Creating Xarray DataArrays

Courses

Intro to Python for Scientists

Dictionary and Xarray Data Arrays

Authors

Dr. Nicholas Del Grosso | Dr. Sangeetha Nandakumar | Dr. Ole Bialas | Dr. Atle E. Rimehaug

Open in JupyterLite Download Exercises

Setup

Import Libraries

import xarray as xr
import numpy as np

Thus far, we have worked with data organized in separate numpy arrays, but when working with real data, keeping track of many arrays stored in separate variables and how they relate to one another can quickly become complicated. Xarray helps you store multiple arrays and information about their relationships together in one DataArray, making it easier for the researcher to work with and understand the data.

In this notebook, we will look at how you can put multiple arrays together in xarray DataArrays, access the data stored in a DataArray, and how save it to in a file.

Section 1: Labeling the Indices of a DataArray’s Dimensions and Accessing the Data

Code	Description
`da = xr.DataArray(data=x, coords={'time': y}, name='sensor')`	Make a DataArray from the equal-length arrays `x` and `y`, describing `x` as a sensor data and `y` as the time points for each measurement.
`da = xr.DataArray(data=x, coords={'time': y, 'channel': z}, name='sensor')`	Make a 2D DataArray from `x`, `y`, and `z`, where z is the channel names in the sensor data.
`da['time']`	Get all time points at which data was recorded
`da['channel']`	Get the names of all channels
`da.loc[1:1.5]`	Get the sensor data from time points 1-1.5 secs.
`da.loc[1:1.5, :]`	Get the sensor data from time points 1-1.5 secs, and all channels.
`da.loc[1:1.5, ['CHAN-2, 'CHAN-4]]`	Get the sensor data from time points 1-1.5 secs and the channels labeled ‘CHAN-2’ and ‘CHAN-4’.
`da.sel(channel='CHAN-3')`	Get the sensor data across the whole time period from the channel labeled ‘CHAN-3’

Exercises

Example: A DataArray can be made by simply passing a regular numpy array with the data to the xarray DataArray constructor.

data = np.random.random(size = 10)
data

array([0.58489453, 0.30931659, 0.35022568, 0.50525519, 0.83252742,
       0.78679708, 0.66900383, 0.65463909, 0.55992887, 0.37307137])

data_xr = xr.DataArray(data)
data_xr

<xarray.DataArray (dim_0: 10)> Size: 80B
array([0.58489453, 0.30931659, 0.35022568, 0.50525519, 0.83252742,
0.78679708, 0.66900383, 0.65463909, 0.55992887, 0.37307137])
Dimensions without coordinates: dim_0

xarray.DataArray

dim_0: 10

0.5849 0.3093 0.3502 0.5053 0.8325 0.7868 0.669 0.6546 0.5599 0.3731

array([0.58489453, 0.30931659, 0.35022568, 0.50525519, 0.83252742,
0.78679708, 0.66900383, 0.65463909, 0.55992887, 0.37307137])

Coordinates: (0)
Indexes: (0)
Attributes: (0)

Example: When we display the resulting DataArray, we see that there is more information that can be added. That’s the strength and benefit of DataArrays; but we’re not taking advantage of it in the example above. In the following example, we include time information - the month - for which a given data point is recorded. In this hypothetical scenario, it’s the sale of hiking boots in a sportswear store over the course of a year.

months = ['Januar', 'February', 'March', 'April', 'May', 'June', 'July', 'August', 'September', 'October', 'November', 'December']
hiking_boots_sold = np.random.randint(low=2, high = 50, size = len(months))
hiking_boots_sold

array([39,  6,  6, 40,  6, 33, 38, 22, 32, 48, 18,  3])

data_boots = xr.DataArray(
    data=hiking_boots_sold, 
    coords={'month': months}, 
    name='sale_hiking_boots',
)
data_boots

<xarray.DataArray 'sale_hiking_boots' (month: 12)> Size: 96B
array([39,  6,  6, 40,  6, 33, 38, 22, 32, 48, 18,  3])
Coordinates:

month    (month) <U9 432B 'Januar' 'February' … 'November' 'December'

xarray.DataArray

‘sale_hiking_boots’

month: 12

39 6 6 40 6 33 38 22 32 48 18 3

array([39,  6,  6, 40,  6, 33, 38, 22, 32, 48, 18,  3])

Coordinates: (1)

month

(month)

<U9

'Januar' 'February' … 'December'

array(['Januar', 'February', 'March', 'April', 'May', 'June', 'July', 'August',
'September', 'October', 'November', 'December'], dtype='<U9')

Indexes: (1)

month

PandasIndex

PandasIndex(Index(['Januar', 'February', 'March', 'April', 'May', 'June', 'July', 'August',
'September', 'October', 'November', 'December'],
dtype='object', name='month'))

Attributes: (0)

Exercise: Make a DataArray out of the following variables containing numpy arrays.

Solution

days = np.linspace(1,365,365,dtype=int)
hours_of_sunlight = np.random.uniform(low=0, high = 16, size=len(days))

data_sun = xr.DataArray(
    data = hours_of_sunlight,
    coords = {'day': days},
    name = 'hours_of_sunlight_over_year'
)
data_sun

<xarray.DataArray 'hours_of_sunlight_over_year' (day: 365)> Size: 3kB array([8.23006743e+00, 7.21694392e+00, 3.74655378e+00, 8.92461200e+00, 9.38304235e+00, 8.92751010e+00, 3.77495059e+00, 1.31636567e+01, 2.45902953e+00, 8.36089037e+00, 7.09654038e+00, 1.48757403e+01, 1.16164646e+01, 9.80515275e+00, 7.08769855e+00, 8.52507807e+00, 3.11079725e+00, 7.91321694e+00, 1.11614207e+01, 1.48038049e+01, 8.91518155e+00, 5.33508301e-01, 2.47056017e+00, 6.91468774e+00, 1.17036281e+01, 8.77655657e+00, 7.17807369e+00, 1.36831349e+00, 4.61231869e-01, 4.64060336e+00, 6.83088895e+00, 6.06679933e+00, 2.45094377e+00, 7.01031672e-01, 1.11239269e+01, 1.43514494e+01, 3.29001670e-01, 1.02294294e+01, 6.81727355e+00, 9.18526246e+00, 7.79739073e+00, 1.41748651e+01, 1.09333584e+01, 5.92642428e+00, 7.79074386e+00, 4.53065382e-01, 1.12618476e+01, 6.21238801e+00, 1.12555484e+01, 1.20050358e+01, 6.02066390e+00, 1.53692216e+01, 1.59851780e+01, 1.76377694e+00, 1.58066919e+01, 7.57641734e+00, 3.90439098e+00, 9.46307778e+00, 1.02639021e+00, 4.35392115e+00, 3.46629355e+00, 8.00917271e+00, 1.56924200e+01, 3.35086470e+00, 5.81554633e+00, 7.80686597e+00, 8.82128041e+00, 1.51389988e+01, 8.24378132e+00, 5.40782027e+00, 5.95643967e+00, 1.33394170e+00, 1.47696565e+01, 1.47813671e+00, 1.48453840e+01, 6.26934476e+00, 5.41620327e+00, 6.11286275e+00, 6.49687445e+00, 1.08877683e+01, … 9.84325244e+00, 1.01528923e+00, 3.15687929e+00, 9.77263943e+00, 9.96574333e+00, 9.42822240e+00, 6.55878955e-01, 1.17890842e+00, 6.50210208e+00, 1.30461117e+01, 1.27016934e+01, 1.74379897e+00, 5.08618410e+00, 1.11982102e+01, 1.18590474e+01, 2.71737149e+00, 1.46177633e+01, 1.31257279e+01, 1.16233467e+01, 6.16799363e+00, 4.28178084e+00, 7.83740481e+00, 1.29375046e+01, 6.06850805e+00, 1.32126785e+01, 6.39787765e+00, 1.56068444e+01, 1.51391095e+01, 1.51110788e+01, 5.43832742e+00, 8.26582415e+00, 1.87781332e+00, 6.74993425e+00, 9.21053554e+00, 1.35123060e+01, 1.53834993e+01, 1.02323070e+01, 1.28699584e+01, 1.17486530e+01, 4.87753529e+00, 6.55916363e+00, 4.16106549e+00, 1.24088568e+01, 2.53296482e+00, 2.81155073e+00, 9.35439130e+00, 9.83815547e+00, 1.20805543e+01, 9.91077458e+00, 1.17941470e+01, 3.73621124e+00, 6.84984768e-01, 7.11782867e+00, 1.49531769e+01, 1.32211583e+01, 3.99443547e+00, 2.04934036e+00, 1.38590038e+01, 1.17655673e+01, 7.11167642e+00, 4.77290576e+00, 3.36814979e+00, 5.25550318e+00, 7.29273883e+00, 1.57836152e+01, 1.29942957e+01, 1.26501592e+01, 1.38997670e+01, 9.35196810e+00, 1.58813368e+01, 3.86108730e+00, 1.54960821e+01, 1.16770869e+01, 1.38766505e+01, 1.25933240e+01, 2.27569293e+00, 8.90981941e+00]) Coordinates:

day (day) int64 3kB 1 2 3 4 5 6 7 8 … 358 359 360 361 362 363 364 365

xarray.DataArray

‘hours_of_sunlight_over_year’

day: 365

8.23 7.217 3.747 8.925 9.383 8.928 … 11.68 13.88 12.59 2.276 8.91

array([8.23006743e+00, 7.21694392e+00, 3.74655378e+00, 8.92461200e+00,
9.38304235e+00, 8.92751010e+00, 3.77495059e+00, 1.31636567e+01,
2.45902953e+00, 8.36089037e+00, 7.09654038e+00, 1.48757403e+01,
1.16164646e+01, 9.80515275e+00, 7.08769855e+00, 8.52507807e+00,
3.11079725e+00, 7.91321694e+00, 1.11614207e+01, 1.48038049e+01,
8.91518155e+00, 5.33508301e-01, 2.47056017e+00, 6.91468774e+00,
1.17036281e+01, 8.77655657e+00, 7.17807369e+00, 1.36831349e+00,
4.61231869e-01, 4.64060336e+00, 6.83088895e+00, 6.06679933e+00,
2.45094377e+00, 7.01031672e-01, 1.11239269e+01, 1.43514494e+01,
3.29001670e-01, 1.02294294e+01, 6.81727355e+00, 9.18526246e+00,
7.79739073e+00, 1.41748651e+01, 1.09333584e+01, 5.92642428e+00,
7.79074386e+00, 4.53065382e-01, 1.12618476e+01, 6.21238801e+00,
1.12555484e+01, 1.20050358e+01, 6.02066390e+00, 1.53692216e+01,
1.59851780e+01, 1.76377694e+00, 1.58066919e+01, 7.57641734e+00,
3.90439098e+00, 9.46307778e+00, 1.02639021e+00, 4.35392115e+00,
3.46629355e+00, 8.00917271e+00, 1.56924200e+01, 3.35086470e+00,
5.81554633e+00, 7.80686597e+00, 8.82128041e+00, 1.51389988e+01,
8.24378132e+00, 5.40782027e+00, 5.95643967e+00, 1.33394170e+00,
1.47696565e+01, 1.47813671e+00, 1.48453840e+01, 6.26934476e+00,
5.41620327e+00, 6.11286275e+00, 6.49687445e+00, 1.08877683e+01,
…
9.84325244e+00, 1.01528923e+00, 3.15687929e+00, 9.77263943e+00,
9.96574333e+00, 9.42822240e+00, 6.55878955e-01, 1.17890842e+00,
6.50210208e+00, 1.30461117e+01, 1.27016934e+01, 1.74379897e+00,
5.08618410e+00, 1.11982102e+01, 1.18590474e+01, 2.71737149e+00,
1.46177633e+01, 1.31257279e+01, 1.16233467e+01, 6.16799363e+00,
4.28178084e+00, 7.83740481e+00, 1.29375046e+01, 6.06850805e+00,
1.32126785e+01, 6.39787765e+00, 1.56068444e+01, 1.51391095e+01,
1.51110788e+01, 5.43832742e+00, 8.26582415e+00, 1.87781332e+00,
6.74993425e+00, 9.21053554e+00, 1.35123060e+01, 1.53834993e+01,
1.02323070e+01, 1.28699584e+01, 1.17486530e+01, 4.87753529e+00,
6.55916363e+00, 4.16106549e+00, 1.24088568e+01, 2.53296482e+00,
2.81155073e+00, 9.35439130e+00, 9.83815547e+00, 1.20805543e+01,
9.91077458e+00, 1.17941470e+01, 3.73621124e+00, 6.84984768e-01,
7.11782867e+00, 1.49531769e+01, 1.32211583e+01, 3.99443547e+00,
2.04934036e+00, 1.38590038e+01, 1.17655673e+01, 7.11167642e+00,
4.77290576e+00, 3.36814979e+00, 5.25550318e+00, 7.29273883e+00,
1.57836152e+01, 1.29942957e+01, 1.26501592e+01, 1.38997670e+01,
9.35196810e+00, 1.58813368e+01, 3.86108730e+00, 1.54960821e+01,
1.16770869e+01, 1.38766505e+01, 1.25933240e+01, 2.27569293e+00,
8.90981941e+00])

Coordinates: (1)
- day
  (day)
  int64
  1 2 3 4 5 6 … 361 362 363 364 365
```
array([  1,   2,   3, …, 363, 364, 365])
```

Indexes: (1)

day

PandasIndex

PandasIndex(Index([  1,   2,   3,   4,   5,   6,   7,   8,   9,  10,
…
356, 357, 358, 359, 360, 361, 362, 363, 364, 365],
dtype='int64', name='day', length=365))

Attributes: (0)

Exercise: Get the array containing the days throughout the year.

Solution

data_sun['day']

<xarray.DataArray 'day' (day: 365)> Size: 3kB
array([  1,   2,   3, …, 363, 364, 365])
Coordinates:

day      (day) int64 3kB 1 2 3 4 5 6 7 8 … 358 359 360 361 362 363 364 365

xarray.DataArray

‘day’

day: 365

1 2 3 4 5 6 7 8 9 10 11 … 356 357 358 359 360 361 362 363 364 365
```
array([  1,   2,   3, …, 363, 364, 365])
```
Coordinates: (1)
- day
  (day)
  int64
  1 2 3 4 5 6 … 361 362 363 364 365
```
array([  1,   2,   3, …, 363, 364, 365])
```

Indexes: (1)

day

PandasIndex

PandasIndex(Index([  1,   2,   3,   4,   5,   6,   7,   8,   9,  10,
…
356, 357, 358, 359, 360, 361, 362, 363, 364, 365],
dtype='int64', name='day', length=365))

Attributes: (0)

Exercise: Get the data on hours of sunlight for day number 3 through 11 using the loc method.

Solution

data_sun.loc[3:11]

<xarray.DataArray 'hours_of_sunlight_over_year' (day: 9)> Size: 72B
array([ 3.74655378,  8.924612  ,  9.38304235,  8.9275101 ,  3.77495059,
13.16365669,  2.45902953,  8.36089037,  7.09654038])
Coordinates:

day      (day) int64 72B 3 4 5 6 7 8 9 10 11

xarray.DataArray

‘hours_of_sunlight_over_year’

day: 9

3.747 8.925 9.383 8.928 3.775 13.16 2.459 8.361 7.097

array([ 3.74655378,  8.924612  ,  9.38304235,  8.9275101 ,  3.77495059,
13.16365669,  2.45902953,  8.36089037,  7.09654038])

Coordinates: (1)
- day
  (day)
  int64
  3 4 5 6 7 8 9 10 11
```
array([ 3,  4,  5,  6,  7,  8,  9, 10, 11])
```

Indexes: (1)

day

PandasIndex

PandasIndex(Index([3, 4, 5, 6, 7, 8, 9, 10, 11], dtype='int64', name='day'))

Attributes: (0)

Exercise: Get the data on hours of sunlight for day number 3 through 11 using regular indexing for arrays. Do you notice a difference in which indeces you use to access the data?

Solution

data_sun[2:11]

<xarray.DataArray 'hours_of_sunlight_over_year' (day: 9)> Size: 72B
array([ 3.74655378,  8.924612  ,  9.38304235,  8.9275101 ,  3.77495059,
13.16365669,  2.45902953,  8.36089037,  7.09654038])
Coordinates:

day      (day) int64 72B 3 4 5 6 7 8 9 10 11

xarray.DataArray

‘hours_of_sunlight_over_year’

day: 9

3.747 8.925 9.383 8.928 3.775 13.16 2.459 8.361 7.097

array([ 3.74655378,  8.924612  ,  9.38304235,  8.9275101 ,  3.77495059,
13.16365669,  2.45902953,  8.36089037,  7.09654038])

Coordinates: (1)
- day
  (day)
  int64
  3 4 5 6 7 8 9 10 11
```
array([ 3,  4,  5,  6,  7,  8,  9, 10, 11])
```

Indexes: (1)

day

PandasIndex

PandasIndex(Index([3, 4, 5, 6, 7, 8, 9, 10, 11], dtype='int64', name='day'))

Attributes: (0)

Exercise: In the hiking boots DataArray from the example, get the number of hiking boots sold in October using the loc method.

Solution

data_boots.loc['October']

<xarray.DataArray 'sale_hiking_boots' ()> Size: 8B
array(48)
Coordinates:
month    <U9 36B 'October'

Exercise: Creating DataArrays with multidimensional data. Let’s say that the company selling hiking boots has stores in multiple cities - Cologne, Berlin, and Munich - and that you want to store data on sales in all three cities throughout the year. In this case, you’re storing multidimensional data; data across time and space, similar to neuroscience data.

months = ['Januar', 'February', 'March', 'April', 'May', 'June', 'July', 'August', 'September', 'October', 'November', 'December']
cities = ['Cologne', 'Berlin', 'Munich']
hiking_boots_sold = np.random.randint(low=2, high = 50, size = (len(months), len(cities)))
hiking_boots_sold

array([[11, 38, 45],
       [37, 25, 13],
       [ 4, 21, 11],
       [45, 19, 20],
       [ 9, 29, 28],
       [ 8,  5, 49],
       [10, 22, 23],
       [28, 25, 12],
       [43,  5, 28],
       [13, 18,  9],
       [ 6, 30, 33],
       [20, 20, 16]])

data_boots_cities = xr.DataArray(
    data=hiking_boots_sold,
    coords={'month': months, 'city': cities},
    name='hiking_boots_sold_different_cities'
)
data_boots_cities

<xarray.DataArray 'hiking_boots_sold_different_cities' (month: 12, city: 3)> Size: 288B
array([[11, 38, 45],
[37, 25, 13],
[ 4, 21, 11],
[45, 19, 20],
[ 9, 29, 28],
[ 8,  5, 49],
[10, 22, 23],
[28, 25, 12],
[43,  5, 28],
[13, 18,  9],
[ 6, 30, 33],
[20, 20, 16]])
Coordinates:

month    (month) <U9 432B 'Januar' 'February' … 'November' 'December'
city     (city) <U7 84B 'Cologne' 'Berlin' 'Munich'

Exercise: Make a 2-D array with data on sunlight throughout the year in Germany, France, and Italy using the variables in the cell below.

Solution

days = np.linspace(1,365,365,dtype=int)
countries = ['Germany', 'France', 'Italy']
hours_of_sunlight = np.random.uniform(low=0, high = 16, size=(len(days), len(countries)))

data_sun_country = xr.DataArray(
    data = hours_of_sunlight,
    coords={'day': days, 'country': countries},
    name='hours_of_sunlight_countries'
)
data_sun_country

<xarray.DataArray 'hours_of_sunlight_countries' (day: 365, country: 3)> Size: 9kB
array([[10.62911822,  0.13318533,  5.50667079],
[10.20166023, 15.54403471,  6.6744995 ],
[15.91543282,  6.52257226, 13.96066533],
…,
[ 2.5423545 ,  9.90954904, 12.38464112],
[ 3.10759408,  8.16055736, 10.55890302],
[13.99674588,  2.43413591,  2.84435315]])
Coordinates:

day      (day) int64 3kB 1 2 3 4 5 6 7 8 … 358 359 360 361 362 363 364 365
country  (country) <U7 84B 'Germany' 'France' 'Italy'

Exercise: Get the data on hours of sunlight from day 3 to 11 for Italy using the loc function.

Solution

data_sun_country.loc[3:11, 'Italy']

<xarray.DataArray 'hours_of_sunlight_countries' (day: 9)> Size: 72B
array([13.96066533,  1.17941739, 11.09906937, 14.0559088 , 11.17591572,
10.79901602, 12.55835353,  1.38706644,  9.55820968])
Coordinates:

day      (day) int64 72B 3 4 5 6 7 8 9 10 11
country  <U7 28B 'Italy'

Exercise: Get the data on hours of sunlight from day 3 to 11 for both Germany and France together.

Solution

data_sun_country.loc[3:11, ['Germany', 'France']]

<xarray.DataArray 'hours_of_sunlight_countries' (day: 9, country: 2)> Size: 144B
array([[15.91543282,  6.52257226],
[ 3.05188703, 12.36757178],
[ 5.86002647, 12.3537221 ],
[ 9.81929209,  9.54686706],
[11.81800274, 13.86272208],
[ 0.78291768, 11.79843803],
[ 8.8559426 ,  5.45258128],
[14.32486253,  4.11157752],
[ 7.35287893,  8.44928813]])
Coordinates:

day      (day) int64 72B 3 4 5 6 7 8 9 10 11
country  (country) <U7 56B 'Germany' 'France'

Section 2: Saving DataArray to File.

After the DataArray is constructed, you want to save it to a file so that you can load it and continue to work on it later or share it with others.

Code	Description
`da.to_netcdf('data/filename.nc')`	Write the DataArray variable named “da” to a file with a filename of your choosing in the data directory
`data = xr.load_dataarray('data/filename.nc')`	Load the DataArray and put it in a variable

Run the cell below to create the data directory if it doesn’t already exist.

Exercises

from pathlib import Path

data_dir = Path('data')

data_dir.mkdir(exist_ok=True, parents=True)

Exercise: Write the DataArray data on sunlight per day to file to save it.

Solution

#Write to file
data_boots_cities.to_netcdf('data/hiking_boots_sold.nc')

Exercise: Load the DataArray data on sunlight per day in different countries you saved to a variable. Display the variable to check that the data was stored correctly.

Solution

data = xr.load_dataarray('data/hiking_boots_sold.nc')
data

<xarray.DataArray 'hiking_boots_sold_different_cities' (month: 12, city: 3)> Size: 288B
array([[11, 38, 45],
[37, 25, 13],
[ 4, 21, 11],
[45, 19, 20],
[ 9, 29, 28],
[ 8,  5, 49],
[10, 22, 23],
[28, 25, 12],
[43,  5, 28],
[13, 18,  9],
[ 6, 30, 33],
[20, 20, 16]])
Coordinates:

month    (month) <U9 432B 'Januar' 'February' … 'November' 'December'
city     (city) <U7 84B 'Cologne' 'Berlin' 'Munich'