TTAA 56121 72662 99892 12262 27505 00047 ///// ///// 92717 ///// ///// 85430 11263 32026 70018 00058 32553 50563 13557 32078 40729 25560 33077 30931 41356 34104 25052 51159 35120 20195 58557 35127 15374 59362 33588 10628 59368 34058 88181 63356 34622 77207 35129 40716 31313 58208 81109 51515 10164 00009 10194 32034 32054=

TAC: Traditional Alphanumeric Codes • Developed for the teletype era but still in use • “TEMP” code used for radiosonde data

• Each sounding has four parts (TTAA, TTBB, TTCC, TTDD) • Parts A and C—mandatory levels below/above 100 hPa

• Pressure, height, temperature, dewpoint depression, winds

• Parts B and D—significant levels, below/above 100 hPa • Pressure, temperature, dewpoint depression

• Precision limited and variable

• “PILOT” code used for winds (pilot balloons, etc.) • Parts A, B, C, D like TEMP (PPAA, PPBB, PPCC, PPDD) • Parts B and D have geopotential height, winds • Used in some countries (e.g., U.S.) for significant level

radiosonde winds

Shortcomings of TAC: • Knowledge of arcane “rules” is required to read reports. • Changes such as increasing precision and adding new

quantities are impossible to make. • Metadata (latitude, longitude, elevation, etc.) are typically

not included, requiring an external station list. • Four to six files are required to reconstitute a sounding. • If significant level winds are provided in PILOT rather than

TEMP messages, the geopotential heights need to be converted to pressure, which is a significant source of error.

043930600372600 12509999999011000041010201033122640017101100 40340025AA A015030000030340024AA A03039-071160340033AA A0558 0-211999320048AA A07180-351130320060AA A09 40-461999310061AA A10340-311140300061AA A11790-521999310047AA A13640-54199931 0046AA A16220-591999300023AA A18470-591999290028CA A20590-59 1999280017 Q FXXXXXX0206118270102000120040VA 093100050020 A 087000020050 08000-041000 07650-041150 C 07110-061160 A 05720-191150 A 05350-191150 A 04000-351130 A 03540-401120 A 02260-541999 A 01150-561999 A 01020-591999 A 00910-561999 A 00780-591999 A 00560-611999 A 00400-561999 A 00380-551999 C 050670204402260-541999300056T 00800-599999280025T XXXXXX 040942026000171340022W 00305330027 00610340023 00914340026 01219340027 01829340021 02134340027 02438340029 027433 40030 03658330029 04267330038 04877320047 06096310059 0 7620310066 09144310061 10668300061 15240300026 164593000 24 19202290020 21031270018 081020707000136105A 00133105B 00163105C 00163105D 18690107ZB05057108BT18550108DTEND REPORT

BUFR: Binary Universal Form for the Representation of meteorological data • First approved for use by the WMO in 1988 • Binary—files are compact • Universal—one decoder used for many data types • Self-describing—through use of “data descriptors” • Expandable—easy to add variables, precision • Flexible—easy to vary content • Sustainable—decoders backward compatible • WMO has approved “templates” for specific data types (e.g., radiosonde data) with lists of defined descriptors to make it easier to encode data

• Precision is improved over TAC. • Includes descriptors for station metadata so an

external station list is no longer necessary. • Thousands of vertical levels can be included. • Radiosonde drift time, lat, lon offsets are included. • Two messages per sounding are intended—one when

the balloon reaches 100 hPa, the second with the full sounding from the surface to balloon burst.

Shortcomings of BUFR: • Not human readable • Requires considerable expertise to implement

Assimilating Real-Time High-Resolution BUFR Radiosonde Data in NAVGEM: Current Results and Future Prospects

Patricia Pauley and Daniel Tyndall Marine Meteorology Division, Naval Research Laboratory, Monterey CA

Justin Reeves and Randal Pauley Fleet Numerical Meteorology and Oceanography Center, Monterey CA

The TAC to BUFR Migration • In 2003, the WMO members approved a migration from TAC to BUFR for data distribution on the Global

Telecommunications System (GTS), the primary communications pathway countries use to share data. • Data were to be encoded and distributed in both TAC and BUFR for a limited time to enable users to

transition between the two formats. • The official timetable called for TAC distribution to cease in November 2014, however not all countries have

started producing BUFR and only a small fraction of TAC has actually stopped.

Why use BUFR for radiosonde data? Characteristics of BUFR radiosonde data BUFR structure issues • The final BUFR message from a given station

should include the whole sounding (purple), according to WMO rules.

• Some countries put each TEMP part in a separate message—”BUFR by parts” (red). Many NWP centers will not use these.

• Other countries use separate messages for the sounding below and above 100 hPa (cyan).

• A few countries only encode data below 100 hPa (green).

“Native” vs. reformatted BUFR • “Native” BUFR data use data from the radiosonde

ground-station and include balloon drift information. • High-density native BUFR data have hundreds to

thousands of levels (blue). • Low-density native BUFR data have a couple hundred

levels (red).

• Reformatted BUFR use data decoded from TAC bulletins (green) and so do not have drift data. • Reformatted BUFR have the same number of levels as

TAC (~60-100).

Use of BUFR radiosonde data in NAVGEM • Decoded data are read and checked for errors in both data and metadata.

• Data errors • Climatological/physical limits checks • Check for known errors

• Pressure divided by 10 for significant level winds above 100 hPa from some stations in India • Significant level wind speeds divided by 2 from Viet Nam, etc.

• Metadata errors (compared to FNMOC station list) • Errors converting deg-min-sec to decimal degrees (e.g., Yuma, AZ (32° 50’ converted to 30.50° instead of 30.86°)

• Sign errors in longitude (e.g., Shemya, AK, is degrees W instead of E)

• Use of Hp (barometer height) vs. Ha (field elevation) • In the U.S., pressure at Hp is adjusted to the elevation Ha, so Ha should be used as the station elevation for U.S. stations even though Hp

is usually used as the station elevation by NWP centers. • e.g., Kodiak, AK, has Hp = 33.8 m and Ha = 5.5 m—reformatted BUFR uses 34 m as the station height, but 6 m should be used since the

surface pressure is adjusted to the radiosonde release point

• Errors in station list used for reformatted BUFR (e.g., Rapid City, SD, located 60 km too far north)

• TAC and BUFR data are duplicate-checked separately, then duplicate-checked again after TAC and BUFR data are consolidated. All TAC data are rejected when native BUFR are available.

• The surface elevation is estimated hydrostatically to aid in metadata reconciliation when BUFR values differ from the values in the FNMOC station list.

• High-resolution data are thinned to approximately 200 levels per sounding. • The combined dataset is then subjected to quality control and assimilated in NAVGEM in the same way TAC

data were processed previously, using NCEP’s Complex Quality Control. • BUFR radiosonde data have been used operationally since 23 Sept 2015 in NAVGEM’s 4DVAR (NAVDAS-AR). • Adding BUFR data increased the ob count (counting T, q, u, and v as separate obs) by ~30k at 00Z and 12Z. • Incorporation of balloon drift is planned for NAVGEM.

• Balloon drift will be used from reports where available and will be estimated from reported winds otherwise. • Example shows how reported drifts differ geographically for stations in southern Canada for data from 00Z 17 March

2015. (U.S. stations do not show drift since it is not reported.)

Shemya (WMO) Shemya (BUFR) The BUFR location has

a sign error in lon —a 790 km error

Future work with BUFR radiosonde data High resolution real-time BUFR capability from NWS stations is currently being tested and should be deployed during spring to late summer 2017 (Aaron Poyer, Observation Systems Test Director, NWS). • High-resolution data currently goes only to NCEI some

hours post-launch using a legacy BUFR template. • The NWS initiative will provide real-time native BUFR data

using the standard WMO template and will include drift. • Example shows data from a test launch from Gaylord, MI.

Test sites include Pittsburgh, PA; Amarillo, TX; Gaylord, MI; Reno, NV; and Fairbanks, AK. Shreveport, LA, is scheduled to be added in January.

• 10 of the 92 NWS radiosonde sites are not included in this software upgrade, since they use a 403 MHz COTS system that was recently installed to mitigate RF interference with GOES-16.

APPROVED FOR PUBLIC RELEASE, DISTRIBUTION IS UNLIMITED The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Department of the Navy, Department of Defense, nor the U.S. Government.

00Z 20 Nov 2016

00Z 20 Nov 2014

00Z 20 Nov 2015

00Z 20 Nov 2014

00Z 20 Nov 2015 00Z 20 Nov 2016

00Z 20 Nov 2014

00Z 20 Nov 2015 00Z 20 Nov 2016

BUFR data issues • Errors are present in the BUFR lat/lon (red) or elevation (gold), primarily in reformatted BUFR where an

external station list is needed. • Some countries do not include the significant level winds with the rest of the sounding (purple). • Incomplete BUFR soundings can be filled in with TAC data (blue).

Reported radiosonde drift color-coded by pressure (hPa)

References Ingleby, B., P. Pauley, A. Kats, J. Ator, D. Keyser, A. Doerenbecher, E. Fucile, J.

Hasegawa, E. Toyoda, T. Kleinert, W. Qu, J. St James, W. Tennant, and R. Weedon, 2016: Progress towards high-resolution, real-time radiosonde reports, 97(11). Bulletin of the American Meteorological Society. (http://dx.doi.org/10.1175/BAMS-D-15-00169.1)

ECMWF, 2016: TAC to BUFR Migration Wiki. (https://software.ecmwf.int/wiki/display/TCBUF/TAC+To+BUFR+Migration#app-switcher)

Observational Needs As seen in the FSOI (Forecast System Observation Impact—Langland and Baker, 2004) figure above, radiosonde data are a critical observing system for NWP. While satellite data (radiances, satellite-derived winds, and GPS-RO) currently represent about 90% of the observations assimilated in NAVGEM, the previous figure shows that conventional observations (radiosonde, aircraft, and surface data) are responsible for 31.4% of the error reduction. NWP has an ongoing need for these operational conventional observations, not only in the U.S. but worldwide.

0.1%

1.7%

2.5%

3.0%

5.8%

7.4%

10.9%

13.1%

17.4%

17.6%

20.5%

-1.25 -1 -0.75 -0.5 -0.25 0Average FSOI by Instrument Group per dtg (J/kg)

IR Radiances

Geo Winds

MW Radiances

Radiosonde

Surface

Aircraft

Satellite TPW

GPS-RO

Polar Winds

Sat Sfc Winds

TC Synth

Hybrid opnl--12 Nov-12 Dec 2016 Total FSOI = -5.3 J/kg

Radiosonde data provide 13% of the error reduction for 12 Nov to 12 Dec 2016 in the NAVGEM 1.4 operational run. This version became operational on 12 Oct 2016 and upgraded to Hybrid 4DVAR using ensemble background error covariances.