Numpy-based vectorized geospatial functions

pip install geog==0.0.2



A pure numpy implementation for geodesic functions. The interfaces are vectorized according to numpy broadcasting rules compatible with a variety of inputs including lists, numpy arrays, and Shapely geometries - allowing for 1-to-1, N-to-1, or the element-wise N-to-N calculations in a single call.

geog uses a spherical Earth model (subject to change) with radius 6371.0 km.

geog draws inspiration from TurfJS


  • distance - Compute the distance in meters between any number of longitude,latitude points
  • course - Compute the forward azimuth between points
  • propagate - Starting from some points and pointing azimuths, move some distance and compute the final points.

Getting Started

Compute the distance in meters between two locations on the surface of the Earth.

>>> import geog

>>> boston = [-71.0589, 42.3601]
>>> la = [-118.2500, 34.0500]

>>> geog.distance(boston, la)

>>> geog.course(boston, la)

geog allows different sizes of inputs conforming to numpy broadcasting rules

Compute the distances from several points to one point.

>>> dc = [-77.0164, 38.9047]
>>> paris = [2.3508, 48.8567]
>>> geog.distance([boston, la, dc], paris)
array([ 5531131.56144631,  9085960.07227854,  6163490.48394848])

Compute the element-wise distance of several points to several points

>>> sydney = [151.2094, -33.865]
>>> barcelona = [2.1833, 41.3833]
>>> geog.distance([boston, la, dc], [paris, sydney, barcelona])
array([  5531131.56144631,  12072666.9425518 ,   6489222.58111716])

geog functions can take numpy arrays as inputs

>>> import numpy as np
>>> points = np.array([boston, la, dc])
>>> points
array([[ -71.0589,   42.3601],
       [-118.25  ,   34.05  ],
       [ -77.0164,   38.9047]])
>>> geog.distance(points, sydney)
array([ 16239763.03982447,  12072666.9425518 ,  15711932.63508411])

geog functions can also take Shapely geometries as inputs

>>> import shapely.geometry
>>> p = shapely.geometry.Point([-90.0667, 29.9500])
>>> geog.distance(points, p)
array([ 2185738.94680724,  2687705.07260978,  1554066.84579387])

Other Uses

Use propagate to buffer a single point by passing in multiple angles.

>>> n_points = 6
>>> d = 100  # meters
>>> angles = np.linspace(0, 360, n_points)
>>> polygon = geog.propagate(p, angles, d)

Compute the length of a line over the surface.

>>> np.sum(geog.distance(line[:-1,:], line[1:,:]))

Quick Documentation

distance(p0, p1, deg=True)

course(p0, p1, deg=True, bearing=False)

propagate(p0, angle, d, deg=True, bearing=False)

For all of the above, p0 or p1 can be:

  • single list, tuple, or Shapely Point of [lon, lat] coordinates
  • list of [lon, lat] coordinates or Shapely Points
  • N x 2 numpy array of (lon, lat) coordinates

If argument deg is False, then all angle arguments, coordinates and azimuths, will be used as radians. If deg is False in course(), then it's output will also be radians.

Consult the documentation on each function for more detailed descriptions of the arguments.


  • All points, or point-like objects assume a longitude, latitude ordering.
  • Arrays of points have shape N x 2.
  • Azimuth/course is measured with 0 degrees as due East, increasing counter-clockwise so that 90 degrees is due North. The functions that operate on azimuth accept a bearing=True argument to use the more traditional definition where 0 degrees is due North increasing clockwise such that that 90 degrees is due East.


geog is hosted on PyPI.

pip install geog

See also