Minutes of progress meeting 6, 18 September 2009, Ålesund, 9:00 - 17:00

Participants

  • Marc Bouvet (MB), ESA
  • Claudia Emde (CE), DLR
  • Bernhard Mayer (BM), DLR
  • Arve Kylling (AK), Ålesund

Agenda

  1. Discussion of Task 3, Algorithm development
  2. Discussion of Task 4, Verification
  3. Intermediate libRadtran version
  4. General issues

Minutes

Discussion of Task 3, Algorithm development

1. Raman scattering (AK)

The code works for Rayleigh atmospheres, but development not finished due to problem with cloud scattering:

  • cloud decreases filling in factor too much, this error is still under investigation (it is not due to the Raman scattering cross sections; the elastic scattering in the qdisort solver is also correct which a comparison with the disort2 solver has shown)

Computation time is large, optimizations preferable but probably beyond the scope of this project.

2. Modification of DISORT (BM)

Considerable work has been spent on improving the DISORT solver: Previously, the scattering phase function was provided to DISORT by means of its Legendre decomposition which required more than 10000 terms e.g. for the highly peaked phase functions of ice crystals. In many cases this caused low computational speed, huge memory requirements, and even numerical uncertainty because a decompoistion into more than 10000 Legendre terms is not trivial. Robert Buras managed to re-write DISORT so that the phase function is now provided directly to disort, using typically 500 angular grid points. The output of the modified DISORT has been extensively compared to unmodified DISORT and differences were negligible. With this development it is now possible to provide optical properties of clouds at much higher spectral resolution, including polarization.

3. New databases of optical properties (CE)

A new binary format (netcdf) has been generated for the optical properties of aerosols and clouds. This format includes the phase matrix, Legendre coefficients, extinction coefficients, single scattering albedo, and the refractive index and can be used by all libRadtran radiative transfer solvers. Water clouds (short wave), OPAC aerosols, and ice clouds (Baum) are already available in this format. In order to close WP3200, the water cloud data in the thermal region needs to be calculated. This will be done until the end of the month (delivery of intermediate libRadtran version).

4. Graphical User Interface (AK)

The GUI has very much advanced since the last meeting and can now actually be used for running uvspec since it includes all input options in a structured way and conveniently allows plotting of the model output. It is now automatically generated from the source code (uvspec_lex.l). Documentation for all options is available which is also extracted from the manual. Some parts should be improved:

  • options that are modified can not be unset
  • the plotting does not work for all solvers since they have different output formats, it would be good to have a commom netcdf output format for all solvers
  • the GUI needs more logic (options should appear only when it makes sense to specify them). This information is not contained in uvspec_lex.l, therefore we need another structure where the options are specified and from which GUI and uvspec_lex.l file are generated. CE suggested to use a python structure to specify the options, because this could be easily used by the GUI.

The libRadtran version to be delivered by end September will include the current GUI version. AK asked MB to provide feedback that can be used to improve the final GUI version.

Line-by-line calculations work with molecular_tau_file (CE should check with DLR-IMF when their tool to generate this file will be available).

5. General discussion about open issues in Task 3 (all)

  • Example for optical properties for aspherical aerosol particles has been generated (WP3220 closed)
  • Polarized BRDFs are available, e.g. from Hasekamp et al. (2007). They will probably be included into libRadtran (final version).
  • MB mentioned that a new surface property database is currently being developed in an ongoing ESA study. BM agreed that new database (including surface BRDFs of whole Earth) should be interfaced to libRadtran which should not be much work since infrastructure is already there
  • Refraction has been included and produced reasonable results. Validation yet to be done CE will contact the University of Bremen in order to perform a model inter-comparison with SCIATRAN.

Discussion of Task 4, Verification

1. Polarization (CE)

The implementation of polarization in MYSTIC has been verified against benchmark results (for Rayleigh scattering and aspherical aerosol particles), measurements and by a model intercomparison (where MYSTIC agreed with PSTAR and SCIATRAN). The validation is complete and MYSTIC works correctly for molecular, aerosol, and cloud scattering. In the comparison between MYSTIC and the polradtran solver, it turned out that the polradtran accuracy might be not sufficient. So MB asked whether another 1D solver for polarization could be included into libRadtran. CE said the PSTAR might be a possibility, but the developers have to be contacted first. Including a new solver is not part of the ESASLight project and can not be promised until the end of the project.

2. Raman scattering (AK)

Most verification cases for Raman scattering for clear sky have been finished. A good agreement with results from the literature has been found. The verification for cloudy cases can only start when clouds are implemented correctly. AK will contact Spurr to ask whether a model-model intercomparison would be possible.

3. Test suite (CE)

The development of a complete test suite has started. In a first step a python structure for uvspec options has been generated. From this structure, the GUI, the testsuite, and the documentation should be generated. A first python script has been generated that produces all possible combinations of options regarding water clouds.

Intermediate libRadtran version

BM presented an overview of the preliminary libRadtran “ESASLight” release. This version will be put on the ESASLight webpage for download by the end of September. Additionally to the public version on the libRadtran webpage, this version will include MYSTIC with the new features (polarisation, refraction). Furthermore the Raman scattering solver qdisort will be included (without cloud and aerosol scattering). Also, the first working version of the GUI will be included.

Task 6 - Future extensions of libRadtran

A new activity will probably start next year.

Possible extensions for libRadtran are listed here. Further discussion points were:

  • improve speed of Raman scattering calculations
  • other planets (MB: trace gas profiles databases for Mars, Venus, Titan available at ESA)
  • coupled atmosphere-ocean RT calculations (MB said that standard analytical functions for water optical properties are available)
  • include other 1D solver with polarization
  • MB appreciated the idea of organizing a libRadtran workshop. This could be done within the next activity. The workshop could take place at ESTEC (according to MB facilities are probably available) or at Schneefernerhaus (Zugspitze), which BM suggested.

Other issues

Next meeting

If there will be much feedback and questions by libRadtran users at ESTEC, the next meeting will be at ESTEC, otherwise at the LMU (University of Munich). The meeting will take place on the 17 of December 2009.

Review of Action Items

  1. AK: Include the possibility to unset options in GUI
  2. BM, CE: Put intermediate libRadtran version on ESASLight webpage
  3. CE: Contact DLR-IMF regarding line-by-line tool
  4. CE: Finish database for water cloud optical properties (thermal range)
  5. CE: Contact UB for a intercomparison with SCIATRAN (validation of refraction)
  6. AK: Contact Spurr regarding a model intercomparison for Raman scattering
  7. AK: Disable Raman scattering in intermediate version, when clouds or aerosols are included
  8. MB: Find out whether a report about the ESASLight project could be put on GSP website
 
minutes_20090918.txt · Last modified: 2009/09/25 12:50 by esaslight     Back to top