GRIB – WMO General Regularly-distributed Information in Binary form
Driver short name
Driver built-in by default
This driver is built-in by default
GDAL supports GRIB1 (reading) and GRIB2 (reading and writing) format raster data, with support for common coordinate system, georeferencing and other metadata. GRIB format is commonly used for distribution of Meteorological information, and is propagated by the World Meteorological Organization.
The GDAL GRIB driver is based on a modified version of the degrib application which is written primarily by Arthur Taylor of NOAA NWS NDFD (MDL). The degrib application (and the GDAL GRIB driver) are built on the g2clib grib decoding library written primarily by John Huddleston of NOAA NWS NCEP.
GRIB2 files without projection on lon/lat grids have the peculiarity
of using longitudes in the [0,360] range and wrapping around the
antimeridian as opposed to the usual wrapping around the prime meridian
of other raster datasets. Starting with GDAL 3.4.0, when reading such
files, a transparent conversion of the longitudes to [-180,180] will be
performed and the data will be rewrapped around the prime meridian -
the split&swap mode. This behavior can be disabled by setting the
GRIB_ADJUST_LONGITUDE_RANGE configuration option to NO.
There are several encoding schemes for raster data in GRIB format. Most common ones should be supported including PNG encoding. JPEG2000 encoded GRIB files will generally be supported if GDAL is also built with JPEG2000 support via one of the GDAL JPEG2000 drivers.
GRIB files may a be represented in GDAL as having many bands, with some sets of bands representing a time sequence. GRIB bands are represented as Float64 (double precision floating point) regardless of the actual values. GRIB metadata is captured as per-band metadata and used to set band descriptions, similar to this:
Description = 100000[Pa] ISBL="Isobaric surface" GRIB_UNIT=[gpm] GRIB_COMMENT=Geopotential height [gpm] GRIB_ELEMENT=HGT GRIB_SHORT_NAME=100000-ISBL GRIB_REF_TIME= 1201100400 sec UTC GRIB_VALID_TIME= 1201104000 sec UTC GRIB_FORECAST_SECONDS=3600 sec
GRIB2 files may also include an extract of other metadata, such as the identification section, product definition template number (GRIB_PDS_PDTN, octet 8-9), and the product definition template values (GRIB_PDS_TEMPLATE_NUMBERS, octet 10+) as metadata like this:
GRIB_DISCIPLINE=0(Meteorological) GRIB_IDS=CENTER=7(US-NCEP) SUBCENTER=0 MASTER_TABLE=8 LOCAL_TABLE=1 SIGNF_REF_TIME=1(Start_of_Forecast) REF_TIME=2017-10-20T06:00:00Z PROD_STATUS=0(Operational) TYPE=1(Forecast) GRIB_PDS_PDTN=32 GRIB_PDS_TEMPLATE_NUMBERS=5 7 2 0 0 0 0 0 1 0 0 0 0 1 0 31 1 29 67 140 2 0 0 238 217 GRIB_PDS_TEMPLATE_ASSEMBLED_VALUES=5 7 2 0 0 0 0 1 0 1 31 285 17292 2 61145
GRIB_DISCIPLINE was added in GDAL 2.3.0 and is the Discipline field of the Section 0 of the message.
GRIB_IDS was added in GDAL 2.3.0 and is the identification section / Section 1 of the message.
GRIB_PDS_TEMPLATE_ASSEMBLED_VALUES was added in GDAL 2.3.0, and use template definitions to assemble several bytes that make a template item into a 16 or 32 bit signed/unsigned integers, whereas GRIB_PDS_TEMPLATE_NUMBERS expose raw bytes
This driver supports georeferencing
This driver supports virtual I/O operations (/vsimem/, etc.)
This paragraph lists the configuration options that can be set to alter the default behavior of the GRIB driver.
GRIB_NORMALIZE_UNITS=YES/NO : Default to YES. Can be set to NO to avoid gdal to normalize units to metric. By default (GRIB_NORMALIZE_UNITS=YES), temperatures are reported in degree Celsius (°C). With GRIB_NORMALIZE_UNITS=NO, they are reported in degree Kelvin (°K).
USE_IDX=YES/NO: (From GDAL 3.4) Enable automatic reading of external wgrib2 external index files when available. GDAL will look for a <GRIB>.idx in the same place as the dataset. These files when combined with careful usage of the API or the CLI tools allow a GRIBv2 file to be opened without reading all the bands. In particular, this allows an orders of magnitude faster extraction of select bands from large GRIBv2 files on remote storage (like NOMADS on AWS S3). In order to avoid unnecessary I/O only the text description of the bands should be accessed as accessing the metadata will require loading of the band header. gdal_translate is supported but gdalinfo is not. Default is YES.
GRIB2 write support
GRIB2 write support is available since GDAL 2.3.0, through the CreateCopy() / gdal_translate interface.
Each band of the input dataset is translated as a GRIB2 message, and all of them are concatenated in a single file, conforming to the usual practice.
The input dataset must be georeferenced, and the supported projections are: Geographic Longitude/Latitude, Mercator 1SP/2SP, Transverse Mercator, Polar Stereographic, Lambert Conformal Conic 1SP/2SP, Albers Conic Equal Area and Lambert Azimuthal Equal Area.
A number of creation options are available as detailed in below sections. Those creation options are valid for all bands. But it is possible to override those global settings in a per-band way, by defining creation options that use the same key and are prefixed by BAND_X_ where X is the band number between 1 and the total number of bands. For example BAND_1_PDS_PDTN
Product identification and definition
Users are strongly advised to provide necessary information to appropriately fill the Section 0 / “Indicator”, Section 1 / “Identification section” and Section 4 / “Product definition section” with the following creation options. Otherwise, GDAL will fill with default values, but readers might have trouble exploiting GRIB2 datasets generating with those defaults.
DISCIPLINE=integer: sets the Discipline field of Section 0. Valid values are given by Table 0.0:
0: Meteorological Products. Default value
1: Hydrological Products
2: Land Surface Products
3, 4: Space Products
10: Oceanographic Product
IDS=string. String with different elements to fill the fields of the Section 1 / Identification section. The value of that string will typically be retrieved from the GRIB_IDS metadata item of an existing GRIB product. For example “IDS=CENTER=7(US-NCEP) SUBCENTER=0 MASTER_TABLE=8 SIGNF_REF_TIME=1(Start_of_Forecast) REF_TIME=2017-10-20T06:00:00Z PROD_STATUS=0(Operational) TYPE=1(Forecast)”. More formally, the format of the string is a list of KEY=VALUE items, with space separator. The accepted keys are CENTER, SUBCENTER, MASTER_TABLE, SIGNF_REF_TIME, REF_TIME, PROD_STATUS and TYPE. Only the numerical part of the value is taken into account (the precision between parenthesis will be ignored). It is possible to use both this IDS creation option and a specific IDS_xxx creation option that will override the potential corresponding xxx key of IDS. For example with the previous example, if both “IDS=CENTER=7(US-NCEP)…” and “IDS_CENTER=8” are define, the actual value used with be 8.
IDS_CENTER=integer. Identification of originating/generating center, according to Table 0. Defaults to 255/Missing
IDS_SUBCENTER=integer. Identification of originating/generating center, according to Table C. Defaults to 65535/Missing
IDS_MASTER_TABLE=integer. GRIB master tables version number, according to Table 1.0. Defaults to 2
IDS_SIGNF_REF_TIME=integer. Significance of reference time, according to Table 1.2. Defaults to 0/Analysis
IDS_REF_TIME=datetime as YYYY-MM-DD[THH:MM:SSZ]. Reference time. Defaults to 1970-01-01T00:00:00Z
IDS_PROD_STATUS=integer. Production status of processed data, according to Table 1.3. Defaults to 255/Missing
IDS_TYPE=integer. Type of processed data, according to Table 1.4. Defaults to 255/Missing
PDS_PDTN=integer. Product definition template number, according to Table 4.0. Defaults to 0/Analysis or forecast at a horizontal level or in a horizontal layer at a point in time. If this default template number is used, and none of PDS_TEMPLATE_NUMBERS or PDS_TEMPLATE_ASSEMBLED_VALUES is specified, then a default template definition is also used, with most fields set to Missing.
PDS_TEMPLATE_NUMBERS=string. Product definition template raw numbers. This is a list of byte values (between 0 and 255 each), space separated. The number of values and their semantics depends on the template number specified by PDS_PDTN, and you have to consult the template structures pointed by Table 4.0. It might be easier to use the GRIB_PDS_TEMPLATE_NUMBERS reported by existing GRIB2 products as the value for this item. If the template structure is known by the reading side of the driver, an effort to validate the number of template numbers against the template structure is made (with warnings if more elements than needed are specified, and error if less are specified). It is also possible to define a template that is not or partially implemented by the reading side of the driver.
PDS_TEMPLATE_ASSEMBLED_VALUES=string. Product definition template assembled values. This is a list of values (with the range of signed/unsigned 1, 2 or 4-byte wide integers, depending on the item), space separated. The number of values and their semantics depends on the template number specified by PDS_PDTN, and you have to consult the template structures pointed by Table 4.0. It might be easier to use the GRIB_PDS_TEMPLATE_ASSEMBLED_VALUES reported by existing GRIB2 products as the value for this item. PDS_TEMPLATE_NUMBERS and PDS_TEMPLATE_ASSEMBLED_VALUES are exclusive. To use this creation option, the template structure must be known by the reading side of the driver.
In GRIB2, a number of data encoding schemes exist (see Section 5 / “Data representation section”). By default, GDAL will select an appropriate data encoding that will preserve the range of input data. with the DATA_ENCODING, NBITS, DECIMAL_SCALE_FACTOR, JPEG200_DRIVER, COMPRESSION_RATIO and SPATIAL_DIFFERENCING_ORDER creation options.
Users can override those defaults with the following creation options are:
DATA_ENCODING=AUTO / SIMPLE_PACKING / COMPLEX_PACKING / IEEE_FLOATING_POINT / PNG / JPEG2000: Choice of the Data representation template number. Defaults to AUTO.
In AUTO mode, COMPLEX_PACKING is selected if input band has a nodata value. Otherwise if input band datatype is Float32 or Float64, IEEE_FLOATING_POINT is selected. Otherwise SIMPLE_PACKING is selected.
SIMPLE_PACKING: use integer representation internally, with offset and decimal and/or binary scaling. So can be used for any datatype.
COMPLEX_PACKING: evolution of SIMPLE_PACKING with nodata handling. By default, a non-spatial differencing encoding is used, but if SPATIAL_DIFFERENCING_ORDER=1 or 2, complex packing with spatial differencing is used
IEEE_FLOATING_POINT: store values as IEEE-754 single or double precision numbers.
PNG: uses the same preparation steps as SIMPLE_PACKING but with PNG encoding of the integer values.
JPEG2000: uses the same preparation steps as SIMPLE_PACKING but with JPEG2000 encoding of the integer values.
NBITS=integer between 1 to 31. Bit width for each sample value. Might be only loosely honored by some DATA_ENCODING. If not specified, the bit width is computed automatically from the range of input values for integral data types, or default to 8 for Float32/Float64.
DECIMAL_SCALE_FACTOR=integer_value. Input values are multiplied by 10^DECIMAL_SCALE_FACTOR before integer encoding (and automatically divised by this value at decoding, so this only affect precision). For example, if the type of the data is a temperature, with floating point data type, DECIMAL_SCALE_FACTOR=1 can be used to specify that the data has a precision of 1/10 of degree. The default is 0 (no premultiplication)
SPATIAL_DIFFERENCING_ORDER=0/1/2. Only used for COMPLEX_PACKING. Defines the order of the spatial differencing. 0 means that the values are encoded independently, 1 means that the difference of consecutive values is encoded and 2 means that the difference of the difference of consecutive values is encoded. Defaults to 0
COMPRESSION_RATIO=integer_value between 1 and 100. Defaults to 1 for lossless JPEG2000 encoding. Only used for JPEG2000 encoding. If a value greater than 1 is specified, lossy JPEG2000 compression is used. The value indicates the desired compression factor with respected to uncompressed data. For example a value of 10 means that the desired JPEG2000 codestream should be 10 times smaller than the corresponding uncompressed file (with NBITS bits per pixel).
JPEG2000_DRIVER=JP2KAK/JP2OPENJPEG/JPEG2000/JP2ECW (possible values depend on the actually available JPEG2000 driver in the GDAL build). To specify which JPEG2000 driver should be used. If not specified, drivers are searched in the order given in the enumeration.
Internally GRIB stores values in the units of the international system (ie Metric system). So temperatures must be stored as Kelvin degrees. But on the reading side of the driver, fields with temperatures are exposed in Celsius degrees (unless the GRIB_NORMALIZE_UNITS configuration option is set to NO). For consistency, the writing side of the driver also assumed that temperature (detected if the first value of a product definition template, ie the Parameter category is 0=Temperature) values in the input dataset will be in Celsius degrees, and will automatically offset them to Kelvin degrees. It is possible to control that behavior by setting the INPUT_UNIT creation option to C (for Celsius) or K (for Kelvin). The default is C.
GRIB2 to GRIB2 conversions
If GRIB2 to GRIB2 translation is done with gdal_translate (or CreateCopy()), the GRIB_DISCIPLINE, GRIB_IDS, GRIB_PDS_PDTN and GRIB_PDS_TEMPLATE_NUMBERS metadata items of the bands of the source dataset are used by default (unless creation options override them).
DECIMAL_SCALE_FACTOR and NBITS will also be attempted to be retrieved from the GRIB special metadata domain.
gdal_translate in.tif out.grb2 -of GRIB \ -co "IDS=CENTER=8(US-NWSTG) SIGNF_REF_TIME=1(Start_of_Forecast) REF_TIME=2008-02-21T17:00:00Z PROD_STATUS=0(Operational) TYPE=1(Forecast)" \ -co "PDS_PDTN=8" \ -co "PDS_TEMPLATE_ASSEMBLED_VALUES=0 5 2 0 0 255 255 1 43 1 0 0 255 -1 -2147483647 2008 2 23 12 0 0 1 0 3 255 1 12 1 0"
Support for GRIB2 write capabilities has been funded by Meteorological Service of Canada.