INTRODUCTION

This is a package to make the transition to bufr reports from alphanumeric text easiest as possible. The intial commit is contributed from ogimet.com

A lot of software is coded assuming that the primary source of meteorological reports is in alphanumeric format. Decode libraries are expecting this. But time is changing, meteorological services are migrating to BUFR and other binary formats. A lot of decode software have to be changed.

This is a software to get meteorological reports in old alphanumeric format from BUFR files. At the moment includes the following reports:

FM 12-XIV SYNOP
FM 13-XIV SHIP
FM 14-XIV SYNOP MOBIL
FM 18-XII BUOY
FM 35-XI TEMP
FM 36-XI TEMP SHIP
FM 38-XI TEMP MOBIL
FM 71-XII CLIMAT

Note that the results from this library is not intended to match at %100 level to original alphanumeric reports. It cannot. Some variables in alphanumeric code rules can be coded in several ways, and there is not a regional even national decision about them. As example, the 'hshs' item (table code 1617) for synop FM-12 can be coded using 00-80 range or 90-99 one. A numeric value for heigh of base clouds can be coded in two ways. And there some few more examples.

Since version 0.23.0 the old support for supress the legacy code to support bufrdc ECMWF library has been supressed.

In this package, the included library bufrdeco to decode BUFR files has been completed and optimized to deal with all BUFR reports.

BUFRDECO LIBRARY

The repository includes a library to decode BUFR reports written from scratch. It is bufrdeco and is a very fast and ligth library capable to decode any bufr report since version 0.20.0 and above.

Since version 0.23.0 two interesting optional features has been added to bufrdeco

Create and use a cache of bufr tables when the library is used to decode more than one BUFR file. If the master version of every file is not the same, then the cache stores the tables used up to moment. This is a very nice optimization because most of CPU time is wasted in reading and decoding the tables before decoding a given bufr report. If the a master version is already used in the session then there is no need to read and decode their tables again.
Create small indexes for non compressed BUFR reports. When data in a BUFR file is not compressed then to get the data for a subset N it is needed to parse the sec 4 data for the subsets 0 to N-1. This is because the extension of a subset data is not known before parse it for non compressed cases. The creation of this index file save a lot of work if we are only interested in the data of a single subset.

Since version 0.24.0 bufrdeco also includes json output fetaures. It can print BUFR sections, expanded tree and expanded data results in json format. You can play with it using the binary bufrdeco_test as is explained in following sections.

BUILD

BUILD USING GNU AUTMAKE/AUTOTOOLS

If you uses git repository, first time you clone the code you need to set the autotools files. To achieve this just from bufr2synop/ directory

    make -f Makefile.cvs

Then it is recommended to build in a separate directory, say build0

    mkdir build0
    cd build0

After this first task, you have to configure and make to build the package. Also if you get the package from a tarball you must do configure and make. From bufr2synop/build0/ directory

    ../configure

To build the package

    make

BUILD USING GNU AUTOMAKE/AUTOTOOLS

Assumimg you are in root source directory bufr2synop/ you should build an install in a separate directory tree. Let assume you'll build in a new bunfr2synop/buld/ directory

    mkdir build
    cd build

Then configure using cmake

    cmake ..

To build the package

    make

WHAT WE GET AFTER A SUCCESSFULL BUILD?

Assume you built the package. you will have up to five binaries:

bufrdeco_json Binary to check and view all data from BUFR files in json format.
bufrtotac . Main binary to extract BUFR report data in Traditional Alphanumeric Code. This binary uses the own bufrdeco library and tables. It is several times faster than old software based in ECMWF library.
bufrnoaa An utility to extract and select bufr files from NOAA bin archives. This is used in Ogimet.com site to extract bufr files from NOAA GTS gateway
build_bufrdeco_tables A binary to convert BUFR table files from ECMWF and WMO to table files used by bufrdeco library. This is used by bufr2synp` developers ans the results included in the directory tables of the package. User do not need to used it.

And also the libraries :

libbufrdeco A light and very fast bufr decode library. Since version 0.20.0 it can decode any BUFR report, but just a subset of them are suited to get the target TAC reports using bufrtorac binary.
libbufr2tac Used by binary bufrtotac to transform a decoded bufr into a TAC (Traditional Alphanumeric Format). Before version 0.7 this library was named bufr2synop. It seems the new name is the right name.

Note that if you build using GNU autotools you will get both static and dynamic vesions of libraries. Using cmake you just get shared libraries.

INSTALL

See INSTALL file in this package to get details. If you use default settings, then install will install all the files in /usr/local/ directory. So you will need root privileges. As example

    sudo make install

    su -c "make install"

Because new shared libraries may be installed, is recommended to execute ldconfig command

    sudo /sbin/ldconfig

    su -c "/sbin/ldconfig"

EXAMPLES OF USE

In a working dir we want to get a BUFR archive from NOAA GTS gateway and see the reports we can extract from it. On some enviroments, do not forget to set BUFR_TABLES variable to directory containing customized bufr tables if you are using them.

Get a bufr bin file from NOAA GTS gateway. Is a cyclic directory file, with file names in the form sn.NNNN.bin, NNNN varying from 0000 to 0120. We choose first one. Don't worry, the file is also included in examples dir. (The file you can download is different for sure, the example in package is from Saturday Oct 18 2014).
```
    wget https://tgftp.nws.noaa.gov/SL.us008001/DF.bf/DC.intl/sn.0000.bin
```
Then use bufrnoaa to get a lot of files. Every file is a bufr report. We do not want all kind of bufr reports, just surface observations (land and oceanographic). We can just do
```
    bufrnoaa  -f -T SO -S ACIMNSVW -O B -i sn.0000.bin
```
If we did this with examples sn.0000.bin file we got 171 bufr files selected from 571 in the archive sn.0000.bin. Let choose one bufr file 20141018211119_ISIN03_EGRR_182100.bufr (also in examples dir)

To see the 'emulated' synop reports from this bufr file
```
    bufrtotac -i 20141018211119_ISIN03_EGRR_182100.bufr
```
We will get a list of 58 synops to stdout.

And to see CLIMAT reports, we can try with another bufr file
```
    bufrtotac -i 20150705121512_ISCD01_LIIB_050000.bufr
```
We will get 19 CLIMAT reports, some of them NIL reports.

We can try also to get a traditional TEMP report from examples. Let use the new bufrtotac
```
    bufrtotac -i 20160402121749_IUSH01_DRRN_021100.bufr
```
We will get the four parts of a TEMP report.

In case you just want to get BUFR details just add '-V -n' arguments
```
    bufrtotac -i 20150705121512_ISCD01_LIIB_050000.bufr -V -n
```
you will get a long output with all details for all subsets of descriptors, including the expanded tree. Option -n supresses the conversion to TAC.

There are some more options for bufrnoaa and bufrtotac. You can see a list using option -h

$> bufr2noaa -h
bufrnoaa -h
bufrnoaa: Version '0.24.0' built using GNU C compiler gcc 12.2.1 at Oct  3 2022 06:41:17 and cmake.
Usage: 

bufrnoaa -i input_file [-h][-v][-f][-l][-F prefix][-T T2_selection][-O selo][-S sels][-U selu]
   -h Print this help
   -v Print information about build and version
   -i Input file. Complete input path file for NOAA *.bin bufr archive file
   -2 Input file is formatted in alternative form: Headers has '#' instead of '*' marks and no sep after '7777'
   -l list the names of reports in input file
   -f Extract selected reports and write them in files, one per bufr message, as 
      example '20110601213442_ISIE06_SBBR_012100_RRB.bufr'. First field in name is input file timestamp 
      Other fields are from header
   -F prefix. Builds an archive file with the same format as NOAA one but just with selected messages
      witgh option  -T. Resulting name is 'prefix_input_filename'
      If no -F option no archive bin file is created.
      If no message is selected, the a void file is created.
      File timestamp is the same than input file
   -T T2_selection. A string with selection. A character per type (T2 code)
      'S' = Surface . 'O'= Oceanographic. 'U'= upper air
      If no -T argument then nothing is selected
   -S sels. String with selection for A1 when T2='S'
      By default all A1 are selected
   -O selo. String with selection for A1 when T2='O'
      By default all A1 are selected
   -U sels. String with selection for A1 when T2='U'
      By default all A1 are selected

$> butrtotac -h
bufrtotac -h
bufrtotac: Version '0.24.0' built using GNU C compiler gcc 12.2.1 at Oct  3 2022 10:41:51 and cmake.
Linked to bufr2tac library version '0.24.0' built using GNU C compiler gcc 12.2.1 at Oct  3 2022 06:57:27 and cmake.
Linked to bufrdeco library version '0.24.0' built using GNU C compiler gcc 12.2.1 at Oct  3 2022 06:57:25 and cmake.

Usage: 
bufrtotac -i input_file [-i input] [-I list_of_files] [-t bufrtable_dir] [-o output] [-s] [-v][-j][-x][-X][-c][-h][more optional args....]
    -c. The output is in csv format
    -D debug level. 0 = No debug, 1 = Debug, 2 = Verbose debug (default = 0)
    -E. Print expanded tree in json format
    -G. Print latitude, logitude and altitude 
    -g. Print WIGOS ID
    -h Print this help
    -i Input file. Complete input path file for bufr file
    -I list_of_files. Pathname of a file with the list of files to parse, one filename per line
    -j. The output is in json format
    -J. Output expanded subset SEC 4 data in json format
    -n. Do not try to decode to TAC, just parse BUFR report
    -o output. Pathname of output file. Default is standar output
    -R. Read bit_offsets file if exists. The path of these files is to add '.offs' to the name of input BUFR file
    -s prints a long output with explained sequence of descriptors
    -S first..last . Print only results for subsets in range first..last (First subset available is 0). Default is all subsets
    -t bufrtable_dir. Pathname of bufr tables directory. Ended with '/'
    -T. Use cache of tables to optimize execution time
    -W. Write bit_offsets file. The path of these files is to add '.offs' to the name of input BUFR file
    -V. Verbose output
    -v. Print version
    -x. The output is in xml format
    -X. Try to extract an embebed bufr in a file seraching for a first '7777' after first 'BUFR'
    -0. Prints BUFR Sec 0 information in json format
    -1. Prints BUFR Sec 1 information in json format
    -2. Prints BUFR Sec 2 information in json format
    -3. Prints BUFR Sec 3 information in json formatbufrtotac: Version '0.23.0' built using GNU C compiler gcc 12.1.1 at Aug 30 2022 18:39:58 and cmake.

Since version 0.23.0 there are four new interesting options in buftotac as you can see.

The option -g prints WIGOS Identifier if it is available in BUFR data. It prints the identifier at the begining of the report line output. See the three lines what follows. The legacy TAC report without -g would begin in column 34, just after '|' character.
First line is from a BUFR with synop data. It have both WIGOS identifier 0-20000-0-06610 and WMO five digits IIiii identifier 06610. Second line is from a BUFR with synop data with five digits WMO identifier IIiii 26645 but without WIGOS idenetifier. 0-0-0-MISSING is set as missing WIGOS identifier. The third is from a BUFR report with synop data but without five digits IIiii WMO identifier. bufrtotac prints the IIiii identifier as 00000 but the existing 0-170-0-23190700 identifier permit to get report station metadata.

0-20000-0-06610                 |202208080300 AAXX 08034 06610 26/// /2500 10129 20121 39625 40200 57002 333 55300 20000 60005 90730 91104=
0-0-0-MISSING                   |202208080300 AAXX 08031 26645 22/65 70000 10105 20098 39985 40242 52003 87030 333 60002=
0-170-0-23190700                |202208080300 AAXX 08031 00000 02/// /3301 10205 29095 39084 333 20205 70000=

The option -T uses the cache for bufr tables. This is very interesting in addition to option -I . With the -I option you enter a file with a list of bufr file paths than bufrtotac will parse in sequential order. If the master version of every bufrfile is not the same then the use of -T option will create a internal cache in memory that will optimize the CPU time.
The options -R and -W read or write respectively a small file of bitoffset index for non compressed BUFR files. The name of index file assocciated to every BUFR file is to concatenate .offs to the original bufr filename. This feature is useful if you need to decode the same bufr file many times and only access to a given subset. First time you decode a file using bufrdeco with -W option. Next times you can use -R option to read the data for subsets in a optimized way

Since 0.24.0 there are also some options to get bufr data in json format.

The options -0 -1 -2 -3 display json objects for the sections 0 to 3 respectively.
The option -E display a json object with the expanded descriptor tree (without data nor replications)
The option -J display a json object with all the expanded and replicated tree of descriptors, sequences and data in the bufr report.

As example, you can view the content of sec3 of the file 20141018211119_ISIN03_EGRR_182100.bufr using the folowing command. In addition of option -3, the option -n is used to not decode the data into TAC. Note that also is used the utility jq to view the results in a readble mode (bufrdeco library output json objects in a single line and is not easy to read by humans in this form).

bufrtotac -i 20141018211119_ISIN03_EGRR_182100.bufr -n -3 | jq
{
  "Sec 3": {
    "Sec3 length": 9,
    "Subsets": 58,
    "Observed": 1,
    "Compressed": 0,
    "Unexpanded descriptors": 1,
    "Unexpanded array": [
      {
        "0": "3 07 080"
      }
    ]
  }
}

The binary bufrdeco_json is build to get data from a bufr report in json format without decode any data in TAC format. Options are

bufrdeco_json -h
Usage: 
bufrdeco_test -i input_file [-h][more optional args...]
   -h Print this help
   -i Input file. Complete input path file for bufr file
   -J. Information, tree and data in json format. Equivalent to option -E01234
   -S first..last . Print only results for subsets in range first..last (First subset available is 0). Default is all subsets
   -T. Print expanded tree in json format
   -X. Extract first BUFR buffer found in input file (from first 'BUFR' item to next '7777')
   -0. Prints BUFR Sec 0 information in json format
   -1. Prints BUFR Sec 1 information in json format
   -2. Prints BUFR Sec 2 information in json format
   -3. Prints BUFR Sec 3 information in json format
   -4. Prints BUFR data in json format

As example, we can use it to view Secs 1 and 3 of file 20150705121512_ISCD01_LIIB_050000.bufr included in example directory.

bufrdeco_json -i 20150705121512_ISCD01_LIIB_050000.bufr -13 |jq
{
  "Sec 1": {
    "Length": 22,
    "Bufr master table": 0,
    "Centre": 80,
    "Sub-Centre": 0,
    "Update sequence": 0,
    "Options": "0x0",
    "Category": 0,
    "Subcategory": 20,
    "Sub-category local": 255,
    "Master table version": 16,
    "Master table local": 0,
    "Year": 2015,
    "Month": 6,
    "Day": 1,
    "Hour": 0,
    "Minute": 0,
    "Second": 0,
    "Aditional space": 0
  }
}
{
  "Sec 3": {
    "Sec3 length": 9,
    "Subsets": 19,
    "Observed": 1,
    "Compressed": 0,
    "Unexpanded descriptors": 1,
    "Unexpanded array": [
      {
        "0": "3 07 073"
      }
    ]
  }
}

And finally, you also can read the doc pages at github project site

gbvalor/bufr2synop