Matplotlib contour lines are not closing up

Question

I would need some help with my Matplotlib contour plots.

The problem is, that the lines of my contour plot (shown beneath) are not closing up and that I have an image kind of cut in half. I am wondering if I can force Matplotlib to close the contours or if my data has too poor statistics to do so. However, I selected all objects with a value of the x-axis > 10.0 and got 10e6 objects to plot.

I am using:

Python 2.7
Ubuntu 14.04
--> plot printed with PdfPages

The code is the following:

import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
from matplotlib.backends.backend_pdf import PdfPages

x        = np.linspace(min(mydata[:,0]), max(mydata[:,0]), 100)
y        = np.linspace(min(mydata[:,1]), max(mydata[:,1]), 100)

H        = np.histogram2d(mydata[:,0], mydata[:,1], bins=100)       
binsize  = int((H.max()-grid_min)/8)

X,Y      = np.meshgrid(x,y, indexing='ij')

#set_zero=2.0 % for example! Amount of objects which are not included in the plot 
#in order to get a not that wide spread last contour (dark blue in the figure enclosed)!    
grid_min = np.floor(int(H.max()/100.0*set_zero))        

CS=cb_axis.contour(X,Y,H, 8, lw=2.5, ls='-', col='w', 
   levels=np.arange(grid_min, np.ceil(H.max()), binsize))

cb_axis.contourf(X,Y,H, 8, levels=np.arange(grid_min, np.ceil(H.max()), 
   binsize), cmap='coolwarm'))       

plt.colorbar(CS, pad=0.01, aspect=35, use_gridspec=True, format='%0.1e')

How do you expect the contours to close if you throw away part of the data? — Roland Smith, Jan 13 '17 at 12:16
How did you select the data? Is there NaN values at x<10? What happens if you set the values to something small, e.g. 1? — periphreal, Jan 13 '17 at 12:21
@RolandSmith Smith: That is part of what I am trying to find out. Can I force the contours to somehow close that the plot looks better, or is it just not possible. So I know that I did nothing wrong, it is just the data with this kind of selection ... — firefly2517, Jan 13 '17 at 12:26
@periphreal: No there are no NaN values in the selection, I took care of them before ... — firefly2517, Jan 13 '17 at 12:27
You could artificially generate values for a short range, say between 9.9 and 10, with a (log-)linear interpolation. — periphreal, Jan 13 '17 at 12:40
Setting the data to zero where the lines need to close might be a good idea. See my answer below. — ImportanceOfBeingErnest, Jan 13 '17 at 15:00

score 1 · Accepted Answer · answered Jan 13 '17 at 14:59

In principle there can be two reasons for the contour lines not to be closed.

The first is a simple visual appearance. Having white lines on a white background may make the look like disappear. This effect is shown in the plot below in the first image, left side of the data. This effect can be visualized by coloring the contour lines (second image).

There is not much you can do about this, appart from changing the lines' or the background's color.

The second reason, which can be observed in the first two images on the right side of the data, is that the lines are really not closed, because they are at the edge of the data. This is of course independent of the lines' color and is due to matplotlib not knowing what to do outside the data (it's only the human brain that thinks those lines should be closed, but there is no evidence for closed lines in the data).

To solve this second problem, one can extend the data or assign the edge values of the data to a minimum value, such that the lines need to close. This is shown in the third and fourth example image, where I set the last column of the data matrix to zero.

For reference, this is the code to produce the above image.

import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
from matplotlib.backends.backend_pdf import PdfPages

n=90
x        = np.arange(n)
y        = np.arange(n)
X,Y      = np.meshgrid(x,y)
H        = (np.sinc((X-70)/50.)**2*np.sinc((Y-50)/50.)**2)*3600
H[:, :50] = np.zeros((n,50))

grid_min = 10
binsize = H.max()/8.        

fig, (ax, ax2, ax3, ax4) = plt.subplots(ncols=4, figsize=(12,5))
plt.subplots_adjust(left=0.03, right=0.97, wspace=0.03)
levels=np.arange(grid_min, np.ceil(H.max()), binsize)
for a in (ax, ax2, ax3, ax4):
    a.contourf(X,Y,H, levels=levels, cmap='viridis')
    a.set_xlim([40,100])
    a.yaxis.set_visible(False)
    a.xaxis.set_visible(False)

ax.set_title("Open white lines")
ax.text(.076, .5, "Closed white line ", rotation=90, transform=ax.transAxes, va="center")
ax.text(.86, .5, "Open white lines", rotation=-90, transform=ax.transAxes, va="center")     
ax.contour(X,Y,H, levels=levels, linewidths=3, linestyles='-', colors=['w' for i in range(len(levels))]) 

ax2.set_title("Open Colored lines")
ax2.text(.06, .5, "Closed colored lines", rotation=90, transform=ax2.transAxes, va="center")
ax2.text(.86, .5, "Open colored lines", rotation=-90, transform=ax2.transAxes, va="center")
ax2.contour(X,Y,H, levels=levels, linewidths=3, linestyles='-', cmap="jet") 

# Setting the last column of data to zero in order to close the lines
H[:, n-1] = np.zeros(n)
ax3.set_title("Closed White lines")
ax3.text(.076, .5, "Closed white lines", rotation=90, transform=ax3.transAxes, va="center")
ax3.text(.86, .5, "Closed white lines", rotation=-90, transform=ax3.transAxes, va="center")    
ax3.contour(X,Y,H, levels=levels, linewidths=3, linestyles='-', colors=['w' for i in range(len(levels))]) 

ax4.set_title("Closed Colored lines")
ax4.text(.06, .5, "Closed colored lines", rotation=90, transform=ax4.transAxes, va="center")
ax4.text(.86, .5, "Closed colored lines", rotation=-90, transform=ax4.transAxes, va="center")   
ax4.contour(X,Y,H, levels=levels, linewidths=3, linestyles='-', cmap="jet") 

plt.show()

Matplotlib contour lines are not closing up

1 Answers1