In [1]:
name = '2019-02-14-matplotlib-subplots-callum'
title = 'Better plotting with matplotlib subplots'
tags = 'matplotlib, dataviz'
author = 'Callum Rollo'
In [2]:
from nb_tools import connect_notebook_to_post
from IPython.core.display import HTML
html = connect_notebook_to_post(name, title, tags, author)
Today we'll be using matplotlib's pyplot to make clearer, prettier figures.
First we import packages and generate some data to plot.
In [2]:
import numpy as np
import matplotlib.pyplot as plt
Pump up the font sizes on plots. Better to do it now than when you need 10 plots for a powerpoint!
In [3]:
plt.rcParams.update({"font.size": 20})
More fine grained control can be exerted with rc from matplotlib
In [4]:
SMALL_SIZE = 16
MED_SIZE = 20
LARGE_SIZE = 26
plt.rc('font', size=MED_SIZE)
plt.rc('xtick', labelsize=SMALL_SIZE)
plt.rc('ytick', labelsize=SMALL_SIZE)
plt.rc('axes', titlesize=MED_SIZE)
plt.rc('legend', fontsize=MED_SIZE)
Make 9 sine curves with random offsets and noise.
In [5]:
time = np.linspace(0, np.pi, 50)
no_of_lines = 9
x1 = np.empty((no_of_lines, len(time)))
for row in range(no_of_lines):
x1[row, :] = np.sin(time + 2 * np.pi * np.random.rand()) + 0.2 * np.random.rand(len(time))
In [6]:
plt.plot(time, x1[0, :])
Out[6]:
Now plot all of them
In [7]:
for row in range(no_of_lines):
plt.plot(time, x1[row, :])
What a mess! Better off with some subplots
In [10]:
# Using subplots to make a 3 X 3 grid of axes for plotting
fig, axs = plt.subplots(3, 3, figsize=(8, 8))
# ravel makes the axs object into an array of 9 objects we can access
# with indices, very handy for plotting within a loop
axs = axs.ravel()
for row in range(no_of_lines):
axs[row].plot(time, x1[row, :])
We can improve this by sharing axis labels and reducing the space between plots
In [11]:
fig, axs = plt.subplots(3, 3, figsize=(8, 8), sharex="col", sharey="row")
axs = axs.ravel()
fig.subplots_adjust(hspace=0.05, wspace=0.05)
for row in range(no_of_lines):
axs[row].plot(time, x1[row, :]);
Subplots don't have to be identical. If we want we can change the relative sizes with gridspec. Especailly useful if you are mixing line and contour plots into one superfigure
In [12]:
# Making a grid of temperature data with meshgrid
delta1 = 0.025
x1 = np.sin(np.arange(np.pi / 4, 3 * np.pi / 4, delta1))
y1 = np.arange(10, 20, delta1)
X1, Y1 = np.meshgrid(x1, y1)
temp = X1 * Y1
distance = np.linspace(0, 90, len(x1))
depth = 20 - y1
In [13]:
fig, axs = plt.subplots(
2,
2,
figsize=(8, 8),
sharex="col",
sharey="row",
gridspec_kw={"height_ratios": [1, 4], "width_ratios": [4, 1]},
)
fig.subplots_adjust(hspace=0, wspace=0)
axs = axs.ravel()
# Plotting the surface temperature and mean temp depth profile alongside the contour plot of temp
axs[0].plot(distance, temp[-1, :], label="surface temp")
axs[2].contourf(distance, depth, temp)
axs[3].plot(np.nanmean(temp, 1), depth)
# Setting limits to keep things tight
axs[2].set(xlim=[distance[0], distance[-1]], ylim=[depth[0], depth[-1]])
axs[3].set(xlim=[np.min(temp), np.max(temp)]);
Some more tricks to make the plot look better
In [14]:
fig, axs = plt.subplots(2, 2, figsize=(8, 8), sharex="col", sharey="row",
gridspec_kw={"height_ratios": [1, 4], "width_ratios": [4, 1]})
fig.subplots_adjust(hspace=0, wspace=0)
axs = axs.ravel()
axs[0].plot(distance, temp[-1, :], label="surface temp")
p1 = axs[2].contourf(distance, depth, temp)
axs[3].plot(np.nanmean(temp, 1), depth)
# Add a colorbar to the right of the bottom right plot
fig.colorbar(ax=axs[3], mappable=p1, label=r"Temperature $\mathrm{^{\circ}C}$")
# Add a label to the first subplot
axs[0].legend()
# Remove the lines and ticks from the unused top right axis
axs[1].axis("off")
# Addding labels and limits
axs[2].set(
xlim=[distance[0], distance[-1]],
ylim=[depth[0], depth[-1]],
ylabel="Depth (m)",
xlabel="Distance (km)",
)
axs[3].set(xlim=[np.min(temp), np.max(temp)], xlabel="mean temp");
We can use this method to make split level plots, if for instance you want to capture near surface variability in greater detail
In [15]:
fig,axs = plt.subplots(2,1,figsize=(8,8),sharex = 'col',
gridspec_kw = {'height_ratios':[1, 2]})
fig.subplots_adjust(hspace = 0, wspace=0)
axs = axs.ravel()
split_level = 0.1
# Plotting the surface temperature and mean temp depth profile alongside the contour plot of temp
axs[0].contourf(distance,depth,temp)
axs[0].set(ylim=[split_level,depth[-1]])
axs[1].contourf(distance,depth,temp)
axs[1].set(ylim=[depth[0],split_level])
# Setting limits to keep things tight
axs[1].set(xlim=[distance[0],distance[-1]])
Out[15]:
A few more examples can be found in this post: More on subplots in matplotlib.
In [13]:
HTML(html)
Out[13]: