Multidimensional groupby

doc/api.rst

@@ -109,6 +109,7 @@ Computation
    Dataset.apply
    Dataset.reduce
    Dataset.groupby
+   Dataset.groupby_bins
    Dataset.resample
    Dataset.diff
 
@@ -245,6 +246,7 @@ Computation
 
    DataArray.reduce
    DataArray.groupby
+   DataArray.groupby_bins
    DataArray.rolling
    DataArray.resample
    DataArray.get_axis_num

doc/examples.rst

@@ -7,3 +7,4 @@ Examples
     examples/quick-overview
     examples/weather-data
     examples/monthly-means
+    examples/multidimensional-coords

doc/examples/multidimensional-coords.rst

@@ -0,0 +1,200 @@
+
+Working with Multidimensional Coordinates
+=========================================
+
+Author: `Ryan Abernathey <http://github.org/rabernat>`__
+
+Many datasets have *physical coordinates* which differ from their
+*logical coordinates*. Xarray provides several ways to plot and analyze
+such datasets.
+
+.. code:: python
+
+    %matplotlib inline
+    import numpy as np
+    import pandas as pd
+    import xarray as xr
+    import cartopy.crs as ccrs
+    from matplotlib import pyplot as plt
+
+    print("numpy version  : ", np.__version__)
+    print("pandas version : ", pd.__version__)
+    print("xarray version   : ", xr.version.version)
+
+
+.. parsed-literal::
+
+    ('numpy version  : ', '1.11.0')
+    ('pandas version : ', u'0.18.0')
+    ('xarray version   : ', '0.7.2-32-gf957eb8')
+
+
+As an example, consider this dataset from the
+`xarray-data <https://github.com/pydata/xarray-data>`__ repository.
+
+.. code:: python
+
+    ! curl -L -O https://github.com/pydata/xarray-data/raw/master/RASM_example_data.nc
+
+.. code:: python
+
+    ds = xr.open_dataset('RASM_example_data.nc')
+    ds
+
+
+
+
+.. parsed-literal::
+
+    <xarray.Dataset>
+    Dimensions:  (time: 36, x: 275, y: 205)
+    Coordinates:
+      * time     (time) datetime64[ns] 1980-09-16T12:00:00 1980-10-17 ...
+        yc       (y, x) float64 16.53 16.78 17.02 17.27 17.51 17.76 18.0 18.25 ...
+        xc       (y, x) float64 189.2 189.4 189.6 189.7 189.9 190.1 190.2 190.4 ...
+      * x        (x) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ...
+      * y        (y) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ...
+    Data variables:
+        Tair     (time, y, x) float64 nan nan nan nan nan nan nan nan nan nan ...
+    Attributes:
+        title: /workspace/jhamman/processed/R1002RBRxaaa01a/lnd/temp/R1002RBRxaaa01a.vic.ha.1979-09-01.nc
+        institution: U.W.
+        source: RACM R1002RBRxaaa01a
+        output_frequency: daily
+        output_mode: averaged
+        convention: CF-1.4
+        references: Based on the initial model of Liang et al., 1994, JGR, 99, 14,415- 14,429.
+        comment: Output from the Variable Infiltration Capacity (VIC) model.
+        nco_openmp_thread_number: 1
+        NCO: 4.3.7
+        history: history deleted for brevity
+
+
+
+In this example, the *logical coordinates* are ``x`` and ``y``, while
+the *physical coordinates* are ``xc`` and ``yc``, which represent the
+latitudes and longitude of the data.
+
+.. code:: python
+
+    print(ds.xc.attrs)
+    print(ds.yc.attrs)
+
+
+.. parsed-literal::
+
+    OrderedDict([(u'long_name', u'longitude of grid cell center'), (u'units', u'degrees_east'), (u'bounds', u'xv')])
+    OrderedDict([(u'long_name', u'latitude of grid cell center'), (u'units', u'degrees_north'), (u'bounds', u'yv')])
+
+
+Plotting
+--------
+
+Let's examine these coordinate variables by plotting them.
+
+.. code:: python
+
+    fig, (ax1, ax2) = plt.subplots(ncols=2, figsize=(14,4))
+    ds.xc.plot(ax=ax1)
+    ds.yc.plot(ax=ax2)
+
+
+
+
+.. parsed-literal::
+
+    <matplotlib.collections.QuadMesh at 0x118688fd0>
+
+
+
+.. parsed-literal::
+
+    /Users/rpa/anaconda/lib/python2.7/site-packages/matplotlib/collections.py:590: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
+      if self._edgecolors == str('face'):
+
+
+
+.. image:: multidimensional_coords_files/xarray_multidimensional_coords_8_2.png
+
+
+Note that the variables ``xc`` (longitude) and ``yc`` (latitude) are
+two-dimensional scalar fields.
+
+If we try to plot the data variable ``Tair``, by default we get the
+logical coordinates.
+
+.. code:: python
+
+    ds.Tair[0].plot()
+
+
+
+
+.. parsed-literal::
+
+    <matplotlib.collections.QuadMesh at 0x11b6da890>
+
+
+
+
+.. image:: multidimensional_coords_files/xarray_multidimensional_coords_10_1.png
+
+
+In order to visualize the data on a conventional latitude-longitude
+grid, we can take advantage of xarray's ability to apply
+`cartopy <http://scitools.org.uk/cartopy/index.html>`__ map projections.
+
+.. code:: python
+
+    plt.figure(figsize=(14,6))
+    ax = plt.axes(projection=ccrs.PlateCarree())
+    ax.set_global()
+    ds.Tair[0].plot.pcolormesh(ax=ax, transform=ccrs.PlateCarree(), x='xc', y='yc', add_colorbar=False)
+    ax.coastlines()
+    ax.set_ylim([0,90]);
+
+
+
+.. image:: multidimensional_coords_files/xarray_multidimensional_coords_12_0.png
+
+
+Multidimensional Groupby
+------------------------
+
+The above example allowed us to visualize the data on a regular
+latitude-longitude grid. But what if we want to do a calculation that
+involves grouping over one of these physical coordinates (rather than
+the logical coordinates), for example, calculating the mean temperature
+at each latitude. This can be achieved using xarray's ``groupby``
+function, which accepts multidimensional variables. By default,
+``groupby`` will use every unique value in the variable, which is
+probably not what we want. Instead, we can use the ``groupby_bins``
+function to specify the output coordinates of the group.
+
+.. code:: python
+
+    # define two-degree wide latitude bins
+    lat_bins = np.arange(0,91,2)
+    # define a label for each bin corresponding to the central latitude
+    lat_center = np.arange(1,90,2)
+    # group according to those bins and take the mean
+    Tair_lat_mean = ds.Tair.groupby_bins('xc', lat_bins, labels=lat_center).mean()
+    # plot the result
+    Tair_lat_mean.plot()
+
+
+
+
+.. parsed-literal::
+
+    [<matplotlib.lines.Line2D at 0x11cb92e90>]
+
+
+
+
+.. image:: multidimensional_coords_files/xarray_multidimensional_coords_14_1.png
+
+
+Note that the resulting coordinate for the ``groupby_bins`` operation
+got the ``_bins`` suffix appended: ``xc_bins``. This help us distinguish
+it from the original multidimensional variable ``xc``.

doc/groupby.rst

@@ -14,10 +14,11 @@ __ http://www.jstatsoft.org/v40/i01/paper
 - Combine your groups back into a single data object.
 
 Group by operations work on both :py:class:`~xarray.Dataset` and
-:py:class:`~xarray.DataArray` objects. Currently, you can only group by a single
-one-dimensional variable (eventually, we hope to remove this limitation). Also,
-note that for one-dimensional data, it is usually faster to rely on pandas'
-implementation of the same pipeline.
+:py:class:`~xarray.DataArray` objects. Most of the examples focus on grouping by
+a single one-dimensional variable, although support for grouping
+over a multi-dimensional variable has recently been implemented. Note that for
+one-dimensional data, it is usually faster to rely on pandas' implementation of
+the same pipeline.
 
 Split
 ~~~~~
@@ -63,6 +64,33 @@ You can also iterate over over groups in ``(label, group)`` pairs:
 Just like in pandas, creating a GroupBy object is cheap: it does not actually
 split the data until you access particular values.
 
+Binning
+~~~~~~~
+
+Sometimes you don't want to use all the unique values to determine the groups
+but instead want to "bin" the data into coarser groups. You could always create
+a customized coordinate, but xarray facilitates this via the
+:py:meth:`~xarray.Dataset.groupby_bins` method.
+
+.. ipython:: python
+
+    x_bins = [0,25,50]
+    ds.groupby_bins('x', x_bins).groups
+
+The binning is implemented via `pandas.cut`__, whose documentation details how
+the bins are assigned. As seen in the example above, by default, the bins are
+labeled with strings using set notation to precisely identify the bin limits. To
+override this behavior, you can specify the bin labels explicitly. Here we
+choose `float` labels which identify the bin centers:
+
+.. ipython:: python
+
+    x_bin_labels = [12.5,37.5]
+    ds.groupby_bins('x', x_bins, labels=x_bin_labels).groups
+
+__ http://pandas.pydata.org/pandas-docs/version/0.17.1/generated/pandas.cut.html
+
+
 Apply
 ~~~~~
 
@@ -149,3 +177,31 @@ guarantee that all original dimensions remain unchanged.
 
 You can always squeeze explicitly later with the Dataset or DataArray
 :py:meth:`~xarray.DataArray.squeeze` methods.
+
+Multidimensional Grouping
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Many datasets have a multidimensional coordinate variable (e.g. longitude)
+which is different from the logical grid dimensions (e.g. nx, ny). Such
+variables are valid under the `CF conventions`__. Xarray supports groupby
+operations over multidimensional coordinate variables:
+
+__ http://cfconventions.org/cf-conventions/v1.6.0/cf-conventions.html#_two_dimensional_latitude_longitude_coordinate_variables
+
+.. ipython:: python
+
+    da = xr.DataArray([[0,1],[2,3]],
+        coords={'lon': (['ny','nx'], [[30,40],[40,50]] ),
+                'lat': (['ny','nx'], [[10,10],[20,20]] ),},
+        dims=['ny','nx'])
+    da
+    da.groupby('lon').sum()
+    da.groupby('lon').apply(lambda x: x - x.mean(), shortcut=False)
+
+Because multidimensional groups have the ability to generate a very large
+number of bins, coarse-binning via :py:meth:`~xarray.Dataset.groupby_bins`
+may be desirable:
+
+.. ipython:: python
+
+    da.groupby_bins('lon', [0,45,50]).sum()