matplotlib subplots with no spacing, restricted figure size and tight_layout()

Question

I am trying to make a plot which has two subplots which share the x axis and shall have no space in between them. I followd the Create adjacent subplots example from the matplotlib gallery. However, my plot needs to have a fixed size and this makes everything complicated. If I just follow the example and add a fixed size figure size, then the labels are cut off. If I include the labels by using tight_layout , then the plots are spaced. How to fix this? Also, the title should be closer to the legend. Any help is much appreciated!

Example program, comment out tight_layout to see the difference.

import numpy as np                                                                                             
import matplotlib.pyplot as plt                                                                                
                                                                                                               
x_min = -2*np.pi                                                                                               
x_max = 2*np.pi                                                                                                
resolution = 101                                                                                               
x_vals = np.linspace(x_min, x_max, resolution)                                                                 
y_upper = np.cos(x_vals)                                                                                       
y_lower = -np.cos(x_vals)                                                                                      
data3 = np.sin(x_vals)                                                                                         
                                                                                                               
fig = plt.figure(figsize=(80/25.4, 80/25.4))  # figsize is needed for later usage of the plot                  
ax = fig.subplots(2, 1, sharex=True)                                                                           
fig.subplots_adjust(hspace=0)                                                                                  
                                                                                                               
ax[0].plot(x_vals, y_upper, label="data 1")                                                                    
ax[0].plot(x_vals, y_lower, label="data 2")                                                                    
                                                                                                               
ax[1].set_xlim([x_min,x_max])                                                                                  
ax[0].set_ylim([-1.6,1.6])                                                                                     
ax[1].set_ylim([-1.3,1.3])                                                                                     
                                                                                                               
ax[1].plot(x_vals, data3, ls='-', label="data 3", color='C2')                                                  
                                                                                                               
ax[1].set_xlabel("xaxis")                                                                                      
ax[0].set_ylabel("yaxis 1")                                                                                    
ax[1].set_ylabel("yaxis 2")                                                                                    
ax[0].legend(bbox_to_anchor=(0, 1.02, 1., 0.102), loc='lower left', ncol=2, mode="expand", borderaxespad=0)    
                                                                                                               
fig.suptitle("Title")                                                                                          
fig.tight_layout()  # comment this out to see the difference                                                   
# fig.savefig('figure.png')                                                                                    
plt.show()

Answer 1

You need to use a GridSpec instead of subplots_adjust() , that way tight_layout() will know that you want zero-space and it keep it that way.

In fact, you are already creating a GridSpec when you use fig.subplots() , so you just need to pass some extra parameter in gridspec_kw=

x_min = -2*np.pi                                                                                               
x_max = 2*np.pi                                                                                                
resolution = 101                                                                                               
x_vals = np.linspace(x_min, x_max, resolution)                                                                 
y_upper = np.cos(x_vals)                                                                                       
y_lower = -np.cos(x_vals)                                                                                      
data3 = np.sin(x_vals)                                                                                         
                                                                                                               
fig = plt.figure(figsize=(80/25.4, 80/25.4))  # figsize is needed for later usage of the plot             
#
# This is the line that changes. Instruct the gridspec to have zero vertical pad
#     
ax = fig.subplots(2, 1, sharex=True, gridspec_kw=dict(hspace=0))                                                                           
                                                                                          
ax[0].plot(x_vals, y_upper, label="data 1")                                                                    
ax[0].plot(x_vals, y_lower, label="data 2")                                                                    
                                                                                                               
ax[1].set_xlim([x_min,x_max])                                                                                  
ax[0].set_ylim([-1.6,1.6])                                                                                     
ax[1].set_ylim([-1.3,1.3])                                                                                     
                                                                                                               
ax[1].plot(x_vals, data3, ls='-', label="data 3", color='C2')                                                  
                                                                                                               
ax[1].set_xlabel("xaxis")                                                                                      
ax[0].set_ylabel("yaxis 1")                                                                                    
ax[1].set_ylabel("yaxis 2")                                                                                    
ax[0].legend(bbox_to_anchor=(0, 1.02, 1., 0.102), loc='lower left', ncol=2, mode="expand", borderaxespad=0)    
                                                                                                               
fig.suptitle("Title")                                                                                          
fig.tight_layout()  # Now tight_layout does not add padding between axes
# fig.savefig('figure.png')                                                                                    
plt.show()

Answer 2

It can be frustrating to get precise results with subplots - using gridspec ( https://matplotlib.org/3.3.3/tutorials/intermediate/gridspec.html ) will give your greater precision.

However, given where you are, I think you can get what you want with this:

import matplotlib.pyplot as plt                                                                                
                                                                                                               
x_min = -2*np.pi                                                                                               
x_max = 2*np.pi                                                                                                
resolution = 101                                                                                               
x_vals = np.linspace(x_min, x_max, resolution)                                                                 
y_upper = np.cos(x_vals)                                                                                       
y_lower = -np.cos(x_vals)                                                                                      
data3 = np.sin(x_vals)                                                                                         
                                                                                                               
fig = plt.figure(figsize=(80/25.4, 80/25.4))  # figsize is needed for later usage of the plot                  
ax = fig.subplots(3, 1, sharex=True)                                                                           
fig.subplots_adjust(hspace=0)                                                                                  
ax[0].text(0,0.5,"Title", ha='center')
ax[0].axis("off")
ax[1].plot(x_vals, y_upper, label="data 1")                                                                    
ax[1].plot(x_vals, y_lower, label="data 2")                                                                    
                                                                                                               
ax[2].set_xlim([x_min,x_max])                                                                                  
ax[1].set_ylim([-1.6,1.6])                                                                                     
ax[2].set_ylim([-1.3,1.3])                                                                                     
                                                                                                               
ax[2].plot(x_vals, data3, ls='-', label="data 3", color='C2')                                                  
                                                                                                               
ax[2].set_xlabel("xaxis")                                                                                      
ax[1].set_ylabel("yaxis 1")                                                                                    
ax[2].set_ylabel("yaxis 2")                                                                                    
ax[1].legend(bbox_to_anchor=(0, 1.02, 1., 0.102), loc='lower left', ncol=2, mode="expand", borderaxespad=0)    

#fig.tight_layout()  # comment this out to see the difference                                                   
# fig.savefig('figure.png')                                                                                    
plt.show()

Answer 3

Of course, gridspec is the correct approach, and if you are in early phases of the script writing, you should adapt this . However, if you want an easy fix, you could also move fig.subplots_adjust() :

#...
fig.suptitle("Title")                                                                                          
fig.tight_layout()    
fig.subplots_adjust(hspace=0)                                            
# fig.savefig('figure.png')                                                                                    
plt.show()

Saved image: