为了补充@abevieiramota的精彩答案，你可以用constrained_layout得到等价的tight_layout。如果你使用imshow而不是pcolormesh，你仍然会得到很大的水平间隙，因为imshow施加了1:1的纵横比。

import numpy as np
import matplotlib.pyplot as plt

fig, axes = plt.subplots(nrows=2, ncols=2, constrained_layout=True)
for ax in axes.flat:
    im = ax.pcolormesh(np.random.random((10,10)), vmin=0, vmax=1)

fig.colorbar(im, ax=axes.flat)
plt.show()

I noticed that almost every solution posted involved ax.imshow(im, ...) and did not normalize the colors displayed to the colorbar for the multiple subfigures. The im mappable is taken from the last instance, but what if the values of the multiple im-s are different? (I'm assuming these mappables are treated in the same way that the contour-sets and surface-sets are treated.) I have an example using a 3d surface plot below that creates two colorbars for a 2x2 subplot (one colorbar per one row). Although the question asks explicitly for a different arrangement, I think the example helps clarify some things. I haven't found a way to do this using plt.subplots(...) yet because of the 3D axes unfortunately.

如果我能以更好的方式定位颜色条就好了……(可能有更好的方法来做到这一点，但至少它应该不会太难遵循。)

import matplotlib
from matplotlib import cm
import matplotlib.pyplot as plt
import numpy as np
from mpl_toolkits.mplot3d import Axes3D

cmap = 'plasma'
ncontours = 5

def get_data(row, col):
    """ get X, Y, Z, and plot number of subplot
        Z > 0 for top row, Z < 0 for bottom row """
    if row == 0:
        x = np.linspace(1, 10, 10, dtype=int)
        X, Y = np.meshgrid(x, x)
        Z = np.sqrt(X**2 + Y**2)
        if col == 0:
            pnum = 1
        else:
            pnum = 2
    elif row == 1:
        x = np.linspace(1, 10, 10, dtype=int)
        X, Y = np.meshgrid(x, x)
        Z = -np.sqrt(X**2 + Y**2)
        if col == 0:
            pnum = 3
        else:
            pnum = 4
    print("\nPNUM: {}, Zmin = {}, Zmax = {}\n".format(pnum, np.min(Z), np.max(Z)))
    return X, Y, Z, pnum

fig = plt.figure()
nrows, ncols = 2, 2
zz = []
axes = []
for row in range(nrows):
    for col in range(ncols):
        X, Y, Z, pnum = get_data(row, col)
        ax = fig.add_subplot(nrows, ncols, pnum, projection='3d')
        ax.set_title('row = {}, col = {}'.format(row, col))
        fhandle = ax.plot_surface(X, Y, Z, cmap=cmap)
        zz.append(Z)
        axes.append(ax)

## get full range of Z data as flat list for top and bottom rows
zz_top = zz[0].reshape(-1).tolist() + zz[1].reshape(-1).tolist()
zz_btm = zz[2].reshape(-1).tolist() + zz[3].reshape(-1).tolist()
## get top and bottom axes
ax_top = [axes[0], axes[1]]
ax_btm = [axes[2], axes[3]]
## normalize colors to minimum and maximum values of dataset
norm_top = matplotlib.colors.Normalize(vmin=min(zz_top), vmax=max(zz_top))
norm_btm = matplotlib.colors.Normalize(vmin=min(zz_btm), vmax=max(zz_btm))
cmap = cm.get_cmap(cmap, ncontours) # number of colors on colorbar
mtop = cm.ScalarMappable(cmap=cmap, norm=norm_top)
mbtm = cm.ScalarMappable(cmap=cmap, norm=norm_btm)
for m in (mtop, mbtm):
    m.set_array([])

# ## create cax to draw colorbar in
# cax_top = fig.add_axes([0.9, 0.55, 0.05, 0.4])
# cax_btm = fig.add_axes([0.9, 0.05, 0.05, 0.4])
cbar_top = fig.colorbar(mtop, ax=ax_top, orientation='vertical', shrink=0.75, pad=0.2) #, cax=cax_top)
cbar_top.set_ticks(np.linspace(min(zz_top), max(zz_top), ncontours))
cbar_btm = fig.colorbar(mbtm, ax=ax_btm, orientation='vertical', shrink=0.75, pad=0.2) #, cax=cax_btm)
cbar_btm.set_ticks(np.linspace(min(zz_btm), max(zz_btm), ncontours))

plt.show()
plt.close(fig)
## orientation of colorbar = 'horizontal' if done by column

上面的答案很好，但大多数都使用fig.colobar()方法应用于fig对象。这个例子展示了如何使用plt.colobar()函数，直接应用到pyplot:

def shared_colorbar_example():
    fig, axs = plt.subplots(nrows=3, ncols=3)
    for ax in axs.flat:
        plt.sca(ax)
        color = np.random.random((10))
        plt.scatter(range(10), range(10), c=color, cmap='viridis', vmin=0, vmax=1)
    plt.colorbar(ax=axs.ravel().tolist(), shrink=0.6)
    plt.show()

shared_colorbar_example()

因为上面的大多数答案都演示了在2D矩阵上的使用，所以我使用了一个简单的散点图。收缩关键字是可选的，可以调整颜色条的大小。

如果没有指定vmin和vmax，该方法将自动分析所有子图，以获得颜色条上使用的最小值和最大值。当使用fig.colorbar(im)时，上述方法只扫描作为颜色条最小和最大值参数的图像。

结果:

只需将颜色条放在它自己的轴上，并使用subplots_adjust为它腾出空间。

举个简单的例子:

import numpy as np
import matplotlib.pyplot as plt

fig, axes = plt.subplots(nrows=2, ncols=2)
for ax in axes.flat:
    im = ax.imshow(np.random.random((10,10)), vmin=0, vmax=1)

fig.subplots_adjust(right=0.8)
cbar_ax = fig.add_axes([0.85, 0.15, 0.05, 0.7])
fig.colorbar(im, cax=cbar_ax)

plt.show()

请注意，即使值的范围是由vmin和vmax设置的，颜色范围也将由最后绘制的图像(产生im的图像)设置。例如，如果另一个图具有更高的最大值，则比im的最大值更高的点将以统一的颜色显示。

使用make_axes甚至更简单，并且可以得到更好的结果。它还提供了自定义颜色条位置的可能性。 还要注意共享x轴和y轴的子图选项。

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

fig, axes = plt.subplots(nrows=2, ncols=2, sharex=True, sharey=True)
for ax in axes.flat:
    im = ax.imshow(np.random.random((10,10)), vmin=0, vmax=1)

cax,kw = mpl.colorbar.make_axes([ax for ax in axes.flat])
plt.colorbar(im, cax=cax, **kw)

plt.show()

您可以使用带有坐标轴列表的figure.colorbar()的axparameter简化Joe Kington的代码。 从文档中可以看到:

斧头 没有|父轴对象，其中新的色条轴的空间将被窃取。如果给出了坐标轴列表，它们将全部调整大小，以便为颜色条坐标轴腾出空间。

import numpy as np
import matplotlib.pyplot as plt

fig, axes = plt.subplots(nrows=2, ncols=2)
for ax in axes.flat:
    im = ax.imshow(np.random.random((10,10)), vmin=0, vmax=1)

fig.colorbar(im, ax=axes.ravel().tolist())

plt.show()