Plotting a masked surface plot using python, numpy and matplotlib

Question

I'm plotting a surface using matplotlib 1.1.0.

The plot Z axis is masked like so:

Zm = ma.masked_where((abs(z_grid) < 1.09) & (abs(z_grid) > 0.91), (z_surface))
surf = ax.plot_surface(X, Y,Zm, rstride=2, cstride=2, cmap=colors,linewidth=0, antialiased=False)

But I'm not seeing the mask applied on the plot. I plotted the mask itself as a subplot

surf = ax.plot_surface(X, Y,ma.getmask(Zm), rstride=2, cstride=2, cmap=colors,linewidth=0, antialiased=False)

Which worked, so I know my mask does actually contain True values.

Full code:

from pylab import *
import matplotlib.pyplot as plt
from matplotlib.widgets import Button
import numpy
from mpl_toolkits.mplot3d.axes3d import Axes3D
from  matplotlib import patches
from matplotlib.figure import Figure
from matplotlib import rcParams


fig = plt.figure(figsize=plt.figaspect(0.5))
ax = fig.add_subplot(1, 2, 1,projection='3d')

pole_positions_orig = [-0.6+0.73j];
zero_positions_orig = [0.29-0.41j];

surface_limit = 1.7;
min_val = -surface_limit;
max_val = surface_limit;

surface_resolution = 0.0333;

X = numpy.arange(min_val,max_val,surface_resolution)
Y = numpy.arange(min_val,max_val,surface_resolution)
X, Y = numpy.meshgrid(X, Y)

z_grid = X + Y*1j;
z_surface = z_grid*0;

pole_positions = numpy.round(pole_positions_orig,1) + surface_resolution/2+(surface_resolution/2)*1j;
zero_positions = numpy.round(zero_positions_orig,1) + surface_resolution/2 +(surface_resolution/2)*1j;

for k in range(0, len(zero_positions)):
    z_surface = z_surface + 20*log10((z_grid - zero_positions[k].real - zero_positions[k].imag*1j));
    z_surface = z_surface + 20*log10((z_grid - zero_positions[k].real + zero_positions[k].imag*1j));

for k in range(0, len(pole_positions)):
    z_surface = z_surface - 20*log10((z_grid - pole_positions[k].real - pole_positions[k].imag*1j));
    z_surface = z_surface - 20*log10((z_grid - pole_positions[k].real + pole_positions[k].imag*1j));    


colors = cm.jet;
colors.set_bad('k');


Zm = ma.masked_where((abs(z_grid) < 1.09) & (abs(z_grid) > 0.91), (z_surface))

z_surface = Zm;

surf = ax.plot_surface(X, Y,z_surface, rstride=2, cstride=2, cmap=colors,linewidth=0, antialiased=False)


ticks = [-1, 1]; 
z_ticks = [-30,-20,-10,0,10,20,30]; 
ax.set_xticks(ticks);
ax.set_yticks(ticks);   
ax.set_zticks(z_ticks);

ax.set_xlabel('Re')
ax.set_ylabel('Im')
ax.set_zlabel('Mag(db)',ha='left')
plt.setp(ax.get_zticklabels(), fontsize=7)
plt.setp(ax.get_xticklabels(), fontsize=7)  
plt.setp(ax.get_yticklabels(), fontsize=7)

ax = fig.add_subplot(1, 2, 2,projection='3d')
surf = ax.plot_surface(X, Y,ma.getmask(z_surface), rstride=2, cstride=2, cmap=colors,linewidth=0, antialiased=False)

ax.grid(b=None);
show();

This is what I have:

python plot

This is what I want (from matlab):

matlab plot

What am I missing?

So the masked data should still be drawn, but in a solid color? — fraxel, Jun 13 '12 at 19:03
yes, is this possible? ... If not, then just not drawing the mask is a good compromise. I played around with "colors.set_bad('k',alpha=0.5)" to give that a try, but it didn't change the plot at all. — stanri, Jun 13 '12 at 19:31
Don't think its going to be possible with masking, looks like `plot_surface()` [doesn't respect masks](https://github.com/matplotlib/matplotlib/issues/487). Probably is possible through a clever workaround, but its beating me at the moment :( — fraxel, Jun 14 '12 at 07:54
I've added an answer to the question explaining how I got around the problem :) It's not ideal, but it worked for my purposes. — stanri, Jun 14 '12 at 18:15

stanri · Answer 1 · 2012-06-15T05:04:03.787

Fraxel mentioned that surface_plot doesn't support masking. In order to get around the issue, this is what I did:

I basically manually masked the z axis data by setting every masked value to numpy.nan like so:

Zm = ma.masked_where((abs(z_grid) < 1.02) & (abs(z_grid) > 0.98), (z_surface))
z_surface[where(ma.getmask(Zm)==True)] = numpy.nan

Cmap Broken

However, it messed up my colormap scaling. To fix that, I did this:

cmap = cm.jet
lev = numpy.arange(-30,30,1);
norml = colors.BoundaryNorm(lev, 256)

surf = ax.plot_surface(X, Y, z_surface,...,norm = norml)

Fixed

Not 100% what I wanted, but a good compromise nonetheless.

score 1 · Accepted Answer · answered Feb 22 '15 at 10:19

You can do it, but you need to do it by manually colouring the surface faces yourself;

the cmap function takes a nubmer between 0 and 1, so we just need to normalise the values before calling the cmap function on them.

z_surface = numpy.real(z_surface)
min_z, max_z = z_surface.min(), z_surface.max()
colours = numpy.zeros_like(z_surface, dtype=object)

for i in range(len(z_surface)):
  for j in range(len(z_surface[0])):
    if 0.91 < numpy.sqrt(X[i,j]**2 + Y[i,j]**2) < 1.09:
      colours[i,j] = "red"  
    else:
      colours[i,j] = plt.get_cmap("jet")((z_surface[i,j]-min_z) / (max_z - min_z))


surf = ax.plot_surface(X, Y, z_surface, rstride=2, cstride=2, facecolors=colours, linewidth=0, antialiased=False)

enter image description here

I should also point out that matplotlib is casting your z array to real - whether or not you are taking advantage of this on purpose though i don't know.

Plotting a masked surface plot using python, numpy and matplotlib

2 Answers2

Linked