the results).
- Wednesday there will be meeting people for each groups.
AMIL2DA(FS)
- Advanced MIPAS level 2 analysis (EU_FP5 project)
- www.fzk.de/imk/imk2/ame/amil2da
- main question:
Are MIPAS level 2 data processed by different institutes/processors
comparable?
- Strategy:
1. forward radiative transfer model intercomparison
2. blind-test retrievals using synthetic spectra
3. intercomparison of retrievals with real mipas data
- Forward models:
KOPRA (Karlsruhe), OFM (IROE, Florence), RFM (Oxford),
MIRART (FS), FM2D (RAL)
- In the intercomp. 30 exercises each with only one spectra at one
frequency range starting from simple settings and proceeding to
more complex/realistic cases.
- non LTE implementation is also checked (Kopra and RFM only)
- KOPRA is taken as the reference model
- Transmission: large difference in the CO2 2400 - 2500 cm-1
region (4 micrometer),
250mb, 256 K was coming from different line rejection criteria.
- Radiance (incl. ILS and FoV effects):
deviations for all models less than 1%
- Blind test retrieval - first results:
all codes giving reasonable/consistent results for ozone etc.
Zeeman Splitting(NK)
- contact Nikolay
- Yasuko told BrO and ClO also has small Zeeman splitting
- Ozone??
RT through the atmosphere of other planets(JU)
- Moliere
- MAMBO - MARS ATMOSPHERIC MICROWAVE BRIGHTNESS TEMPERATURE
EXPLORER, Limb looking
- Launch will be in sept 2007, Ariane5, arrival 2 years later
- present and past Martian atm, temperature and winds, general
circ, water cycle(0-60 km), photochemical state- CO, H2O2,
surface(polarisation)
- profiles H20 seasonal variation, temperature is decreasing
with altitude
- nadir operations also
- Moliere is used to calculate spectra
- wind is measured by Doppler shift in the CO line
- Instrument description
- different data levels upto level 4
- mars radius - 3397?
- ground pressure - 7hpa
- H20 and CO2 continuum are implemented
- LTE can be used???
- surface emissivity can be a problem
- strong variability of H2O can be a problem in the retrieval
- retrieval of the temperature from two CO isotopes
- for H2O 5 km grid is used
- The first test retrievals shows temperature and water vapor
can be retrieved, spectroscopy part is seemed to be more
complicated
Forward modeling in the Visible region(Ute)
- unpolarized source
- Polarisation in remote sensing
- Instruments: reflection, refraction, self polarisation
- picture from POLDER to illustrate the effect of polarisation
- polarising processes in nature:
- UV Rayleigh scattering
- Rayleigh scattering - maximum at 90 degrees
- scattering by cloud particles - degree of polarisation is
very sensible to microphysics of the cloud
- non sphericity can be measured through polarisation
- reflection from sand is slightly polarised.
- moist surfaces polarize more
- polarized light from a canopy is coming from the specular
reflection from the leaf surfaces
- on water surface- Fresnel's law, angle dependence
- ground is modeled as
- for aerosols types and optical thickness are retrieved
ARTS Application
Odin Inversions(PE)
- Odin spectra
- don't get exactly the pointing offset, has to be looked at
Ground based inversions in Sweden(PE)
- plot for measurements
- mesospheric CO is measured
- Vertical profiles of CO above Onsala, a downward transport of
CO is seen
-
SMILES Inversion(CV)
- Instrumental description
- three spectral bands
- very low system noise temperature
- nice spectrum
- temperature retrieval is tried.
- temperature weighting function(with and without HSE)
AMSU Intercomparison(MK)
- AMSU and Lindenberg data
- comparison between ARTS and AMSU
- plots for NOAA 15 and NOAA 16
ARTS for AMSU(VOJ)
-
Outgoing Longwave(SB/PE)
- Balance of incoming and outgoing radiation
- difference of real spectra from black body curve
- plot of sensitivity to double H2O, maximum sensitivity 900
hpa to 300 hpa
- same plot for doubling of CO2, only in a narrow frequency
range
- H2O is more important than CO2.
Tangent Linear and Adjoint Models(CM)
- contact Christian
- R M Errico "What is an adjoint model?" BAMS, 1997
- C D Rodgers, "Inverse Methods for Atmospheric Sounding",
World Scientific, 2000.
- Some web sites
- Adjoint model compiler
- Papers by .... told by Franz.
Go to HTML and PDF version of the presentation.
Retrieval Tool(PE)
- Qpack
- Implemented in matlab and uses ARTS as forward model
- only OEM
- example control file
- How to call functions
- creation of covariance matrices
- Franz:
+ PyReS: 'Python based retrieval system'
+ application: FOCUS, MIPAS, THOMAS,...
+ forward models: MIRART, MODTRAN, .....
(steered by PyReS using appropriate template file)
+ inversion codes: (standard) nonlinear least squares libraries,
Birger Schimpf's Tikhonov regularizatio code, ....
+ NOTE: only partially implemented right now
(currently works for Focus retrievals only)
- Fully integrated systems: forward and inverse programs
together: Moliere, Karlsruhe
Retrieval Error Analysis for Spectroscopic Parameters(CV)
- Rodger's OEM algorithm
- Linear mapping method
- inverse model linearized around a priori state
- OEM method will be better since the error covariance can be
treated as Gaussian
Working Groups
- intercomparison working group will be met and try to figure
out problems regarding writing of the paper
------------
ACTION ITEMS
------------
o Intercomparison paper:
- iup/sat Bremen will unify the plotting style for all the
necessary plots of the paper
- all participants of the intercomparison have to obey the
time schedule for the delivering of the model information
and final data sets.
- main editor (one of the iup/sat Brmen member) of the paper
has to be named by iup/sat Brmen.
- Franz will send the AMIL2DA paper and the template for the
model description used in the AMIL2DA comparison to
iup/sat Bremen.
-----===== WEDNESDAY 2002-07-10 =====-----
- Group work
1. Scattering from the Microwave up to the vis/UV
Participants: Claudia, Sreerekha, Christian, Ute, Jana,
Franz, Nikolay, Joachim
The focus of this group was on two scattering models,
1. ARTS.1.1 which is in the development stage at IUP, Bremen and
2. STORM from DLR.
The group discussed at length the different scattering schemes
implemented in these two models and could draw out conclusions
regarding the differences and similiarities in the different
approaches. A brief note can be found here(ps format). A detailed
description of the STORM model is also accessible(ps file). You may
need these figures also for understanding the geometry(ps file) and
integration scheme(ps file).
-----===== THURSDAY 2002-07-11 =====-----
Working Group Scattering
------------------------
CE: discussion of the different approaches of the RT equation from the
mm (ARTS) and vis/UV (STORM model) region for the different parts,
extinction/scattering/absorption.
Successive order scattering model in ARTS-1.1 (e.g. loops until
convergence)
Ute: This loop is sensitive to the evaluation of the phase matrix
since you loop many times over the phase matrix integral.
CE: STORM and ARTS handle diffuse polarized light
ARTS handles also rotationally symmetric aspherical particles while
STORM handles spherical particles.
STORM uses Fourier decomposition of the phase matrix and intensity
between azimuth and zenith angles (only possible for spherical particles).
NK: JM has shown that in the mm range the calculation shows that
the scattered waves are coherent in the forward-backward
directions. ARTS assumes incoherence in all directions.
CE: plans:
- use of Gauss-Loballo quadratures method used for integration
- check if the convergence of RT equation is also possible for
larger optical depths.
- use of Tim Rother (DRL) data base for cloud particles
Ute: interaction principle in STORM:
describe the way of the light through the different layers of
the atmosphere in terms of emission, reflection and
transmission. This includes already multiple scattering.
It should also be possible to use this approach for ARTS.
Ute: one should be aware of how the treatment is with contributers to
ARTS from outside Bremen/Goeteborg. How are contributers treated
with respect to publications and ARTS distribution and such
stuff? How is the policy?
PE: Bremen should create a www page about this workshop results with
results and links.
PE: we hope for a next workshop next year at Bredbeck.
PE : Instead of just discussing the forward models, we can include
retrieval methods in the future workshop
A brief note on our discussions can be found here(ps format).
A detailed description of the STORM model is also accessible(ps file).
The figures can help in understanding the geometry(ps file) and
integration scheme(ps file).
Working Group Validation and Comparison Model vs. Data
------------------------------------------------------
MK: AMSU-A/B data analysis
match of ARTS/radiosonde data vs. AMSU-B data
good agreement for channels 17/18/19/20
channel 16 agrees with ARTS by using an optimized emissivity
AMSU-A: ch. 4/5/6 we see a slope in the comparison,
data is higher in TB than ARTS/radiosonde predict.
suggestion to do the analysis for another location
Cabauw (NL) where SC and TH have some measurements done.
VJ: emissivity retrieval for Cabauw (NL) to see the robustness of the
retrieved emissivity at Lindenberg.
SC: CLIVANET measurements (part of BALTEX experiment) at Cabauw (NL)
Intercomparison with standard abs. models. for 20-90GHz
differences of up to 2K@30GHz and 7K@90GHz(?) are seen.
At 90GHz the diff. can not be explained at the moment.
TH: MARSS ground based measurements at Cabauw (NL).
Additionally airborne based measurements with MARSS were performed
numerously at different climate conditions (equator to poles).
H2O foreign continuum were tried to adjust to data of all the
flights.
Difficulty is that all the lab. data for cont.fit is >273K which
leads probably to differences in the low temperatures of polar
regions.
However, it can not be unambiguously decided from the MARSS data
set which component of the continua has to be updated.
TK: presents comparison between the MARSS/POLEX measurements with ARTS.
Future plans is to compare Atmospheric Radiation Measurement
Program (ARM) data with ARTS different water vapor
absorption models, including the new parameterization of
JQSRT, 2002, vol. 74, pp 545-562, 2002.
Working Group SMILES
--------------------
ED: - New limb submm sounder on ISS developed by Japan NASDA/CRL
which will be launched in 2006.
- New mission plan end of July 2002 available (ask YK for a copy).
- Comparison of retrieval simulation results. In general good
agreement between Moliere, Smiles Simulator and ARTS.
- Structure of the ground based segment discussed (which
inst. processes which level in the data processing chain).
- Baseline retrieval is performed -> will be an additional
option for Moliere.
Working Group ARTS
------------------
DF: installing is quite easy. Documentation for the user installation
and user experiences with ARTS is missing.
action items formulated from this group:
DF: interface to BEAMCAT
OL: easier handling of MSDOS/UNIX line breaks in files
OL: set up ARTS user mailing list
OL: create development platform to allow external developers easier access to
the ARTS cvs repository
PE: clarify hidden spectral line parameters in arts and make it
more transparent (e.g. line broadening for JPL catalog, isotopic ratios).
PE: QPAC running on MS Windows machines
PE confirms that all the ARTS developers from Goeteborg/Bremen will
give as much support as possible for all the ARTS users. However
ARTS-1-0 development will stop now and ARTS-1-1 is in progress. For
normal users ARTS-1-0 should be sufficient and is more stable than
ARTS-1-1 which is a big project for the next years.
------------
ACTION ITEMS
------------
- all participants can contribute to the workshop home page with
additional information either by sending appropriate electronic
versions of a www link whwich will be included in the workshop home
page.
- SC will provide all participants with the necessary information to
access the home page of the CLIVANET measurement campain at
Cabauw (NL).
Monday
Tuesday
Wednesday
Thursday
Whole Week
