xrspatial.proximity.direction#

xrspatial.proximity.direction(raster: DataArray, x: str = 'x', y: str = 'y', target_values: list = None, max_distance: float = inf, distance_metric: str = 'EUCLIDEAN') → DataArray[source]#

Calculates, for all cells in the array, the downward slope direction Calculates, for all pixels in the input raster, the direction to nearest source based on a set of target values and a distance metric.

This function attempts to calculate for each cell, the the direction, in degrees, to the nearest source. The output values are based on compass directions, where 90 is for the east, 180 for the south, 270 for the west, 360 for the north, and 0 for the source cell itself. 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 direction will be computed in pixels.

Direction support NumPy backed, and Dask with NumPy backed xarray DataArray. The return values of direction 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.

Similar to proximity, the implementation for NumPy-backed is ported from GDAL, which uses a dynamic programming approach to identify nearest target of a pixel from its surrounding neighborhood in a 3x3 window The implementation for Dask-backed uses dask.map_overlap to compute proximity direction 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 direction is computed toward the target with the lowest flat (row-major) index, 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 reported direction is deterministic regardless of which backend computes it.

Parameters:

raster (xr.DataArray or xr.Dataset) – 2D array image with raster.shape = (height, width). 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, proximity 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 direction 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., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0.],
    [0., 0., 0., 0., 0.],
    [1., 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 direction
>>> direction_agg = direction(raster)
>>> direction_agg
<xarray.DataArray (y: 5, x: 5)>
array([[ 45.      ,  26.56505 , 360.      , 333.43494 , 315.      ],
       [ 63.434948,  45.      , 360.      , 315.      , 296.56506 ],
       [ 90.      ,  90.      ,   0.      , 270.      , 270.      ],
       [360.      , 135.      , 180.      , 225.      , 243.43495 ],
       [  0.      , 270.      , 180.      , 206.56505 , 225.      ]],
      dtype=float32)
Coordinates:
  * y        (y) int64 4 3 2 1 0
  * x        (x) int64 0 1 2 3 4