Xarray’s Data structures

Xarray’s Data structures#

In this lesson, we cover the basics of Xarray data structures. Our learning goals are as follows. By the end of the lesson, we will be able to:

Understand the basic data structures (DataArray and Dataset objects) in Xarray

Introduction#

Multi-dimensional (a.k.a. N-dimensional, ND) arrays (sometimes called “tensors”) are an essential part of computational science. They are encountered in a wide range of fields, including physics, astronomy, geoscience, bioinformatics, engineering, finance, and deep learning. In Python, NumPy provides the fundamental data structure and API for working with raw ND arrays. However, real-world datasets are usually more than just raw numbers; they have labels which encode information about how the array values map to locations in space, time, etc.

Here is an example of how we might structure a dataset for a weather forecast:

https://docs.xarray.dev/en/stable/_images/dataset-diagram.png

You’ll notice multiple data variables (temperature, precipitation), coordinate variables (latitude, longitude), and dimensions (x, y, t). We’ll cover how these fit into Xarray’s data structures below.

Xarray doesn’t just keep track of labels on arrays – it uses them to provide a powerful and concise interface. For example:

Apply operations over dimensions by name: x.sum('time').
Select values by label (or logical location) instead of integer location: x.loc['2014-01-01'] or x.sel(time='2014-01-01').
Mathematical operations (e.g., x - y) vectorize across multiple dimensions (array broadcasting) based on dimension names, not shape.
Easily use the split-apply-combine paradigm with groupby: x.groupby('time.dayofyear').mean().
Database-like alignment based on coordinate labels that smoothly handles missing values: x, y = xr.align(x, y, join='outer').
Keep track of arbitrary metadata in the form of a Python dictionary: x.attrs.

The N-dimensional nature of xarray’s data structures makes it suitable for dealing with multi-dimensional scientific data, and its use of dimension names instead of axis labels (dim='time' instead of axis=0) makes such arrays much more manageable than the raw numpy ndarray: with xarray, you don’t need to keep track of the order of an array’s dimensions or insert dummy dimensions of size 1 to align arrays (e.g., using np.newaxis).

The immediate payoff of using xarray is that you’ll write less code. The long-term payoff is that you’ll understand what you were thinking when you come back to look at it weeks or months later.

Data structures#

Xarray provides two data structures: the DataArray and Dataset. The DataArray class attaches dimension names, coordinates and attributes to multi-dimensional arrays while Dataset combines multiple arrays.

Both classes are most commonly created by reading data. To learn how to create a DataArray or Dataset manually, see the Creating Data Structures tutorial.

Xarray has a few small real-world tutorial datasets hosted in this GitHub repository pydata/xarray-data. We’ll use the xarray.tutorial.load_dataset convenience function to download and open the air_temperature (National Centers for Environmental Prediction) Dataset by name.

import numpy as np
import xarray as xr

To Pandas and back#

DataArray and Dataset objects are frequently created by converting from other libraries such as pandas or by reading from data storage formats such as NetCDF or zarr.

To convert from / to pandas, we can use the to_xarray methods on pandas objects or the to_pandas methods on xarray objects:

import pandas as pd

series = pd.Series(np.ones((10,)), index=list("abcdefghij"))
series

a    1.0
b    1.0
c    1.0
d    1.0
e    1.0
f    1.0
g    1.0
h    1.0
i    1.0
j    1.0
dtype: float64

arr = series.to_xarray()
arr

<xarray.DataArray (index: 10)>
array([1., 1., 1., 1., 1., 1., 1., 1., 1., 1.])
Coordinates:
  * index    (index) object 'a' 'b' 'c' 'd' 'e' 'f' 'g' 'h' 'i' 'j'

arr.to_pandas()

index
a    1.0
b    1.0
c    1.0
d    1.0
e    1.0
f    1.0
g    1.0
h    1.0
i    1.0
j    1.0
dtype: float64

We can also control what pandas object is used by calling to_series / to_dataframe:

to_series: This will always convert DataArray objects to pandas.Series, using a MultiIndex for higher dimensions

ds.air.to_series()

time                 lat   lon  
2013-01-01 00:00:00  75.0  200.0    241.199997
                           202.5    242.500000
                           205.0    243.500000
                           207.5    244.000000
                           210.0    244.099991
                                       ...    
2014-12-31 18:00:00  15.0  320.0    297.389984
                           322.5    297.190002
                           325.0    296.489990
                           327.5    296.190002
                           330.0    295.690002
Name: air, Length: 3869000, dtype: float32

to_dataframe: This will always convert DataArray or Dataset objects to a pandas.DataFrame. Note that DataArray objects have to be named for this.

ds.air.to_dataframe()

			air
time	lat	lon
2013-01-01 00:00:00	75.0	200.0	241.199997
		202.5	242.500000
		205.0	243.500000
		207.5	244.000000
		210.0	244.099991
...	...	...	...
2014-12-31 18:00:00	15.0	320.0	297.389984
		322.5	297.190002
		325.0	296.489990
		327.5	296.190002
		330.0	295.690002

3869000 rows × 1 columns

Since columns in a DataFrame need to have the same index, they are broadcasted.

ds.to_dataframe()

			air
lat	time	lon
75.0	2013-01-01 00:00:00	200.0	241.199997
		202.5	242.500000
		205.0	243.500000
		207.5	244.000000
		210.0	244.099991
...	...	...	...
15.0	2014-12-31 18:00:00	320.0	297.389984
		322.5	297.190002
		325.0	296.489990
		327.5	296.190002
		330.0	295.690002

3869000 rows × 1 columns

Xarray’s Data structures

Contents

Xarray’s Data structures#

Introduction#

Data structures#

Dataset#

What is all this anyway? (String representations)#

DataArray#

String representations#

Named dimensions#

Coordinates#

Attributes#

To Pandas and back#