Python Bindings for NCO

Posted by Joe Hamman on January 29, 2014

Like many in the geophysical sciences these days, my preferred data format is netCDF (Network Common Data Form). netCDF offers “a set of software libraries and self-describing, machine-independent data formats that support the creation, access, and sharing of array-oriented scientific data.” The fact that the format is self-describing and machine-independent has allowed for the development generic processing tools. Considerable work has gone into developing some very nice and quite advanced netCDF file operators. The two libraries I find myself using on a daily basis are NCO (netCDF Operators) and CDO (Climate Data Operators).

Both NCO and CDO were developed for use on the command line or in the shell (e.g. bash). While the shell is really nice for doing basic operations, I still don’t find it convenient for more complicated file handling operations. In walks Python. Python has been my scripting language of choice and I have found that it offers some nice solutions for file handling. A few years back, Ralf Mueller introduced a set of Python bindings to CDO. allows the user to call any of the CDO operators from within python. As far as I know, no one has done this for NCO, so I gave it a go.


pynco is a clone of the cdo-bindings Python module. It follows basically the same syntax as with a few adjustments. The biggest difference is in its handling of keyword arguments. If you have used NCO before, you know that there are dozens of keyword arguments. In, any long name (e.g. --append) command line argument can be passed to as a keyword argument (e.g. append=True). Here are some examples:

  • Everything starts with initializing a nco object

      from nco import Nco
      nco = Nco()
  • Append two files:

    On the command line it would look like this:

      ncks -A

    With we can do the same thing in one of many ways:

      nco.ncks(input='', output='', options='-A')
      nco.ncks(input='', output='', append=True)
  • Subset using NCO hyperslabs: On the command line it would look like this:

      ncks -F -d lon,1,2
      ncks -d lon,1,2

    With we can do the same thing in one of many ways:

      nco.ncks(input='', output='', fortran=True,
      nco.ncks(input='', output='',

The module also supplies easy temporary file handling. For example,

    temporary_file1 = nco.ncrcat(input=a_long_list_of_files)
    temporary_file2 = nco.ncrcat(input=b_long_list_of_files)
    nco.ncra(input=[temporary_file1, temporary_file2], output='')

A particularly nice feature of both the CDO and NCO python modules is their ability to return open netCDF file objects or numpy arrays. For example, here I take the ensemble average of a list of files and return:

  • netCDF file object (either scipy or netCDF4-python)

      testCdf = nco.ncea(input=['', '', ''], returnCdf=True)
  • numpy array

      temps = nco.ncea(input=['', '', ''], returnArray='temperature')
  • numpy masked array

      masked_temps = nco.ncea(input=['', '', ''], returnMaArray='temperature')

The module is still in beta development phase but it’s passing all its tests (Python 2.6, 2.7, and 3.3) so I thought I would mention it here. I’ve been using it for the last few weeks in my processing scripts and have really liked its functionality. All NCO operators are available including

['ncap2', 'ncatted', 'ncbo', 'nces', 'ncecat', 'ncflint', 'ncks', 'ncpdq', 'ncra', 'ncrcat', 'ncrename', 'ncwa', 'ncea', 'ncdump']

The code can be found on github

Instructions for installation along with a few more examples are listed in the readme file.