Minutes of the seventh International Radiative Transfer Workshop,
June 2005
Monday
Tuesday
Wednesday
Thursday
Whole Week
BREDBECK 2005 Minutes
=====================
Monday, 20/06/05
================
Comparison of stratospheric H2O profiles from an airborne microwave radiometer with
the ECMWF humidity product
Dietrich G. Feist
Airborne water vapor measurements near 183 GHz
- flight compaigns 1998-2004
- one campaign per year since 98
- typical flight route covers most of the nothern hemisphere
#Water vapor spectra in the upper side band (183 GHz)
AMSOS vertical resolution
- averiging kernel (broadband radiometer)
AMSOS measurements September 2002
- typical climatology of H2O
- can see upper troposphere in tropics
H2O measurement in the polar vortex (March 2000)
- summer time looks defferent than winter time
ECMWF humidity product
- H2O volume mixing ratio can be derived from ECMWF specific humidty
- assimilated water vapor in the troposphere
- parameterized water vapor (methane oxidation and transport) in the stratosphere
- good vertical resolution and altitude range since ERA-40 dataset
AMSOS/ECMWF intercomparison (single profile)
- good agreement in mid. altitude, bad in low and high
Mission 1 (Aug 98): AMSOS vs. ECMWF
- AMSOS H2O volume mixing ratio vs. ECMWF H2O volume mixing ratio
- horiz structure in experiment folows nicely ECMWF
- in general ageement is rather good
- 3 hour diff
- inside of polar vortex looks different
Mean difference by mission
- fall, summer good agreement
- spring bad, seasonal dependancy
Standard deviation by mission
- all look the same
Effects of ECMWF model versions
- huge change from one ECMWF version to next
Conclusion
- AMSOS measurments and ECMWF humidity product agree well in lower stratosphere.
Small scale horizontal structure is well resolved.
- systematic deviation in the upper statosphere. ECMWF probably does not model removal
processes for statospheric HO well.
- strong deviation inside and near polar vortex. Problem with ECMWF vertical
transport?
- quality of ageement has a strong correlation with ECMWF model version
Discussion
Stefan: What temperature profile is used?
Profiles are from Metoffice.
Stefan: Leveleing off of H2O?
60 km, apriori contribution is 40%.
Stefan: Any PSC's in ECMWF?
In EMWF, they not correlate with measurements. I dont think so.
--------------------------
Problems with kernel matrices for a retrieval of H2O using ARTS and Qpack.
Stefan Müller.
H2O radiometer at 183 GHz
- measurements by aircraft once a year during 1 week
- 2003 modificatons, problems started
- 2003-2004 problems with ARTS & Qpack
Measurement example
- strong base line from standing wave
- O3 line from side-band, Qpack can deal with these problems
Simulation of a retrieval
- input: apriori temperature profile, sensor character, etc.
- output: profile, A, Kx, Dy
Typical A from a simulation
- strange peaks in A
- platform altitude 10km, problem?
Where does the problem come from?
- peak from Kx
Search for the problem
- line strength
- platform altitude discontinuity?
- comparison of retrieval setup (working 22 GHz vs. 183 GHz)
- calculaton of Kx numerically
Line strength
- simulation with
a reduction of line strength of 183 by 10 and 100
- weaker line
Platform altitude
- simulation with different platform altitude
Comparison of setup at 22 GHz
- uses a polynomfit of the spectrum
- peak is smaller
Simulation around hygropause
- platform at 13km, no polfit - peak is still there
Comparison to numerical calculation
- Kx from ARTS
Averaging kernel with numerical Kx'
- A' = DyKx'
Summary
- unwanted peaks in averaging kernel matrix
- simulations showed
o problem comes from the matrix Kx
o line strength has no effect
o the peaks disappear when being above hygropause
- numerical calculation of Kx does not show the peaks
Discussion
Patrik: What H2O profiles are used?
US standard profile, the same as Dietrich used.
Patrik: Numerical calculation should be the same. Small change of ppm is
important. Jacobians of different units look different, altitude dependant.
More discussions off line.
--------------------------
Comparison of AMSU-B brightness temperatures simulated by ARTS and by RTTOV-7
Nathalie Courcoux.
Motivation
- UTH climatology
ECMWF fields
RT model
scale the obtained bt to UTH
- cannot use ARTS, too time consuming
- use fast RT model, RTTOV-7
- compare ARTS with RTTOV
Setup
- global comparison
- 1 day, 1 time, Jan. 1. 2000
- input: temperature, humidity, pressure ECMWF ERA-40 fields
- for both models ECMWF profiles were interpolated and smoothed to RTTOV pressure
level
- emissivity 0.6 and 0.95
Channel 18, emissivity 0.6
- good ageement
- well spread negative bias
- clear and visible positive bias in specific regions
Channel 18, dependence to IWV
- clear dependence of the possitive bias on low humidity
- dependence documented by Garand et al. for RTTOV 5 and 6
Channel 16, 17, 19, 20; emissivity 0.6
- negative bias in the tropics
- positive bias in the polar regions and mid latititude
- similar pattern for channel 18
Channel 16, 17, 19, 20 dependence to IWV
- for sounding channels strong dependence of positive bias on low IWV
- threshold at which bias changes sign is moving to higher IWV for channels
with lower sounding altitude
- for surface channels, positive biases occur for IWV lower than the mean IWV
Global distribution of IWV from ERA-40 prifiles
- December 99 to november 2000
- highest ocurrences of dry profiles in the polar regions
- only a yearly picture, there are strong seasonal vatiations
Emissivity
- snow emissivity is important beacase most of profiles leading to a
positive bias are located in snow covered area
- snow emissivity highly variable forom .45 to .95 at 150 GHz
Channel 18, emissivity 0.95
- no positive bias
- few profiles with low IWV
Channel 16, 17, 19, and 20, emissivity 0.95
- also less psitive bias left
Comparison of low and high emissivity simulations
- the low surface emissivity shows the largest discrepancies
- consistent with Garand et al.
- do the biases maximise with decreasing emissivity?
Comparison for emissivity 0.1
- maximum bias with low emissivity
Comparison of the continua
- flat negative bias due to continua
Summary
- low emissivity -> larger discrepancies
- biases and their standard deviations between the models are modest
- certain biases always occure over specific regions
- positive bias maximises with decreasing emissivity
- flat biases are due to use of different continuum models
Conclusions
- bias emissivity dependant, models handle surface differently
- channel 18: positive bias in dry regions is up to 1.5 K
- reported biases are important for numerical weather prediction models and climatology
from satellites
Discussion
Christian: 6 kg/m2 in channel 19 sees the ground
Christian: Why discrapancy in surface channels?
ARTS and RTTOV treat surface differntly.
--------------------------
Total water vapor retrieval in polar regions using AMSU-B
Christian Melsheimer.
TWV retrieval from AMSU-B
- basic: RTE
- assume linear relaton between opacity and total H2O
Algorithm
- measure Tb at 3 different frequencies, ground emissivity is similar,
but H2O absoprtion different
- relationship between Tb and TWV (opacity)
AMSU-B
- channels sorted: channel 16, 17, 20, 19, 18
- channel 3, 4, 5 for low TWV
- channel 2, 3, 4 for high TWV
Algorithm development for AMSU-B
- use radiosond (RS) profile, integrate TWV, simulate Tbs with ARTS
- linear fit for each RS profile
- find focal point
- linear fit for TWV
- use TWV accordingly
Comparison with NCEP reanalysis data
- dayly averages, good agreement
Extension to higher TWV using emissivity information
- use channels 1, 2, 3, but channel 1 is different from the others
- algorithm not independent on emissivity
- from SEPOR/POLEX campaign emission of various surface types in winter
was determined for frequency needed here
Summary
- TWV retrieval is possible for TWV < 6 kg/m2
- TWV retrieval from radiometer data with info on emissivity can be extended to TWV
of 10 kg/m2
- EU project IOMASA about assimulation of dirived TWV into NWP models
- TWV data might also be used together with regional models for water
cycle investigation
Discussion
Nathalie: More studies for emissivity other than Selbach's?
This project will do this.
Stefan: ECMWF ok for retrievals?
No, but it is better than just using emissivity 1.
Tuesday, 21/06/05
=================
Working groups.
Wednesday, 22/06/05
===================
Participants:
-------------
CD - Cory Davis
SB - Stefan Bühler
DF - Dietrich Feist
AD - Amy Dorothy
OL - Oliver Lemke
PE - Patrick Eriksson
AB - Alessandro Batta
NC - Nathalie Courcoux
JM - Jana Mendrok
MK - Mashrab Kuvatov
SM - Stefan Müller
CM - Christian Melsheimer
UR - Uwe Raffalski
SR - Sreerekha Ravi
AJ - Adam Jaczewski
BR - Bengt Rydberg
MP - Mattias Palm
CE - Claudia Emde
Chair: CD
Presentation of the working groups
===================================
CD: Scattering models: ARTS-MC, ARTS-DOIT, Alessandros model,
Janas model, RTTOV-SCAT
- Discussion about comparisons
- suggestion by Alessandro, uplooking simulation, 3D,
with polarization, rain
SB: ARTS users
- problems in using ARTS/QPack for up-looking instruments
- always problems with compatibility, feedback needed
- contact webpage to be created
- DF: installation of ARTS on Mac
- practical part: sort out problems in retrieval
AD: Cloud microphysical assumptions
- many free parameters
- consider bulk properties
- more information from other sources should be used, e.g. radar
DF: ARTS Beamcat
- Script for producing arts linefiles from beamcat produced by DF and OL
CE: New ARTS development
- new agenda concept
- analytical jacobians, ppath array structure
- data format, xml not practical for some cases
- Monte Carlo: include sensor characteristics for EOS MLS and surface
- include new option: sensor inside cloudbox
- port absorption from ARTS-1-0 to ARTS-1-1
- general ARTS-1-1 paper
- Documentation, new ARTS Wiki, try to keep user guide complete
- discussion about implementation of fast scattering model
MK: AMSU UTH and climatology
- water vapor daily cycle in Mediterranian Sea
- investigate this area in AMSU-B data, is it possible to investigate
boundary layer?
- trend over several days
- cycle could not be seen in AMSU data
- suggestion by AB: use meteosat data
DF: ECMWF AMSU comparisons
- AJ found bias, 3K in centre of H2O line, 15K in surface channels
- Start with 3K bias
- bias is a technical bug, could not be sorted out
- NC: no bias between RTTOV and ARTS
- possible explanation: antenna patterns, less impact on nadir radiances
Talk session
=============
PE: Odin-SMR cloud ice retrieval
---------------------------------
- Odin: 2 instruments: SMR, OSIRIS
- used Odin-SMR data, stratospheric mode (501.18 - 502.38 GHz)
- tangent points below 9.5 km
- measurement principle, blackbody radiation below ~10km
- with cloud BT decreases, would become more complicated for higher
tangent altitudes
- lower retrieval limit about 10 km
- example spectra and retrievals, cloud detection
- ice columns above 260 hPa - similar results for SMR-MH97 and ECMWF
- similar retrievals are done for EOS-MLS
- particle size distributions, large deviations, main uncertainty
- outlook: improve Odin-SMR retrievals, collaboration with EOS-MLS,
Chair: PE
CD: AURA-MLS cloud observations
-------------------------------
- dI for thin clouds approx. proportional to IMC
- polarization signals can be quite large, explained by horizontally
aligned ice particles
- AURA-MLS: First dual polarized measurements
- Focus on radiometer 1, measures H and V, centered at O2 line
-> not ideal for cloud detection
- polarized simulation for different particle shapes
- BT depression does not depend much on shape, but polarization
- polarization signal lerge for horizontally aligned particles
- understanding Q: low tangent altitudes -> Q>0
- measurements: polarization signal abou 10% of BT depression
- interpretation of observations: moderate aspect ratios reproduce
measurements
- conclusion: randomly oriented particle assumption seems to be justified
- MLS observations and MODIS cloud height
Discussion
(CE) higher polarization signal for other radiometers - may be assumption of
randomly oriented particles not valid
(PE) with 122 GHz mainly convective clouds are observed -> randomly oriented
particles, probably different in thin cirrus
BR: Gereration of cirrus retreival test data base
--------------------------------------------------
- database content: ~100 000 cloud cases, depending on number of parameters
to be retrieved
- retrieval parameters: Column IWC, IWC profiles,
size parameters (mean mass diameter, median mass diameter, ...), shape
- Cloudnet radar data, Products: radar reflectivity factor,
inverted IWC and LWC
- microphysical assumptions: use PyARTS for computations, but not very
realistic size assumption, gamma size distribution
- try to match gamma-distr. parameters by Heymsfield 2003
- simulations using DOIT for CIWSIR frequencies
- database is under construction, plan: use information from radar
Discussion:
- (CE): Gamma distribution can be reproduced using one particle type
(CD): Does not work because you have altitude dependant size distributions
(JM) Pseudo-spherical RT modelling for emitting and scattering atmospheres
--------------------------------------------------------------------------
- devide solar term and emission term cause problems in IR (2.5 - 4.0 microns)
- features from troposphere can be observed in limb geometry because of
scattering into the LOS
- rte with four partial derivatives
- simplified spherical RTE (integral form of RTE, 1D, local panarity of
atmosphere
- existing modules used: Absorption (F. Schreier)
precalculated optical properties, DISORT
- parabolic parameterization of extinction for calculation of optical depth
- source terms: emission, solar radiation in spherical geometry
multiple scattering term in plane-parallel geometry
- with solar radiation 2D problem
- validation using ARTS and KOPRA
model without multiple scattering compared to KOPRA
with multiple scattering compared to ARTS
- comparison with MCScia for different solar angles with single scattering
with multiple scattering
problems with very low sun cases
- reconstruction of measured MIPAS spectra
- intercomparison with McSCIA for more setups
(MP) Taking Bayes's Theorem seriously
-------------------------------------
- inverse problem -> Bayes theorem
- application: electromagnetic conductivity imaging (ECI)
- discretizations in cells of constant conductivities
- ECI as probability problem
- Likelyhood : Gaussian noise A priori: Pott's model
- Motivation for using MCMC
- Definition and properties of MCMC algorithm
- Example: largest errors at conductivity boundaries
- Finding optimal size of state space, problem: correlated samples
- Metropolis coupled MCMC
(PE) Article using method in JGR, Tamminen and Kyrölä
(SB) Include this into MC retrieval scheme similar to Evans methods
Thursday, 23/06/05
==================
Chair: CM
WG: Cloud retrieval
PE: CIWSIR: different retrieval methods, neural, MC integration, OEM (only gaus probabilities, for forward models only,limb sounding sensors), 3 first fro small state vector
Relevant set of retrieav products: IWC, size, shape and orientation of the particles (45degrees polarization better because of symmetry)median better than mean of non-gaussian statistics, WVC, profile, T, H2O, external info is needed, f.e. from ECMWF
correlation between water vapour and liquid clouds, how to model it still unknown
discrepancy between scattering models solved (MC and DOIT)
CM: want to know more about CIWSIR, PE answers: higher frequencies as for AMSU (WV transitions: 3 channels, total water vapour: 9 channels, res: 10 km, down looking), expectes in 6-8 years
WG: ARTS user group
OL: Dietrich problem solved and other problems, units of weighting function etc.
UR: QPack working on his windows
Green and red cards part :-)
Future plans:
PE: ARTS 2.0, cloud box part, new definition of agenda set, adding scattering part module, inversion modul for limb sounding sensors, will build QPack for ARTS 2.0, Wiki page for QPack,
Thanks to Nathalie from Workshop participants
Monday
Tuesday
Wednesday
Thursday
Whole Week
