4. Navigating a cube

After loading any cube, you will want to investigate precisely what it contains. This section is all about accessing and manipulating the metadata contained within a cube.

4.1. Cube string representations

We have already seen a basic string representation of a cube when printing:

>>> import iris
>>> filename = iris.sample_data_path('rotated_pole.nc')
>>> cube = iris.load_cube(filename)
>>> print(cube)
air_pressure_at_sea_level / (Pa)    (grid_latitude: 22; grid_longitude: 36)
     Dimension coordinates:
          grid_latitude                           x                   -
          grid_longitude                          -                   x
     Scalar coordinates:
          forecast_period: 0.0 hours
          forecast_reference_time: 2006-06-15 00:00:00
          time: 2006-06-15 00:00:00
     Attributes:
          Conventions: CF-1.5
          STASH: m01s16i222
          source: Data from Met Office Unified Model 6.01

This representation is equivalent to passing the cube to the str() function. This function can be used on any Python variable to get a string representation of that variable. Similarly there exist other standard functions for interrogating your variable: repr(), type() for example:

print(str(cube))
print(repr(cube))
print(type(cube))

Other, more verbose, functions also exist which give information on what you can do with any given variable. In most cases it is reasonable to ignore anything starting with a “_” (underscore) or a “__” (double underscore):

dir(cube)
help(cube)

4.2. Working with cubes

Every cube has a standard name, long name and units which are accessed with Cube.standard_name, Cube.long_name and Cube.units respectively:

print(cube.standard_name)
print(cube.long_name)
print(cube.units)

Interrogating these with the standard type() function will tell you that standard_name and long_name are either a string or None, and units is an instance of iris.unit.Unit.

You can access a string representing the “name” of a cube with the Cube.name() method:

print(cube.name())

The result of which is always a string.

Each cube also has a numpy array which represents the phenomenon of the cube which can be accessed with the Cube.data attribute. As you can see the type is a numpy n-dimensional array:

print(type(cube.data))

Note

When loading some file formats in Iris, the data itself is not loaded until the first time that the data is requested. Hence you may have noticed that running the previous command for the first time takes a little longer than it does for subsequent calls.

For this reason, when you have a large cube it is strongly recommended that you do not access the cube’s data unless you need to. For convenience shape and ndim attributes exists on a cube, which can tell you the shape of the cube’s data without loading it:

print(cube.shape)
print(cube.ndim)

You can change the units of a cube using the convert_units() method. For example:

cube.convert_units('celsius')

As well as changing the value of the units attribute this will also convert the values in data. To replace the units without modifying the data values one can change the units attribute directly.

Some cubes represent a processed phenomenon which are represented with cell methods, these can be accessed on a cube with the Cube.cell_methods attribute:

print(cube.cell_methods)

4.3. Accessing coordinates on the cube

A cube’s coordinates can be retrieved via Cube.coords. A simple for loop over the coords can print a coordinate’s name():

for coord in cube.coords():
    print(coord.name())

Alternatively, we can use list comprehension to store the names in a list:

coord_names = [coord.name() for coord in cube.coords()]

The result is a basic Python list which could be sorted alphabetically and joined together:

>>> print(', '.join(sorted(coord_names)))
forecast_period, forecast_reference_time, grid_latitude, grid_longitude, time

To get an individual coordinate given its name, the Cube.coord method can be used:

coord = cube.coord('grid_latitude')
print(type(coord))

Every coordinate has a Coord.standard_name, Coord.long_name, and Coord.units attribute:

print(coord.standard_name)
print(coord.long_name)
print(coord.units)

Additionally every coordinate can provide its points and bounds numpy array. If the coordinate has no bounds None will be returned:

print(type(coord.points))
print(type(coord.bounds))

4.4. Adding metadata to a cube

We can add and remove coordinates via Cube.add_dim_coord, Cube.add_aux_coord, and Cube.remove_coord.

>>> import iris.coords
>>> new_coord = iris.coords.AuxCoord(1, long_name='my_custom_coordinate', units='no_unit')
>>> cube.add_aux_coord(new_coord)
>>> print(cube)
air_pressure_at_sea_level / (Pa)    (grid_latitude: 22; grid_longitude: 36)
     Dimension coordinates:
          grid_latitude                           x                   -
          grid_longitude                          -                   x
     Scalar coordinates:
          forecast_period: 0.0 hours
          forecast_reference_time: 2006-06-15 00:00:00
          my_custom_coordinate: 1
          time: 2006-06-15 00:00:00
     Attributes:
          Conventions: CF-1.5
          STASH: m01s16i222
          source: Data from Met Office Unified Model 6.01

The coordinate my_custom_coordinate now exists on the cube and is listed under the non-dimensioned single valued scalar coordinates.

4.5. Adding and removing metadata to the cube at load time

Sometimes when loading a cube problems occur when the amount of metadata is more or less than expected. This is often caused by one of the following:

• The file does not contain enough metadata, and therefore the cube cannot know everything about the file.
• Some of the metadata of the file is contained in the filename, but is not part of the actual file.
• There is not enough metadata loaded from the original file as Iris has not handled the format fully. (in which case, please let us know about it)

To solve this, all of iris.load(), iris.load_cube(), and iris.load_cubes() support a callback keyword.

The callback is a user defined function which must have the calling sequence function(cube, field, filename) which can make any modifications to the cube in-place, or alternatively return a completely new cube instance.

Suppose we wish to load a lagged ensemble dataset from the Met Office’s GloSea4 model. The data for this example represents 13 ensemble members of 6 one month timesteps; the logistics of the model mean that the run is spread over several days.

If we try to load the data directly for surface_temperature:

>>> filename = iris.sample_data_path('GloSea4', '*.pp')
>>> print(iris.load(filename, 'surface_temperature'))
0: surface_temperature / (K)           (time: 6; forecast_reference_time: 2; latitude: 145; longitude: 192)
1: surface_temperature / (K)           (time: 6; forecast_reference_time: 2; latitude: 145; longitude: 192)
2: surface_temperature / (K)           (realization: 9; time: 6; latitude: 145; longitude: 192)

We get multiple cubes some with more dimensions than expected, some without a realization (i.e. ensemble member) dimension. In this case, two of the PP files have been encoded without the appropriate realization number attribute, which means that the appropriate coordinate cannot be added to the resultant cube. Fortunately, the missing attribute has been encoded in the filename which, given the filename, we could extract:

filename = iris.sample_data_path('GloSea4', 'ensemble_001.pp')
realization = int(filename[-6:-3])
print(realization)

We can solve this problem by adding the appropriate metadata, on load, by using a callback function, which runs on a field by field basis before they are automatically merged together:

import numpy as np
import iris
import iris.coords as icoords

def lagged_ensemble_callback(cube, field, filename):
    # Add our own realization coordinate if it doesn't already exist.
    if not cube.coords('realization'):
        realization = np.int32(filename[-6:-3])
        ensemble_coord = icoords.AuxCoord(realization, standard_name='realization')
        cube.add_aux_coord(ensemble_coord)

filename = iris.sample_data_path('GloSea4', '*.pp')

print(iris.load(filename, 'surface_temperature', callback=lagged_ensemble_callback))

The result is a single cube which represents the data in a form that was expected:

0: surface_temperature / (K)           (realization: 13; time: 6; latitude: 145; longitude: 192)