dfid-outlooks

Code to process data for DFID Climate Outlooks

Reforecasts and operational forecasts are obtained from four models:

Model	Reforecast ensemble	Operational ensemble	Method
Met Office (UKMO)	7 members, 9th of month, 1993-2015, downloaded each time	20 members: 2 members per day, lagged ensemble of last 10 days	From public sector catalogue on JASMIN
National Centers for Environmental Prediction	4 members, every 5 days, 1982-2011, fixed hindcast set	24 members: 4 members per day, lagged ensemble of last 6 days	From NCEP HTTP site
European Centre for Medium-range Weather Forecasts	25 members, 1st of month, 1993-2015, downloaded each time	51 members, 1st of month (but only publicly available on 13th at 12Z)	From Copernicus Data Service (CDS)
Meteo-France	25 members, 1st of month, 1993-2015, downloaded each time	51 members, 1st of month (but only publicly available on 13th at 12Z)	From Copernicus Data Service (CDS)

UKMO

1. Use “get_ukmo_operational.ksh” to download and process Met Office operational forecasts and re-forecasts from the JASMIN catalogue.
2. You can download and process the re-forecasts in advance by setting “fcast_get” and “fcast_proc” to zero and “hcast_get” and “hcast_proc” to one.

NCEP

1. Use “get_ncep_operational.ksh” to download and process NCEP operational forecasts from the NCEP HTTP site.
2. Use “get_ncep_hindcasts.ksh” to download and process NCEP re-forecasts from the NCEP HTTP site. NCEP use a fixed reforecast set, so you need to do this only once per re-forecast set.
3. Use "hindcasts_make_ensmeans_ts.ksh" to process the NCEP re-forecasts to create timeseries of ensemble means, which are read by the python code that produces the Climate Outlooks. The shell script takes a single argument of the re-forecast start date (MMDD).

ECMWF and Meteo-France

1. Use “get_copernicus_dfid.py” to download ECMWF and Meteo-France data from the CDS.
2. You can download re-forecast data in advance by commenting out the lines that download forecast data, or by setting an “old” forecast date you have already downloaded, as the script will not overwrite files.
3. Use “organise_and_convert_c3s_hcast.ksh” to process re-forecast data from either model. The script is designed to convert one re-forecast, so it can be run in parallel as the re-forecast data are downloaded as one GRIB file per year.
4. Use “organise_and_convert_c3s_fcast.ksh” to process forecast data from either model.
5. Use “compute_mmeans_anoms_c3.ksh” to process forecast and re-forecast data into monthly means and ensemble means.

NB: The KSH scripts above all take various arguments, which are commented in the code.

Dependencies

• Python packages:
  • salem (for country averaging): https://salem.readthedocs.io/en/v0.2.3/
  • cf-python (for plotting maps): https://cfpython.bitbucket.io/
  • cf-plot (for plotting maps): http://ajheaps.github.io/cf-plot
• Fortran:
  • fix_netcdf_dimension_monotonic (for overwriting netCDF coordinate variables) - included in repository but needs to be compiled on your system.