BUFR Format Reference

This page provides reference information for the BUFR (Binary Universal Form for the Representation of meteorological data) format as relevant to earth2bufr.

BUFR Editions

earth2bufr supports BUFR Edition 3 and Edition 4. The key differences:

Feature	Edition 3	Edition 4
Year field	2-digit (YY)	4-digit (YYYY)
Sub-category	Local only	International + local
Seconds	Not present	Present
Section 2	Optional	Optional

BUFR Message Structure

Every BUFR message contains five (optionally six) sections:

Section	Name	Contents
0	Indicator	“BUFR” marker, total message length, edition number
1	Identification	Originating centre, data category, date/time, table versions
2	Optional	Local-use data (centre-specific)
3	Data Description	Unexpanded descriptor list, number of subsets, compression flag
4	Data	Encoded observation values
5	End	“7777” marker

Descriptor Types

BUFR descriptors are 16-bit values encoded as F-XX-YYY where F (2 bits), X (6 bits), and Y (8 bits) determine the type:

F	Type	Purpose
0	Table B (Element)	Single data element (e.g., temperature, pressure)
1	Replication	Repeat a group of descriptors N times
2	Operator	Modify encoding rules (change width, scale, etc.)
3	Table D (Sequence)	Macro expanding to a list of descriptors

Table B Elements

Table B defines individual data elements. Each entry specifies:

• Mnemonic – Short name, which names the decoded column (e.g., B012101, or TOB in NCEP DX tables)

• Units – Physical units or CCITT IA5 for strings, Code table / Flag table for categorical data

• Scale – Decimal scaling factor (value = (raw + reference) / 10^scale)

• Reference value – Offset added to the raw value before scaling

• Bit width – Number of bits used to encode the value

The full WMO Table B is defined in the WMO Manual on Codes, Vol. I.2, also available on Github.

Table D Sequences

Table D defines reusable sequences of descriptors. For example, a common sequence might expand latitude + longitude + station height into a single descriptor for convenience.

The full WMO Table D is maintained alongside Table B in the WMO Manual on Codes, available on Github.

Replication (F=1)

Replication descriptors repeat a group of descriptors:

• Fixed replication (Y > 0): Repeat the next X descriptors Y times

• Delayed replication (Y = 0): The repeat count is encoded in the data stream itself (read from the next descriptor)

Operator Descriptors (F=2)

Operator descriptors modify how subsequent elements are encoded:

Operator	Name	Effect
2-01-YYY	Change data width	Add (YYY - 128) bits to subsequent numeric widths; CCITT IA5, code, and flag tables are exempt
2-02-YYY	Change scale	Add (YYY - 128) to subsequent non-code numeric scales; code and flag tables are not rescaled
2-03-YYY	Change reference value	Following values define new reference values
2-04-YYY	Associated field	Add YYY bits of associated data before each element
2-05-YYY	Text string	Read YYY*8 bits as character data
2-06-YYY	Signify local descriptor	Next descriptor uses YYY bits
2-07-YYY	Increase scale and width	For ordinary numeric data, scale += YYY, reference = 10^YYY, and width += floor((10YYY + 2)/3); CCITT IA5, code, and flag tables are exempt
2-08-YYY	Change string width	Override CCITT IA5 string width to YYY*8 bits

WMO Data Categories

The data_category field in Section 1 identifies the observation type. Standard WMO categories (defined in WMO Common Code Table C-13 and WMO Manual on Codes, Vol. I.2):

Code	Category
0	Surface data – land
1	Surface data – sea
2	Vertical soundings (other than satellite)
3	Vertical soundings (satellite)
4	Single-level upper-air (other than satellite)
5	Single-level upper-air (satellite)
6	Radar data
7	Synoptic features
8	Physical/chemical constituents
9	Dispersal and transport
10	Radiological data
11	BUFR tables (including DX-tables)
12	Surface data – sea (waves)
13	Forecasts
14	National/bilateral arrangements (non-standard)
15-19	Reserved
20	Status information
21-30	Reserved
31	Oceanographic data

PrepBUFR Format

PrepBUFR is an NCEP-specific extension of BUFR used for quality-controlled observational data in the NCEP Global Data Assimilation System (GDAS).

Key Differences from Standard BUFR

1. Embedded DX-tables – Local Table B and Table D definitions are stored as BUFR messages with dataCategory=11 before the data messages they govern; a file can hold several DX blocks

2. NCEP-local descriptors – Uses mnemonics like SID, XOB, YOB, TOB, POB instead of standard WMO names

3. Operator 2-07 – Enhanced precision fields (e.g., longitude/latitude stored with extra scale and width)

PrepBUFR Data Categories

PrepBUFR files use NCEP-specific data category codes. The table below shows the message-level codes (values that appear in the BUFR Section 1 header and in the _data_category column) alongside the NCEP logical type codes. Use the message-level codes with data_category_filter; get_metadata(path).data_messages lists each message’s category.

Message-Level Codes (Section 1 dataCategory)

These are the values you will see in the file and pass to data_category_filter:

Code	Mnemonic	Description
102	ADPUPA	Upper-air (radiosondes, pilot balloons, dropsondes)
103	PROFLR	Wind profiler / acoustic sounder
104	AIRCFT	Conventional aircraft (AIREP, PIREP, AMDAR, TAMDAR)
105	SATWND	Satellite-derived winds
107	VADWND	VAD (NEXRAD) winds
109	ADPSFC	Surface land (synoptic, METAR, ASOS)
110	SFCSHP	Surface marine (ship, buoy, C-MAN, tide gauge)
112	GPSIPW	GPS-IPW (integrated precipitable water)
113	SYNDAT	Synthetic tropical cyclone bogus
119	RASSDA	RASS virtual temperature profiles
121	ASCATW	ASCAT scatterometer winds

NCEP Logical Type Codes

The NCEP documentation also defines a separate “logical” numbering (0-19) used internally by the PrepBUFR processing system. These do not appear in the BUFR message headers but are documented here for cross-reference with NCEP literature:

Code	Mnemonic	Description
0	ADPUPA	Upper-air (radiosondes, pilot balloons, dropsondes)
1	AIRCAR	MDCRS ACARS aircraft
2	AIRCFT	Conventional aircraft (AMDAR, etc.)
3	SATWND	Satellite-derived winds
4	PROFLR	Wind profilers (NOAA, JMA)
5	VADWND	VAD winds (NEXRAD)
6	SATEMP	TOVS/ATOVS sounding / satellite radiance
7	ADPSFC	Surface land (METAR, SYNOP, CMAN)
8	SFCSHP	Surface marine (ships, buoys)
9	SFCBOG	Surface bogus
10	SPSSMI	SSM/I retrievals
11	SYNDAT	Synthetic tropical cyclone bogus
12	ERS1DA	ERS scatterometer
13	GOESND	GOES sounder/radiance
14	QKSWND	QuikSCAT
15	MSONET	Mesonets
16	GPSIPW	GPS-derived integrated precipitable water
17	RASSDA	RASS virtual temperature
18	WDSATR	WindSat
19	ASCATW	ASCAT scatterometer winds

Tip

Use earth2bufr.get_metadata to discover which categories are present in a specific file without decoding any data:

info = earth2bufr.get_metadata("gdas.prepbufr")
print(sorted({message.data_category for message in info.data_messages}))

Sources for PrepBUFR category definitions:

• NCEPLIBS-bufr PrepBUFR table

• PrepBUFR Processing at NCEP (PDF), Table 1

• GSI User’s Guide, Chapter 3

DX-Table Message Format

The Section 4 data of a DX-table message (dataCategory=11) holds local Table A, B and D definitions as fixed-width CCITT IA5 records:

n_a (1 byte), n_a x 67-byte Table A records,
n_b (1 byte), n_b x 112-byte Table B records,
n_d (1 byte), Table D records (descriptor, mnemonic, description, member count, FXXYYY members)

• Table A: category, mnemonic and description per record – usually empty

• Table B: descriptor, mnemonic, name, units, scale, reference and width

• Table D: a sequence descriptor and its member descriptors

earth2bufr extracts these tables entirely in Rust – no external Python dependencies are needed.

Common PrepBUFR Mnemonics

Mnemonic	Description	Units
SID	Station identifier	CCITT IA5
XOB	Longitude	Degrees East
YOB	Latitude	Degrees North
DHR	Observation time minus cycle time	Hours
ELV	Station elevation	Meters
TYP	Report type	Code table
POB	Pressure observation	mb (hPa)
TOB	Temperature observation	Degrees C
QOB	Specific humidity	mg/kg
ZOB	Height observation	Meters
UOB	U-component of wind	m/s
VOB	V-component of wind	m/s
PWO	Precipitable water observation	mm
CAT	Data category	Code table

External References

WMO Standards

• WMO Manual on Codes (WMO-306) – Authoritative source for BUFR Table B, Table D, and code/flag tables

• WMO BUFR/CREX Tables (CSV) – Machine-readable Tables from WMO GitHub

• WMO Common Code Tables – Data categories, originating centres, and other common code tables

NCEP/NOAA PrepBUFR

• NCEPLIBS-bufr Tables – NCEP-maintained local BUFR tables used in PrepBUFR files

• NCEPLIBS-bufr Documentation – The library that reads/writes PrepBUFR (successor to legacy BUFRLIB)

• PrepBUFR Processing at NCEP (PDF) – Overview of the PrepBUFR observation processing system

• NNJA (NOAA-NASA Joint Archive) – Archive for GDAS PrepBUFR files

• NOMADS PrepBUFR – Real-time access to PrepBUFR from NCEP operational runs

EUMETSAT EPS-SG / MTG

• EPS-SG Test Data in BUFR Format – EUMETSAT page with links to EPS-SG and MTG test BUFR data files

• README: BUFR Data Description (Aug 2025, PDF) – EUMETSAT readme describing the BUFR data format for EPS-SG products

• EUMETSAT BUFR Vocabulary Manager – EUMETSAT vocabulary/table browser for BUFR descriptors and code tables

ECMWF BUFR

• ECMWF BUFR Software (ecCodes) – ECMWF’s BUFR/GRIB encoding/decoding library

• ECMWF BUFR Users Guide – ECMWF’s BUFR guide

• ECCodes Github BUFR Examples