Although I'm still in the early stages of comparing BUFR radiosonde data with TAC data in the context of the U.S. Navy's global data assimilation system, I have found several significant errors that I wanted to share with the community.  I have made an effort to report these to relevant agencies so that the errors can be corrected, but these errors are still present in current data as of 25 June 2014.  The maps below depict the current coverage of BUFR radiosonde data decoded at FNMOC, with soundings reporting in the WMO standard format labeled as "glbl" and indicated by blue solid circles and with soundings reported in parts in the KWBC alternate format (described below) labeled as "kwbc" and indicated by red open circles.  Thanks to Cary McGregor at FNMOC for these coverage diagrams.

Errors found in BUFR radiosonde data as of 25 June 2014:

• Negative dew point depressions (temperature minus dew point temperature) on some significant wind levels in BUFR data from France and French overseas departments and collectivities (WMO block 07, plus stations such as 78897 (Le Raizet, Guadeloupe), 81405 (Cayenne, French Guiana), 91592 (Noumea, New Caledonia).

• Negative dew point depressions on various level types in BUFR data from India with values of -70 C or even lower (WMO blocks 42 and 43). It appears that these dew point depressions are actually dew point temperatures in degrees C.
• Miscoded variables for tropopause levels in BUFR data from India (WMO blocks 42 and 43).  It appears that temperature is placed in the height variable and dew point temperature in the temperature variable.  The wind speeds agree with the TAC data, but the wind directions are missing on these levels.

NWSTG in Washington, DC, places BUFR radiosonde data on the GTS in two different formats tagged with KWBC as the originating center–the WMO standard format and another format with separate BUFR messages for each of the TEMP message parts.  Both have errors that might cause problems.  These errors have been reported to the NWSTG help desk, which has taken out trouble tickets to fix them.

• KWBC version of the WMO standard format (includes WMO blocks 70, 72, 74, 78, plus stations such as 61641 (Dakar, Senegal), 85469 (Isla De Pascua, Chile), 91165 (Lihue, Hawaii), 91212 (Guam), 91285 (Hilo, Hawaii), 91334 (Weno Island, Micronesia), 91376 (Marshall Islands), 91408 (Babelthuap Island, Palau), 91413 (Yap Island, Micronesia), 91765 (Pago Pago).  Note that these are only a subset of the stations reporting in the standard WMO format, indicated on the maps below by blue solid circles.)
  
  • Geopotential heights do not agree with the TAC heights or the heights from the KWBC alternate format (which do agree with the TAC heights).  The disparity increases with height from a few meters at 1000 hPa to a couple hundred meters at 200 hPa and approximately 500 m at 10 hPa.
  • 7.0 hPa heights are missing.
• KWBC alternate format (in parts) (See stations designated by red circles as "kwbc" on the maps below)
  
  • Launch times are placed at the end of the message with hours and minutes in separate variables.  The launch time minutes are currently placed in the launch time hours variable.
  • 7.0 hPa heights are assigned a leading 4 digit rather than a 3 in the translation from the TAC message.  For example, at 0000 UTC 25 June 2014 station 70273 reports a 7.0 hPa height of 44410m in BUFR; the FNMOC decoded TAC data gives a 7.0 hPa height of 34410.
    
    
    

Additional errors found in BUFR radiosonde data as of 11 July 2014:

• Some significant discrepancies between TAC and BUFR metadata are present for certain U.S. radiosonde stations.  The metadata supplied for TAC reports come from a master station list maintained at FNMOC, while the metadata supplied in BUFR reports come from a master station list maintained at the NOAA Telecommunication Operations Center (TOC).   The following stations had latitude or longitude discrepancies greater than 0.1 degrees, with values listed for latitude, longitude, and station elevation.  The first three appear to be a result station moves away from airport locations that had not been updated at the TOC but had been updated in the WMO radiosonde list; the fourth appears to be uncertainty associated with U.S. military operations, but the FNMOC elevation seems to be more consistent with the reported radiosonde heights.  In all four cases, the FNMOC master station list value listed as the TAC value appears to be correct; this problem was reported to the TOC help desk.
  
  • 72274 (Tucson, Arizona, USA)
    • TAC: 32.23°, 249.03°, 751m
    • BUFR: 32.11°, 249.07°, 802m
  • 72327 (Nashville, Tennessee, USA)
    • TAC: 36.25°, 273.43°, 180m
    • BUFR: 36.11°, 273.32°, 182m
  • 72662 (Rapid City, South Dakota, USA)
    • TAC: 44.07°, 256.79°, 1029m
    • BUFR: 44.05°, 256.95°, 976m
  • 74004 (Yuma Proving Ground, Arizona, USA)
    • TAC: 32.87°, 245.67°, 136m
    • BUFR: 32.50°, 246.00°, missing (231m listed online at http://weather.noaa.gov/cgi-bin/nsd_lookup.pl (discontinued service)
• Metadata discrepancies also exist for non-U.S. radiosonde stations that are more difficult to reconcile.  This will be the topic of a future post, along with discrepancies in station elevation which I have not yet examined.