How to run

Input data

Currently, ARC3O is tailored for model output from ECHAM, the atmospheric module of the Max Planck Institute Earth System Model. However, it can be used for other models if the variable names are changed accordingly either in the ARC3O functions or the model’s output.

The input data for ARC3O should be divided into monthly files, found in the folder inputpath. Also, to prepare the seasons and ice types mask, ARC3O needs one file in which all the data is merged, found in the folder inputpath0.

These monthly and overview files should contain fields [time,lat,lon] of the following variables:

  • snifrac: snow fraction on ice [0-1]

  • siced: sea-ice thickness [m]

  • sni: snow water equivalent [m]

  • tsi: surface temperature at surface of snow (if present) or of ice (if no snow) [K]

  • qvi: columnar water vapor [mm]

  • wind10: wind speed [m/s]

  • xlvi: columnar liquid water [mm]

  • tsw: sea-surface temperature [K]

  • seaice: sea-ice concentration [0-1]

  • slm: sea-land mask

  • ameltfrac: melt-pond fraction [0-1]

Running ARC3O

You can run ARC3O with the function arc3o.core_functions.satsim_complete_parallel() as follows:

import xarray as xr
import arc3o as arc3o

# inputpath for the monthly chucked files of the climate model output
inputpath = 'pathtofolder1'
# inputpath for the merged file of the climate model output for the whole time series
inputpath0 = 'pathtofolder2'
# outputpath for output folders of your different experiments
outputpath0 = 'pathtofolder3'

# outputpath for your experiment files
#'yes': create a new path for the output files (will create a new folder in outputpath0, called
# yyyymmdd-hhmm)
#'no': keeps the path given in the third option
outputpath = arc3o.new_outputpath('no',outputpath0,'20190516-1047')


# read in the whole time period
orig_data = xr.open_dataset(inputpath0+'assim_SICCI2_50km_echam6_200211-200812_selcode_Arctic.nc')
# transform the unusual MPI-ESM timestamp into a better readable one
orig_data = arc3o.prep_time(orig_data)

# give the first and last year of the time period of your data
start_year = 2003
end_year = 2008

# explain how the monthly chunked file names are built around yyyymm (the ones in inputpath),
# example files are called 'assim_SICCI2_50km_echam6_yyyymm_selcode_Arctic.nc' where yyyy is the year and mm the month
file_begin = 'assim_SICCI2_50km_echam6_'
file_end = '_selcode_Arctic.nc'

# frequency of interest in GHz (must fit one of the AMSR-E frequencies: 6.9, 10.7, 18.7, 23.8, 36.5, 50.3, 52.8, 89.0)
freq_of_int = 6.9

# run ARC3O
arc3o.satsim_complete_parallel(orig_data,               # climate model output, whole time series
                            freq_of_int,                # frequency of interest
                            start_year,end_year,        # first and last year of the time period of interest
                            inputpath,                  # where to find the monthly chunked climate model output
                            outputpath,                 # where to write out the results
                            file_begin,file_end,        # file name as wrapped around yyyymm for the monthly chunked files
                            timestep=6,                 # timestep of climate model data in hours
                            write_mask='yes',           # 'yes' if you want to compute and write out the ice type and season mask, 'no' if you already have a file 'period_masks_assim.nc' in outputpath
                            write_profiles='yes',       # 'yes' if you want to compute and write out the profiles, 'no' if you already have monthly chunked files 'profiles_for_memls_snowno_yyyymm.nc' and 'profiles_for_memls_snowyes yyyymm.nc' in outputpath
                            compute_memls='yes',        # 'yes' if you want to compute and write out the cold conditions ice surface brightness temperature, 'no' if you already have monthly chunked files 'TB_assim_yyyymm_f.nc' in outputpath
                            e_bias_fyi=0.968,           # factor affecting the temperature profiles to bias-correct the brightness temperature (for first-year ice)
                            e_bias_myi=0.968,           # factor affecting the temperature profiles to bias-correct the brightness temperature (for multiyear ice)
                            snow_emis=1,                # snow emissivity for periods of melting snow
                            snow dens=300.,             # snow density in kg/m3
                            pool_nb=12)                 # amount of pool workers to compute several months parallelly

Note

The process takes up a lot of memory so I recommend using a supercomputer if possible. I am working on a “lighter” version but not sure when it will be ready so you’ll currently need to reduce the number of pool workers in the Pool. To do so, choose a pool_nb between 1 and 12. The lower the pool_nb, the less pool workers are activated at the moment (i.e. less memory usage but also longer time until results are ready).

If you only want to run one month instead of the whole time series, you can also run ARC3O with arc3o.core_functions.satsim_complete_1month():

import xarray as xr
import arc3o as arc3o

# inputpath for the monthly chucked files of the climate model output
inputpath = 'pathtofolder1'
# inputpath for the merged file of the climate model output for the whole time series
inputpath0 = 'pathtofolder2'
# outputpath for output folders of your different experiments
outputpath0 = 'pathtofolder3'

# outputpath for your experiment files
#'yes': create a new path for the output files (will create a new folder in outputpath0, called
# yyyymmdd-hhmm)
#'no': keeps the path given in the third option
outputpath = arc3o.new_outputpath('no',outputpath0,'20190516-1047')

# read in the whole time period
orig_data = xr.open_dataset(inputpath0+'assim_SICCI2_50km_echam6_200211-200812_selcode_Arctic.nc')
# transform the unusual MPI-ESM timestamp into a better readable one
orig_data = arc3o.prep_time(orig_data)

# year and month of interest
yyyy = 2004
mm = 6

# explain how the monthly chunked file names are built around yyyymm (the ones in inputpath),
# example files are called 'assim_SICCI2_50km_echam6_yyyymm_selcode_Arctic.nc' where yyyy is the year and mm the month
file_begin = 'assim_SICCI2_50km_echam6_'
file_end = '_selcode_Arctic.nc'

### frequency of interest in GHz (must fit one of the AMSR-E frequencies)
freq_of_int = 6.9

### run the operator
arc3o.satsim_complete_1month(orig_data,            # climate model output, whole time series
                          freq_of_int,              # frequency of interest
                          yyyy,mm,                  # year and month of interest
                          inputpath,                # where to find the monthly chunked climate model output
                          outputpath,               # where to write out the results
                          file_begin,file_end,      # file name as wrapped around yyyymm for the monthly chunked files
                          timestep=6,               # timestep of climate model data in hours
                          write_mask='yes',         # 'yes' if you want to compute and write out the ice type and season mask, 'no' if you already have a file 'period_masks_assim.nc' in outputpath
                          write_profiles='yes',     # 'yes' if you want to compute and write out the profiles, 'no' if you already have monthly chunked files 'profiles_for_memls_snowno_yyyymm.nc' and 'profiles_for_memls_snowyes yyyymm.nc' in outputpath
                          compute_memls='yes',      # 'yes' if you want to compute and write out the cold conditions ice surface brightness temperature, 'no' if you already have monthly chunked files 'TB_assim_yyyymm_f.nc' in outputpath
                          e_bias_fyi=0.968,         # factor affecting the temperature profiles to bias-correct the brightness temperature (for first-year ice)
                          e_bias_myi=0.968,         # factor affecting the temperature profiles to bias-correct the brightness temperature (for multiyear ice)
                          snow_emis=1,              # snow emissivity for periods of melting snow
                          snow dens=300.)           # snow density in kg/m3

Output

The output of ARC3O is written into several netcdf files to outputpath:

  • 'period_masks_assim.nc': Masks for ice type and seasons.

  • 'profiles_for_memls_snowno_yyyymm.nc': Snow-free profiles of ice and snow properties.

  • 'profiles_for_memls_snowyes_yyyymm.nc': Snow-covered profiles of ice and snow properties.

  • 'TB_assim_yyyymm_f.nc': Ice surface brightness temperatures (H and V polarization) for grid cells with ice in cold conditions.

  • 'TBtot_assim_yyyymm_f.nc': Brightness temperatures (H and V polarization) at the top of atmosphere (incl. other seasons than cold conditions and ocean and atmosphere contribution) for all ocean grid cells.

Note

Please remain aware that the assumptions used in ARC3O have only been evaluated for the frequency of 6.9 GHz, vertical polarization at the moment! The use for other frequencies and polarizations is at your own risk!