Meeting Minutes

Session: GLI Algorithms: GAIT Atmosphere Meeting

Chair: Dr. Teruyuki Nakajima

December 8, 1999 (16:30-18:30)

Room: E

  1. ATSK1 and ATSK 3p ( r ), Dr. Tairadate(Fujitsu)
  2. ATSK5 and ATSK16 by Dr. Liu(EORC/GAIT)
  3. ATSK 3e and ATSK 4, Dr. Zhang, EORC/GAIT

Summary of Major Discussions

1.ATSK1 and ATSK 3p ( r ), Dr. Tairadate(Fujitsu)

Dr. Tairadate presented the above two research topics on behalf of his team. He summarized the definitions of the products and input/output files. Dr. Tairadate made suggestions to the PI of ATSK 1 algorithm. The following are some of his questions and suggestions:
  • Which channel should be used in GLI?
  • Confirm whether "bright50.f" and "get_data.f" support GLI Channel or not.
  • Deliver revised "get_scf.f" for GLI Land/Water flag.
  • Tell us when you deliver new code.
  • Confirm whether you can eliminate context processing from ATSK 1 or not.
    If so, when do you deliver the new code?
Dr. Tairadate summarized the discussion from the previous day's meeting.
  • Currently, 12 channels are used in GLI: 8,13,19,24,27,28,29,30,31,34,35,36.
  • GLI revision analysis: GAIT delivered read module for GSD data set, and PI (Ackerman) revises the analysis.
  • Revision code for 2km sampling channels: GAIT delivered read module for 2km channel. PI does the analysis module for 2km channel processing.
  • Context processing: Is being used because the processing is crucial for ocean scene.
  • Cloud Flag: Current one is a 16 bit. There is a plan to expand to a 32 bit.
    Reserve 4 new bits for ATSK 4.
During the presentation, the following issue arose as a major concern:
  • The original meteorological is hPa. Should kilometers be used in expressing the axial instead of hPa?
Questions and Answers
Q: Regarding ATSK 3p(r), do you analyze 1.6 and 2.2 µm radiances? (Dr. Teruyuki Nakajima)
A: Not for this version.
  • Dr. Teruyuki Nakajima's comment: I recommend the creation of 1.6 and 2.2 µm analysis units.
Q: Which data have do you use for the surface albedo to correct for the retrieval of cloud optical thickness? (Dr. Fischer)
A: A Standard database is applied.
Q: Why do you use height as vertical dimension instead of pressure for the cloud top retrieval? (Fischer)
A: Will be considered.
  • Dr. Fisher's Comment and question : Surface albedo should be derived from GLI data and updated because of seasonal changes. Which effective radius is used to generate the lookup tables?
Q: What is the source of Ag libraries? (Dr. King, NASA GSFC)
A: A standard database of four seasons. We plan to use surface albedo derived from GLI data.

* Dr. King's Recommendations
  • Do not restrict algorithms to using three bands. Add 1.6 and 2.2µm if possible, even if available only for every other pixel.
  • Rayleigh scattering is important above cloud when (i) cloud optical thickness is low, and (ii) Åg is large. Consider using 0.865 (band 19) instead of 13 over the ocean.
  • How do you select cloud phase? Must be done before Tc/re retrieval because libraries are different for water drops and ice crystals.
  • The Algorithm as implemented will not work well (for Tc) over snow and sea ice surface. Consider using bond 26 (MODIS algorithm) for "visible"channel in these ecosystems. (Ag<0.5 rather than 1.0 as in band 13).
Q: How do you change CLOP, CLER, etc. when bad fit, and what constitutes a bad fit? (Dr. R. Marchand) A: Newton's techniques. A bad fit consists of 1% at each wavelength.
Q: Does it continue, and what size distribution do you use? (Dr. Marchand)
A: Yes, it continues about 95% of time with test data. We log normal with fixed σ.
Q: What do you do for averages for a loss of a few pixels? (Dr. Marchand)
A: Will think about reporting.

Back to index

2. ATSK5 and ATSK16 by Dr. Liu (EORC/GAIT)

Dr. Liu gave a presentation on ATSK 5 and ATSK 16. Firstly, he briefly defined the ATSK 5 products, their input and output files and algorithm. He also provided the theoretical background of the algorithm; "By theoretical analysis and numerical simulation, the satellite's received apparent reflectance is composed of six items." Secondly Dr. Liu proceeded to the presentation on ATSK 16. He explained its overview including the product information, algorithm input files and output files.
Questions and Answers
Q: (Dr. King) Have you tried to use any of the MODIS aerosol algorithms, which includes aerosol over the land?
A: I have read some algorithm, papers and codes. But presently, it is because my task for now is implementation. Afterwards, I will have time to work on it.
Q: What is the source of ancillary data?
A: I use three kinds of data: ozone amount, water vapor amount, and the 10m height wind speed.
Q: ( Dr. Roger Marchand ) For this kind of classification, you need to process a lot of data. Do you do every pixel segment?
A: Yes, but not 25.
Q: (Dr.Teruyuki Nakajima ) Why do you limit the analysis by the latitude range of 60ºS to 60ºN? It depends on the angle condition.
A: Yes. The retrieval should include a wider latitude if the angle conditions are suitable. So far, we have applied to reprocessed AVHRR data in < +60.
*Dr. Teruyuki Nakajima recommends limiting by solar zenith angle of < 70.

Q: (Dr. Teruyuki Nakajima) Do you classify by pressure or temperature?
A: I use the Cloud Top Temperature.
*Dr.Teruyuki Nakajima recommends having the cloud top pressure products from ATSK 3.
Q: (Dr. Nakajima) Do you analyze all pixels in a segment box?
A: Yes.

Back to index

3. ATSK 3e and ATSK 4,Dr. Zhang, EORC/GAIT

For both ASTK 3e and ATSK4, Dr. Zhang introduced the algorithms, product definition and input/output files. He also explained the algorithm implementations with EORC by showing the directory structure and performance comparison. In addition, Dr. Zhang provided algorithm implementation schedule for ATSK 4.
Questions and Answers
Q: (Dr. Teruyuki Nakajima) How do you test ATSK 3e?
A: With GSD (if possible, calculate from simulation), AVHRR, OCTS data, and MODIS data.

Q: ( Mr.Takashi Nakajima ) Do you have a chart of cloud phase bit?
A: Yes, but not on a transparent sheet to show you right now.

End of Session

Back to index