== FLEXPART meteorological input from ECMWF ==

||[[span(style=color: #333366, Do not remove or rename this page as it is referenced in [https://doi.org/10.5194/gmd-2017-323 a GMD paper])]]||

[FpInput ^back to FpInput]

=== 1. Get access to ECMWF MARS ===

In order to extract data from the ECMWF MARS archive, one needs an account at ECMWF. 

'''Reanalysis''' and MACC/CAMS '''data are publicly available.'''  The [https://software.ecmwf.int/wiki/display/WEBAPI/How+to+retrieve+ECMWF+Public+Datasets procedure for registration] is explained on the ECMWF web site (register and perform steps 1 and 2). Consider the [https://software.ecmwf.int/wiki/display/WEBAPI/Available+ECMWF+Public+Datasets applicable licence].

For other data (e.g., operational IFS data) a member-state or special-project account is be required. Contact your national meteorological service if you need this and your country is among ECMWF member states.

=== 2. How to retrieve data ===

Software for retrieving ECMWF data in the specific format needed by FLEXPART was originally created by Gerhard Wotawa. The later versions were written by Leo Haimberger on behalf of CTBTO.

    * [download:15 ECMWF data retrieval software, V2.1]
    * [download:16 ECMWF data retrieval software, V4], allows to retrieve compressed ECMWF data on old AIX ecgate in mixed GRIB1/GRIB2 coding 
    * [download:46 ECMWF data retrieval software, V4.1], allows to retrieve compressed ECMWF data on new Linux ecgate in mixed GRIB1/GRIB2 coding 
    * [download:50 flex_extract_ecgate_V6.0.tar.gz],   a new version for retrieval with 1h time resolution and other enhancements 
    * [download:55 flex_extract V7.0.2] New, python-based version. Allows access to publicly available data without a member state account through the python framework of ECMWF (`ecmwf-api-client-python`). Note that if you have the choice between working through this interface or directly on `ecaccess`/ through a gateway server, one of the latter options will usually be considerably faster. In any case, extracting large data sets takes time.