python - Contour plot masked on the basis of grid

Question

I am trying to generate a contour plot based on (x,y) coordinates surface with a cube in it that dosent include any data z . Following is its scatterplot.

I use the following code to generate a mesh and interpolate data to plot such a contour map. I try to mask the interpolated data Zi but it still gives me an unmasked contour plot. I also tried to mask x and y coordinates but that dosent do any good.

x = centre_unadj['X [mm]']
y = centre_unadj['Y [mm]']
z = centre_unadj['LDA1-RMS [m/s]']

plt.figure(num=None, figsize=(20, 15), dpi=80, facecolor='w', edgecolor='k')
xi,yi = np.meshgrid(x,y)
mask =(yi> 0) & (yi< 25) & (xi > -53) & (xi < -25) 
#mask_xi = (xi > -53) & (xi < -25) 
#mask_yi = (yi> 0) & (yi< 25)
#yi = ma.masked_array(yi,mask =(yi> 0) & (yi< 25) )
#xi = ma.masked_array(xi,mask=((xi > -53) & (xi < -25) ))
zi = scipy.interpolate.griddata((x,y), z, (xi, yi) , method='cubic')
zi = ma.masked_array(zi, mask = ((yi> 0) & (yi< 25) & (xi > -53) & (xi < -25)) )

#zi[mask]=np.nan

plt.contourf( xi,yi,zi,100)

plt.colorbar()
plt.show()

This is the plot I get after running the above code.

I just dont want any contour interpolation inside the cubic area where there are no datapoints.

Answer 1

The problem is in the meshgrid generation. The values in y go from 0 to 40 more than a dozen times. Thus, the generated xi and yi will be really unintuitive matrixes.

The proper way of generating the meshgrid is the following:

xi,yi = np.meshgrid(np.linspace(x.min(),x.max(),200),np.linspace(y.min(),y.max(),200))

Example

I have generated some data with a similar shape:

import scipy.signal as sgn
import scipy.interpolate as intr
import numpy.ma as ma
x = np.linspace(-100,0,500)
y = sgn.sawtooth(2 * np.pi * .2 * x)
mask = (x>-50) & (x<-25)
y[mask] = (sgn.sawtooth(2 * np.pi * .2 * x[mask])+1)/2
y = (y+1)*25
plt.plot(x,y)
z = np.sin(2*np.pi*.1*x)+np.sin(2*np.pi*.1*y)

Such that the plot x vs y looks like:

The code you are actually using generates the following plot:

xi,yi = np.meshgrid(x,y)
mask =(yi> 0) & (yi< 25) & (xi > -53) & (xi < -25) 
zi = intr.griddata((x,y), z, (xi, yi) , method='cubic')
zi = ma.masked_array(zi, mask = mask )
plt.contourf( xi,yi,zi,100); plt.colorbar()

The interpolation to obtain the grid data yields unexpected and incorrect results, which result in the obtained contourf. In fact, plotting plt.imshow(mask) reveals the positions in the matrix where the values inside the square (y > 0) & (y < 25) & (x > -53) & (x < -25) are placed in the matrix.

When the meshgrid is defined as proposed, the result is this one instead: