Frequency of Data Processing Updates

AIRS processing software is updated every three to five years and the entire mission set is reprocessed using the new software. The Level 1B radiance processing software updates occur less frequently than the Level 2 and Level 3 retrieval software updates.

Current Processing Version

Please note: AIRS Version 7 has been released to the public, but the previous Version 6 will continue to be available until January 2021.

Level 1B Data: Version 5
(Data are not processed to Version 7 because algorithms that create Level 1B are unchanged as of the Version 7 release)

Level 1C Data: Version 6.7

Level 2 Data: Version 7

Level 3 Data: Version 7

About AIRS Version 7

The AIRS Project, in conjunction with the AIRS Science Team and NASA GES DISC, have made Version 7 Level 2 data available to the public. Since the release of AIRS Version 6, significant improvements and modifications have been applied to the AIRS retrieval algorithm, especially the IR-only, which continues to provide high quality observations and retrievals after the complete loss of AMSU-A2 on September 24, 2016.

A summary of noteworthy changes from Version 6 to Version 7 is provided in the document AIRS/AMSU/HSB Version 7 Changes from Version 6.

A more detailed characterization of the differences between Version 6 and Version 7 of various AIRS core data products can be found in the Version 7 Performance Testing and Validation Report.

Highlights of Version 7 are as follows:

Removal of Bias in Water Vapor Products

Analysis of AIRS Version 6 water vapor products after its release revealed spurious bias in day-time products and large difference in the retrieval performance between day- and night-time cases. That bias has been removed in Version 7 as a result of removal of shortwave (2607~2657 cm-1) channels used in the water vapor retrieval of both profiles and column integrals due to larger errors and higher uncertainties in the radiative transfer calculations over these channels.

Improved Stochastic Cloud Clearing Neural Network First Guess

With the loss of microwave information from AMSU-A2 and several AMSU-A1 channels, one of the goals with Version 7 has been to improve the quality of infrared-only algorithm, which continues to provide high quality global retrievals with reduced dependence on microwave observations. The current Version 7 infrared-only product uses a Stochastic Cloud Clearing Neural Network (SCCNN) first guess that utilizes stable microwave channels of AMSU-A1, while its physical retrieval step only uses infrared channels and does not use microwave information. Plans are in place to transition to a pure infrared-only SCCNN first guess if there is further loss of microwave channels in the future.

Algorithm Improvements Leading to Better Temperature, Water Vapor and Ozone Products

Comparisons of temperature and water vapor vertical profiles with radiosonde and reanalysis data show an improved performance over all in Version 7. For ozone, Version 7 includes more channels sensitive to O3 in the upper troposphere and lower stratosphere in the 9.6µm O3 band and uses additional vertical basis functions to better represent the profile. The ozone retrieval also benefits from improved polar stratospheric temperatures in the Version 7 SCCNN and better quality control of the retrievals. Comparisons with OMPS and OMI both indicate an improved column ozone retrieval in Version 7.

Removal of Ambiguity in Surface Classification

The ambiguity in differentiating frozen and non-frozen surface classification in the infrared-only algorithm has been removed by using ice concentration over ocean and water equivalent accumulated snow depth over land obtained from the ancillary Global Forecasting System (GFS) files. We note that until the loss of AMSU-A2 in 2016, this information was derived from the microwave observations.