Jet Propulsion Labratory

MEETING SUMMARY - written by George Aumann, AIRS Project Scientist

The AIRS Science Team held a very informative meeting at Pasadena Oldtown Marriott Hotel from 21-23 February 2001 to discuss the status of the algorithms, the performance of the product generation software and the validation plans. The focus of the meeting was readiness of the science software for launch. Click here for the final agenda and the attendee list. Moustafa Chahine, AIRS Science Team Leader, made readiness of the software for the 17 July 2001 launch the focus for the meeting. He alerted the science team members to t the expiration of the phase C/D contracts with NASA on 31 December 2001.

Ramesh Kakar, EOS Aqua Program Scientist at NASA HQ, indicated that Jack Kaye, Code YS Science Director, is committed to

1) keep the science team funded until launch + 2 years with minimal disturbance to get the AIRS/AMSU-A and HSB data properly validated and

2) to release an AO at launch + 1.5 years to open the data analysis to peer reviewed proposals by launch + 2 years.

Funding required to insure that the instruments work properly, and that the calibration software and the level 2 algorithms are maintained is protected.

JPL will submit a proposal to NASA around July 2001 which covers all operational and data processing support functions for AIRS/AMSU-A and HSB, including science team support. Science team members need to submit statements of work to JPL, covering the period between 1/1/2002 and Launch + 2 years, to be included in the continuation proposal.

George Aumann, AIRS Project Scientist, reviewed the status of the AIRS/AMSU-A and HSB hardware at TRW and summarized the EOS Aqua Science Working Group meeting 8 February 2001 at GSFC. The EOS Aqua and the EOS Aura (previously known as Chem) orbits will be maintained relative to WRS-2 and the Cloudsat will be put into the EOS Aqua orbit trailing the Aqua by about 15 seconds. This leads to interesting science synergisms. Following the lead of the EOS Terra the EOS Aqua plans to hold a press conference at Launch + 90 days to show "first light" images. The purpose is to familiarize the public with the science objectives and capabilities of the EOS Aqua. The AIRS science team plans to submit a collection of papers at launch + 1 year for a special section in JGR. The signup sheet at present contains the lead author names for fifteen papers.

There are in principle three ways to get AIRS/AMSU-A/HSB data: From the DAAC, from JPL and from NOAA. The data from the DAAC is ultimately the only way to get the data, but the data will only be processed through level 1b and, until launch + 12 months, with not or only partially validated software. Getting the data directly from JPL, until launch + 12 months, has the advantage of using the most up-to-date calibration software, but is limited to "golden days". A "golden day" is one day of AIRS/AMSU-A/HSB data which has been processed and reprocessed to debug and optimize the calibration and retrieval processing software. The first golden day will be available at about launch + 4 months. The data from NOAA is a special thinned version of the level 1b product in support of the direct assimilation effort.

Navid Dehghani, AIRS Data System Manager at JPL, summarized the status of the software delivery to the DAAC and the development of the TDS (Team Leader Science Computing Facility Data System). Quentin Sun described data access to the TDS via a browser interface and a command line interface. (Action Item).

Dave Gregorich, AIRS calibration team member, has tested data access to the DAAC using MODIS data. The learning curve for getting usable data is fairly steep, several weeks, even with reasonable familiarity with HDF.

George Serafino,GSFC/DAAC, outline the data visualization support available to the AIRS team at the DAAC and access to the data at the DAAC. The DAAC will have copies of all user guides available on-line, as soon as the AIRS team delivers them. (Action Item). George also pointed out that the data integrity will be checked at the DAAC, but the DAAC depends on AIRS science team for still to be defined routine science quality check products to be provided by the science team. (Action Item). Interested parties are welcome to browse in http://eos/nasa.gov/imswe/come/ The 15Dec2000 simulation will be used as a sample "golden day" to allow future users to practice getting data from the DAAC.

Larrabbe Strow , UMBCD, discussed the status of the forward algorithm. The RTA for three frequency sets has been delivered to JPL. The ReferenceRTA software is expected to be available in 2 months. It produces the same result as the RTA, works with a table defined frequency set, but takes 30 minutes to calculate one spectrum. It uses the latest upgrade to the Masuda sea surface emissivity.

Mitch Goldberg , NOAA/NESDIS, discussed the use of the "Near_Realtime system" at NOAA/NESDIS for data transmission to the GCM assimilation teams, for data debug and for "first light" animation. The data system strips out 281 channels (the same as those selected by Susskind's retrieval) and transmits the clearest of each 3 x 3 AIRS footprints overlaying the AMSU-A footprint to the assimilation teams. In order to insure that he has the most up-to-date level 1b calibration software, the software release policy between JPL and NOAA has to be worked out. (Action Item).

Eric Fetzer , JPL, summarized the status of the validation effort. Due to the likely Aqua launch delay the ARM/CART etc. campaign has been delayed with little impact. There are no validation ship cruises or aircraft campaigns scheduled in support of AIRS validation. The selection by NASA HQ of the validation support teams based on the Aqua validation AO has not been announced. EuMetsat has selected two validation support sites in Toulouse, France and Garmisch, Germany. The validation kickoff meeting is next week at the EuMetsat facility in Darmstadt, Germany.

Bob Knutsen , UW, eported on the validation support effort at the University of Wisconsin in three area: State of the atmosphere based on ARM/CART data, ARM site validation, and land surface emissivity. Sample atmospheric spectra can be found on ftp://tyler.ssec.wisc.edu/pub/outgoing/airs/.

Regular team exercises using one full day of simulated data are used to test the accuracy of the retrieval software, and the flow of the data through JPL and to the science team at their institutions.

Team Exercise Reports.

Mike Gunson , JPL, presented an overview of the 20january01 team exercise. It used the global meteorological conditions given by the aviation forecast for 15dec00. Evan Fishbein , JPL, presented details of the simulation of the infrared and microwave channels, which combines the aviation forecast with various other files to create L2 product "truth" files. The truth files are then converted to level 1b using the RTA and adding detector noise. It takes about 2 hours to create one day's worth of Level 2 data, one hour for each granule (6 minutes) for Level 1b production using a single R20000 CPU. The simulation is very sophisticated with multiple cloud layers and variable surface emissivity based on an NDVI model on land and the Masuda model over ocean. The temperature in the 3 x 3 AIRS footprints varies. Action item to include the standard deviation of the temperature into the truth error field for diagnostics. The ice/water transition occurs at 245K. The simulation uses 363 ppm CO2 concentration (with north/south variability based on climatology). Mark Hofstadter , JPL, presented details of the VIS simulation.

The purpose of the AIRS visible channels is to diagnose the performance of the cloud clearing methodology using the combination of microwave and infrared data. Identification of "cloud free" (less than 2% clouds) AIRS 15km footprints is the one step in the diagnostics. Mark Hofstadter , JPL, presented results with the cloud-identification algorithm using the simulated data. Detection of as little as 1.6% high clouds over ocean is easy, detection of low clouds over land is much harder. However, the visible simulation is not sophisticated enough to give a reliable performance measure. Catherine Gautier , UCSB, presented the results of using the AIRS cloud identification algorithm on MODIS Terra data. The algorithm detected all clouds detected by careful visual inspection of the false color images. High cirrus can be distinguished from low clouds by including one of the AIRS 11 micron window channels.

The first step in analyzing the level 1b data is to check the radiometric quality. Denise Hagan , JPL, presented a method which uses a histogram of the difference between the SST given by the weekly Reynolds product and the SST inferred from the AIRS 2616cm-1 super window channel after TOA correction (about 0.2K). This method allows detection of a radiometric bias a low as 1K, independent of knowledge of the cloud amount in the FOV. Hagan used the same method to assess the assess the quality of the cloud-cleared product and the level 2 SST. While cloud-clearing represents a very large improvement over the not-cloud-cleared data, the presence of the long cold tails in the histograms indicates that the current algorithm needs more work. Larrabbee Strow , UMBC, plans to assess the radiometric and spectral calibration quality of the level 1b data by evaluating the statistical properties of (AIRS_observed - calculated) radiances for "clear" fields, using the ECMWF forecast for the calculation. This requires the availability of a stand-alone "clear" FOV algorithm. (Action Item)

Phil Rosenkranz , MIT, presented retrieval results using the microwave only. Surface temperature error was 4K rms, which requires further evaluation.

The direct assimilation of radiances into the weather forecast GCM at present is limited to the assimilation of "cloud-free" footprints, less than 10% of the data. Mitch Goldberg , NOAA/NESDIS, presented the results with an algorithm which predicts the temperature of the 2390.9cm-1 channel (weighting function peak at 900mb) as a linear combination of AMSU-A4, AMSU-A5 and AMSU-A6 channels, plus a sun angle and scan angle term. If the predicted temperature agrees with the observed within 2K, then the footprint is declared clear. The results look very encouraging. The regression used to obtain the initial solution for T(p), q(p) is trained on "clear" data and has an rms error of less than 0.6K for clear data. The quick start method for training the regression in-orbit uses one day of AIRS/AMSU-A data (f"clear" footprints only) using the ECMWF forecast as truth. This procedure eliminates the need for explicit bias correction or knowing the noise covariance matrix.

Joel Susskind, GSFC, has a "clear" FOV identifier which is a direct byproduct of the cloud-clearing algorithm. If cloud-clearing parameter alpha<0.1, then the FOV is declared to be clear. For the simulated data this occurred in 8% of the cases, in which case the rms T(p) retrieval error in the tropopause was 0.6K. However, during analysis of the overall performance of the T(p), q(p) retrieval accuracy, Susskind noted that the current cloud-clearing algorithm puts about 0.5K to 1K cold bias into the radiances. This is the consistent with Hagan's independent analysis using SST, and needs to be fixed.

Mike Gunson , JPL, summarized the status of the level 2 software and the results of the team exercise. The results are very encouraging, but more work is needed, particularly in the area of cloud clearing and "clear" identification. Both topics will be the focus of the next team exercise in May 2001. The issues arrising at each meeting are collected into an action item list. The project and the science team have been successful in getting action items resolved. Members of the AIRS Science team and the AIRS project support team should check the current status of the action items.

The next AIRS data assimilation workshop will be held at the World Weather Building Room 707, Camp Springs MD on 16,17 May 2001. This will be followed on 18 May 2001 by a calibration team meeting to evaluate the preflight status of the AIRS radiometric and spectral calibration. The next AIRS Science team meeting will be on 19, 20, 21 June 2001 in Pasadena, CA.