Fork me on GitHub


Related Topics

This Page

Shading ExampleΒΆ

  • ../../_images/sphx_glr_shading_example_001.png
  • ../../_images/sphx_glr_shading_example_002.png
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.colors import LightSource
from matplotlib.cbook import get_sample_data

# Example showing how to make shaded relief plots
# like Mathematica
# (
# or Generic Mapping Tools
# (

def main():
    # Test data
    x, y = np.mgrid[-5:5:0.05, -5:5:0.05]
    z = 5 * (np.sqrt(x**2 + y**2) + np.sin(x**2 + y**2))

    filename = get_sample_data('jacksboro_fault_dem.npz', asfileobj=False)
    with np.load(filename) as dem:
        elev = dem['elevation']

    fig = compare(z,
    fig.suptitle('HSV Blending Looks Best with Smooth Surfaces', y=0.95)

    fig = compare(elev,, ve=0.05)
    fig.suptitle('Overlay Blending Looks Best with Rough Surfaces', y=0.95)

def compare(z, cmap, ve=1):
    # Create subplots and hide ticks
    fig, axes = plt.subplots(ncols=2, nrows=2)
    for ax in axes.flat:
        ax.set(xticks=[], yticks=[])

    # Illuminate the scene from the northwest
    ls = LightSource(azdeg=315, altdeg=45)

    axes[0, 0].imshow(z, cmap=cmap)
    axes[0, 0].set(xlabel='Colormapped Data')

    axes[0, 1].imshow(ls.hillshade(z, vert_exag=ve), cmap='gray')
    axes[0, 1].set(xlabel='Illumination Intensity')

    rgb = ls.shade(z, cmap=cmap, vert_exag=ve, blend_mode='hsv')
    axes[1, 0].imshow(rgb)
    axes[1, 0].set(xlabel='Blend Mode: "hsv" (default)')

    rgb = ls.shade(z, cmap=cmap, vert_exag=ve, blend_mode='overlay')
    axes[1, 1].imshow(rgb)
    axes[1, 1].set(xlabel='Blend Mode: "overlay"')

    return fig

if __name__ == '__main__':

Total running time of the script: ( 0 minutes 0.265 seconds)

Gallery generated by Sphinx-Gallery