Xarray

Parallel Programming with Dask in Python

James Fulton

Climate Informatics Researcher

Xarray - like pandas in more dimensions

pandas

Applies index labels to tabular data

Xarray

Applies index labels to high dimensional arrays

DataFrame

A DataFrame which is made up of four columns.

DataSet

A DataSet which is made up of three DataArrays

Loading a DataSet from Zarr

import xarray as xr
ds = xr.open_zarr("data/era_eu.zarr")


print(ds)

<xarray.Dataset>
Dimensions:  (lat: 30, lon: 45, time: 504)
Coordinates:
  * lat      (lat) float64 35.5 36.5 37.5 38.5 39.5 ... 60.5 61.5 62.5 63.5 64.5
  * lon      (lon) float64 -14.5 -13.5 -12.5 -11.5 -10.5 ... 26.5 27.5 28.5 29.5
  * time     (time) datetime64[ns] 1979-05-31 1979-06-30 ... 2021-04-30
Data variables:
    precip   (time, lat, lon) float32 dask.array<chunksize=(12, 15, 15), ... >
    temp     (time, lat, lon) float32 dask.array<chunksize=(12, 15, 15), ... >

DataFrame vs. DataSet

pandas DataFrame

# Select a particular date
df.loc['2020-01-01']

# Select by index number
df.iloc[0]

# Select column
df['column1']

Dask DataSet

# Select a particular date
ds.sel(time='2020-01-01')

# Select by index number
ds.isel(time=0)

# Select variable
ds['variable1']

DataFrame vs. DataSet

pandas DataFrame

# Perform mathematical operations
df.mean()




# Groupby and mean
df.groupby(df['time'].dt.year).mean()


# Rolling mean
rolling_mean = df.rolling(5).mean()

Dask DataSet

# Perform mathematical operations
ds.mean()
ds.mean(dim='dim1')
ds.mean(dim=('dim1', 'dim2'))


# Groupby and mean
ds.groupby(ds['time'].dt.year).mean()


# Rolling mean
rolling_mean = ds.rolling(dim1=5).mean()

rolling_mean.compute()

Plotting

ds['variable'].plot()

Makes a line plot if 1D
Makes a heatmap if 2D
Makes a histogram if 3D+

Example An example line plot shows the temperature in one location in northern Finland over a few years around the Great Cold Outbreak.

Plotting

ds['variable'].plot()

Makes a line plot if 1D
Makes a heatmap if 2D
Makes a histogram if 3D+

Example An example heatmap shows the average temperature across Europe in January.

Plotting

ds['variable'].plot()

Makes a line plot if 1D
Makes a heatmap if 2D
Makes a histogram if 3D+

Example An example histogram shows the distribution in temperatures measured in all locations around Europe.

Let's practice!

Parallel Programming with Dask in Python