xrspatial.proximity.allocation

xrspatial.proximity.allocation(raster: DataArray, x: str = 'x', y: str = 'y', target_values: list = None, max_distance: float = inf, distance_metric: str = 'EUCLIDEAN') → DataArray

Calculates, for all pixels in the input raster, the nearest source based on a set of target values and a distance metric.

This function attempts to produce the value of nearest feature of all pixels in the image to a set of pixels in the source image. The following options are used to define the behavior of the function. By default all non-zero pixels in raster.values will be considered as”target”, and all allocation will be computed in pixels.

Allocation supports NumPy backed, and Dask with NumPy backed xarray DataArray. The return values of allocation are of the same type as the input type. If input raster is a NumPy-backed DataArray, the result is NumPy-backed. If input raster is a Dask-backed DataArray, the result is Dask-backed.

allocation uses the same approach as proximity, which is ported from GDAL. A dynamic programming approach is used for identifying nearest target of a pixel from its surrounding neighborhood in a 3x3 window. The implementation for Dask-backed uses dask.map_overlap to compute allocation chunk by chunk by expanding the chunk’s borders to cover the max_distance.

Tie-breaking: when two or more targets are exactly equidistant from a pixel, the target with the lowest flat (row-major) index wins, i.e. the first target encountered when scanning the raster top-to-bottom and left-to-right. This policy is identical across all backends (numpy, cupy, dask+numpy, dask+cupy), so the allocated value is deterministic regardless of which backend computes it.

Parameters:

• raster (xr.DataArray or xr.Dataset) – 2D array of target data. If a Dataset is passed, the function is applied to each data variable independently, returning a Dataset. The 1D x and y coordinates must be monotonic (strictly increasing or strictly decreasing); a non-monotonic axis raises a ValueError.

• x (str, default='x') – Name of x-coordinates.

• y (str, default='y') – Name of y-coordinates.

• target_values (list) – Target pixel values to measure the distance from. If this option is not provided, allocation will be computed from non-zero pixel values. All entries must be finite; a non-finite value (inf or nan) raises ValueError.

• max_distance (float, default=np.inf) –

  The maximum distance to search. Proximity distances greater than this value will be set to NaN. Must be a non-negative, non-NaN number; a negative or NaN value raises a ValueError. Should be given in the same distance unit as input. For example, if input raster is in lat-lon and distances between points within the raster is calculated using Euclidean distance metric, max_distance should also be provided in lat-lon unit. If using Great Circle distance metric, and thus all distances is in meters, max_distance should also be provided in meters.

  When scaling with Dask, whether the function scales well depends on the max_distance value. If max_distance is infinite by default, this function only works on a single machine. It should scale well, however, if max_distance is relatively small compared to the maximum possible distance in two arbitrary points in the input raster. Note that if max_distance is equal or larger than the max possible distance between 2 arbitrary points in the input raster, the input data array will be rechunked.

• distance_metric (str, default='EUCLIDEAN') – The metric for calculating distance between 2 points. Valid distance metrics are: ‘EUCLIDEAN’, ‘GREAT_CIRCLE’, and ‘MANHATTAN’. An unrecognized value raises ValueError.

Returns:

If raster is a DataArray, returns a DataArray. If raster is a Dataset, returns a Dataset with each variable processed independently. 2D array of allocation values. All other input attributes are preserved.

Return type:

xr.DataArray or xr.Dataset

References

• OSGeo: OSGeo/gdal # noqa

Examples

>>> import numpy as np
>>> import xarray as xr
>>> data = np.array([
    [0., 0., 0., 0., 0.],
    [0., 1., 0., 2., 0.],
    [0., 0., 3., 0., 0.],
    [0., 0., 0., 0., 0.],
    [0., 0., 0., 0., 0.]
])
>>> n, m = data.shape
>>> raster = xr.DataArray(data, dims=['y', 'x'], name='raster')
>>> raster['y'] = np.arange(n)[::-1]
>>> raster['x'] = np.arange(m)

>>> from xrspatial import allocation
>>> allocation_agg = allocation(raster)
>>> allocation_agg
<xarray.DataArray (y: 5, x: 5)>
array([[1., 1., 2., 2., 2.],
       [1., 1., 1., 2., 2.],
       [1., 1., 3., 2., 2.],
       [1., 3., 3., 3., 2.],
       [3., 3., 3., 3., 3.]])
Coordinates:
  * y        (y) int64 4 3 2 1 0
  * x        (x) int64 0 1 2 3 4