python - matplotlib contour/contourf of **concave** non-gridded data

Question

I'd like to do a matplotlib contour or contourf plot of non-gridded 3D data (x, y, z) which is somehow C-shaped in x and y (see sketch) -- therefore part of the enclosing hull around the data is concave in x and y.

Usually I do plots of non-gridded 3D data by first interpolating it with

from matplotlib.mlab import griddata
griddata...

but this generates artifacts in the concave part of the data such that the concave part is filled by the interpolation.

Is it possible to do the interpolating or contourf/contour plots such that the concave part of the data is respected?

Answer 1

Masking by condition

Below is an example on how to use tricontourf with masking to obtain a concave shape without interpolated portions outside the data. It relies on the ability to mask the data depending on a condition.

import matplotlib.pyplot as plt
import matplotlib.tri as tri
import numpy as np

# create some data
rawx = np.random.rand(500)
rawy = np.random.rand(len(rawx))
cond01 = (rawx-1)**2 + rawy**2 <=1
cond02 = (rawx-0.7)**2 + rawy**2 >0.3
x = rawx[cond01 & cond02]
y = rawy[cond01 & cond02]
f = lambda x,y: np.sin(x*4)+np.cos(y)
z = f(x,y)
# now, x,y are points within a partially concave shape

triang0 = tri.Triangulation(x, y)
triang = tri.Triangulation(x, y)
x2 = x[triang.triangles].mean(axis=1) 
y2 = y[triang.triangles].mean(axis=1)
#note the very obscure mean command, which, if not present causes an error.
#now we need some masking condition.
# this is easy in this case where we generated the data according to the same condition
cond1 = (x2-1)**2 + y2**2 <=1
cond2 = (x2-0.7)**2 + (y2)**2 >0.3
mask = np.where(cond1 & cond2,0,1)
# apply masking
triang.set_mask(mask)


fig, (ax, ax2) = plt.subplots(ncols=2, figsize=(6,3))
ax.set_aspect("equal")
ax2.set_aspect("equal")

ax.tricontourf(triang0, z,  cmap="Oranges")
ax.scatter(x,y, s=3, color="k")

ax2.tricontourf(triang, z,  cmap="Oranges")
ax2.scatter(x,y, s=3, color="k")

ax.set_title("tricontourf without mask")
ax2.set_title("tricontourf with mask")
ax.set_xlim(0,1)
ax.set_ylim(0,1)
ax2.set_xlim(0,1)
ax2.set_ylim(0,1)

plt.show()

Masking by maximum side-length

If you do not have access to the exact condition, but have a maximum side-length (distance) between points, the following would be a solution. It would mask out all triangles for which at least one side is longer than some maximum distance. This can be well applied if the point density is rather high.

import matplotlib.pyplot as plt
import matplotlib.tri as tri
import numpy as np

# create some data
rawx = np.random.rand(500)
rawy = np.random.rand(len(rawx))
cond01 = (rawx-1)**2 + rawy**2 <=1
cond02 = (rawx-0.7)**2 + rawy**2 >0.3
x = rawx[cond01 & cond02]
y = rawy[cond01 & cond02]
f = lambda x,y: np.sin(x*4)+np.cos(y)
z = f(x,y)
# now, x,y are points within a partially concave shape

triang1 = tri.Triangulation(x, y)
triang2 = tri.Triangulation(x, y)
triang3 = tri.Triangulation(x, y)

def apply_mask(triang, alpha=0.4):
    # Mask triangles with sidelength bigger some alpha
    triangles = triang.triangles
    # Mask off unwanted triangles.
    xtri = x[triangles] - np.roll(x[triangles], 1, axis=1)
    ytri = y[triangles] - np.roll(y[triangles], 1, axis=1)
    maxi = np.max(np.sqrt(xtri**2 + ytri**2), axis=1)
    # apply masking
    triang.set_mask(maxi > alpha)

apply_mask(triang2, alpha=0.1)
apply_mask(triang3, alpha=0.3)

fig, (ax1, ax2, ax3) = plt.subplots(ncols=3, figsize=(9,3))

ax1.tricontourf(triang1, z,  cmap="Oranges")
ax1.scatter(x,y, s=3, color="k")

ax2.tricontourf(triang2, z,  cmap="Oranges")
ax2.scatter(x,y, s=3, color="k")

ax3.tricontourf(triang3, z,  cmap="Oranges")
ax3.scatter(x,y, s=3, color="k")

ax1.set_title("tricontourf without mask")
ax2.set_title("with mask (alpha=0.1)")
ax3.set_title("with mask (alpha=0.3)")

for ax in (ax1, ax2, ax3):
    ax.set(xlim=(0,1), ylim=(0,1), aspect="equal")

plt.show()

As can be seen, finding the correct parameter (alpha) here is may need some tweaking.