A description of granule re-processing
Due to various reasons some of the data granules for MERRA products have been re-processed. When this occurs the version of the data granule name is updated with an increment of 1. Below are a list of cases in which data has been re-processed and the reason for the re-process. The information is also contained in the metadata of the data granule.
- ALL Standard Products EXCEPT MAT1NXOCN (tavg1_2d_ocn_Nx): Earlier in 2010, a compiler was upgraded on our supercomputing platform. Code modifications implemented at the time of the compiler upgrade introduced a problem into the post processing program that compresses and prepares data files for transfer. The data was reprocessed for June, July and August, including the monthly May, June, July and August data. An alert went out to the users and onto the web interface in case the data was already downloaded. For those that may have downloaded data for the period June 1 - August 31, 2010, the only variables that were confirmed to be corrupted are the roughness lengths for momentum and heat. The new data that was reprocessed has data files that begin with MERRA301* and these are the files GMAO recommends.
- MAT3FECHM (tavg3_3d_chm_Fe): for date 02/22/1992 the granule was replaced and renamed to MERRA201.prod.assim.tavg3_3d_chm_Fe.19990222.hdf due to the original file begin corrupt. The file was reprocessed on 10/21/2010.
- MAT1NXRAD (tavg1_2d_rad_Nx): The data files for the tavg1_2d_Nx_rad data product are currently being replaced so you may experience some difficulty with accessing the data files. The new files will be reversioned to 101, 201 and 301 to record the change. The initial version of this file had incorrect long_name metadata for SWTNT, SWTNTCLR, SWTNTCLN, SWTNTCLRCLN. This corrected file was generated on 04/06/2011 and affects the files up to Jan 1, 2010.
- MAT1NXLND (tavg1_2d_lnd_Nx):The SPSNOW variable in the tavg1_2d_lnd_Nx product had all values set to zero. This was in the daily files, the monthly files, and the diurnal mean files. The range of months containing this error will be:
1979/01 - 1983/12, 1989/01 - 1993/11, 1998/01 - 2001/09
MAT1NXFLX (tavg1_2d_flx_Nx) : The FRSEAICE variable had been left out of the tavg1_2d_flx product during the first initial processing. GMAO fixed it initially but somehow one file right on the transition day of the fix got missed. That was 08/12/1980. They caught the file and corrected it but because it was after initial delivery the versioning on the file was changed to identify a new granule. Please use the data file MERRA101.prod.assim.tavg1_2d_flx_Nx.19800812.hdf.
MST1NXMLD (tavg1_2d_mld_Nx), MSTMNXMLD (tavgM_2d_mld_Nx), and MSTUNXMLD (tavgU_2d_mld_Nx):
The MERRA-Land data products distributed for the months of January, February, and March 2012 had improper missing values included in the data arrays. The metadata indicates missing_value should be 1.e+15. Data points that should have been flagged as missing were assigned the value of -9999 in the hourly files which resulted in erroneous values in the monthly averages. Affected variables include the surface temperature components (TPSNOW, TUNST, TSAT, TWLT) and the corresponding area fractions (FRSNO, FRUNST, FRSAT, FRWLT).
The data have been corrected and the new data file names begin with "MERRA301" instead of "MERRA300". Files with the correct information after March 2012 will resume with the file name beginning with "MERRA300".