Iris Data Structures#

The top level object in Iris is called a cube. A cube contains data and metadata about a phenomenon.

In Iris, a cube is an interpretation of the Climate and Forecast (CF) Metadata Conventions whose purpose is to:

require conforming datasets to contain sufficient metadata that they are self-describing… including physical units if appropriate, and that each value can be located in space (relative to earth-based coordinates) and time.

Whilst the CF conventions are often mentioned alongside NetCDF, Iris implements several major format importers which can take files of specific formats and turn them into Iris cubes. Additionally, a framework is provided which allows users to extend Iris’ import capability to cater for specialist or unimplemented formats.

A single cube describes one and only one phenomenon, always has a name, a unit and an n-dimensional data array to represents the cube’s phenomenon. In order to locate the data spatially, temporally, or in any other higher-dimensional space, a collection of coordinates exist on the cube.


A coordinate is a container to store metadata about some dimension(s) of a cube’s data array and therefore, by definition, its phenomenon.

  • Each coordinate has a name and a unit.

  • When a coordinate is added to a cube, the data dimensions that it

    represents are also provided.

    • The shape of a coordinate is always the same as the shape of the associated data dimension(s) on the cube.

    • A dimension not explicitly listed signifies that the coordinate is independent of that dimension.

    • Each dimension of a coordinate must be mapped to a data dimension. The only coordinates with no mapping are scalar coordinates.

  • Depending on the underlying data that the coordinate is representing, its values may be discrete points or be bounded to represent interval extents (e.g. temperature at point x vs rainfall accumulation between 0000-1200 hours).

  • Coordinates have an attributes dictionary which can hold arbitrary extra metadata, excluding certain restricted CF names

  • More complex coordinates may contain a coordinate system which is necessary to fully interpret the values contained within the coordinate.

There are two classes of coordinates:


  • Numeric

  • Monotonic

  • Representative of, at most, a single data dimension (1d)


  • May be of any type, including strings

  • May represent multiple data dimensions (n-dimensional)


A cube consists of:

  • a standard name and/or a long name and an appropriate unit

  • a data array who’s values are representative of the phenomenon

  • a collection of coordinates and associated data dimensions on the cube’s data array, which are split into two separate lists:

    • dimension coordinates - DimCoords which uniquely map to exactly one data dimension, ordered by dimension.

    • auxiliary coordinates - DimCoords or AuxCoords which map to as many data dimensions as the coordinate has dimensions.

  • an attributes dictionary which, other than some protected CF names, can hold arbitrary extra metadata. This implements the concept of dataset-level and variable-level attributes when loading and and saving NetCDF files (see CubeAttrsDict and NetCDF save() for more).

  • a list of cell methods to represent operations which have already been applied to the data (e.g. “mean over time”)

  • a list of coordinate “factories” used for deriving coordinates from the values of other coordinates in the cube

Cubes in Practice#

A Simple Cube Example#

Suppose we have some gridded data which has 24 air temperature readings (in Kelvin) which is located at 4 different longitudes, 2 different latitudes and 3 different heights. Our data array can be represented pictorially:


Where dimensions 0, 1, and 2 have lengths 3, 2 and 4 respectively.

The Iris cube to represent this data would consist of:

  • a standard name of air_temperature and a unit of kelvin

  • a data array of shape (3, 2, 4)

  • a coordinate, mapping to dimension 0, consisting of:

    • a standard name of height and unit of meters

    • an array of length 3 representing the 3 height points

  • a coordinate, mapping to dimension 1, consisting of:

    • a standard name of latitude and unit of degrees

    • an array of length 2 representing the 2 latitude points

    • a coordinate system such that the latitude points could be fully located on the globe

  • a coordinate, mapping to dimension 2, consisting of:

    • a standard name of longitude and unit of degrees

    • an array of length 4 representing the 4 longitude points

    • a coordinate system such that the longitude points could be fully located on the globe

Pictorially the cube has taken on more information than a simple array:


Additionally further information may be optionally attached to the cube. For example, it is possible to attach any of the following:

  • a coordinate, not mapping to any data dimensions, consisting of:

    • a standard name of time and unit of days since 2000-01-01 00:00

    • a data array of length 1 representing the time that the data array is valid for

  • an auxiliary coordinate, mapping to dimensions 1 and 2, consisting of:

    • a long name of place name and no unit

    • a 2d string array of shape (2, 4) with the names of the 8 places that the lat/lons correspond to

  • an auxiliary coordinate “factory”, which can derive its own mapping, consisting of:

    • a standard name of height and a unit of feet

    • knowledge of how data values for this coordinate can be calculated given the height in meters coordinate

  • a cell method of “mean” over “ensemble” to indicate that the data has been meaned over a collection of “ensembles” (i.e. multiple model runs).

Printing a Cube#

Every Iris cube can be printed to screen as you will see later in the user guide. It is worth familiarising yourself with the output as this is the quickest way of inspecting the contents of a cube. Here is the result of printing a real life cube:

air_potential_temperature / (K)     (time: 3; model_level_number: 7; grid_latitude: 204; grid_longitude: 187)
    Dimension coordinates:
        time                             x                      -                 -                    -
        model_level_number               -                      x                 -                    -
        grid_latitude                    -                      -                 x                    -
        grid_longitude                   -                      -                 -                    x
    Auxiliary coordinates:
        forecast_period                  x                      -                 -                    -
        level_height                     -                      x                 -                    -
        sigma                            -                      x                 -                    -
        surface_altitude                 -                      -                 x                    x
    Derived coordinates:
        altitude                         -                      x                 x                    x
    Scalar coordinates:
        forecast_reference_time     2009-11-19 04:00:00
        STASH                       m01s00i004
        source                      'Data from Met Office Unified Model'
        um_version                  '7.3'

Using this output we can deduce that:

  • The cube represents air potential temperature.

  • There are 4 data dimensions, and the data has a shape of (3, 7, 204, 187)

  • The 4 data dimensions are mapped to the time, model_level_number, grid_latitude, grid_longitude coordinates respectively

  • There are three 1d auxiliary coordinates and one 2d auxiliary (surface_altitude)

  • There is a single altitude derived coordinate, which spans 3 data dimensions

  • There are 7 distinct values in the “model_level_number” coordinate. Similar inferences can be made for the other dimension coordinates.

  • There are 7, not necessarily distinct, values in the level_height coordinate.

  • There is a single forecast_reference_time scalar coordinate representing the entire cube.

  • The cube has one further attribute relating to the phenomenon. In this case the originating file format, PP, encodes information in a STASH code which in some cases can be useful for identifying advanced experiment information relating to the phenomenon.