# xarray.computation.rolling.DataArrayRolling.reduce
#### DataArrayRolling.reduce(func, keep_attrs=None, , sliding_window_view_kwargs=None, \*\*kwargs)[[source]](https://github.com/pydata/xarray/blob/main/xarray/computation/rolling.py#L513-L616)
Reduce each window by applying func.
Equivalent to `.construct(...).reduce(func, ...)`.
* **Parameters:**
* **func** ([`callable()`](https://docs.python.org/3/builtins/functions.html#callable)) – Function which can be called in the form
func(x, \*\*kwargs) to return the result of collapsing an
np.ndarray over the rolling dimension.
* **keep_attrs** ([`bool`](https://docs.python.org/3/builtins/functions.html#bool), *default*: [`None`](https://docs.python.org/3/builtins/constants.html#None)) – If True, the attributes (`attrs`) will be copied from the original
object to the new one. If False, the new object will be returned
without attributes. If None uses the global default.
* **sliding_window_view_kwargs** – Keyword arguments that should be passed to the underlying array type’s
`sliding_window_view` function.
* **\*\*kwargs** ([`dict`](https://docs.python.org/3/builtins/stdtypes.html#dict)) – Additional keyword arguments passed on to func.
* **Returns:**
**reduced** ([`DataArray`](xarray.DataArray.html.md#xarray.DataArray)) – Array with summarized data.
#### SEE ALSO
[`numpy.lib.stride_tricks.sliding_window_view`](https://numpy.org/doc/stable/reference/generated/numpy.lib.stride_tricks.sliding_window_view.html#numpy.lib.stride_tricks.sliding_window_view), [`dask.array.lib.stride_tricks.sliding_window_view`](https://docs.dask.org/en/latest/generated/dask.array.lib.stride_tricks.sliding_window_view.html#dask.array.lib.stride_tricks.sliding_window_view)
### Notes
With dask arrays, it’s possible to pass the `automatic_rechunk` kwarg as
`sliding_window_view_kwargs={"automatic_rechunk": True}`. This controls
whether dask should automatically rechunk the output to avoid
exploding chunk sizes. Automatically rechunking is the default behaviour.
Importantly, each chunk will be a view of the data so large chunk sizes are
only safe if *no* copies are made later.
### Examples
```pycon
>>> da = xr.DataArray(np.arange(8).reshape(2, 4), dims=("a", "b"))
>>> rolling = da.rolling(b=3)
>>> rolling.construct("window_dim")
Size: 192B
array([[[nan, nan, 0.],
[nan, 0., 1.],
[ 0., 1., 2.],
[ 1., 2., 3.]],
[[nan, nan, 4.],
[nan, 4., 5.],
[ 4., 5., 6.],
[ 5., 6., 7.]]])
Dimensions without coordinates: a, b, window_dim
```
```pycon
>>> rolling.reduce(np.sum)
Size: 64B
array([[nan, nan, 3., 6.],
[nan, nan, 15., 18.]])
Dimensions without coordinates: a, b
```
```pycon
>>> rolling = da.rolling(b=3, min_periods=1)
>>> rolling.reduce(np.nansum)
Size: 64B
array([[ 0., 1., 3., 6.],
[ 4., 9., 15., 18.]])
Dimensions without coordinates: a, b
```