[Python] Seaborn heatmap with a logarithmic colorbar with each color resembling one order of...

I´m trying to plot some data in a heatmap. The expected data can be in the range between 1x100 and 1x109, so I´d like to use a discrete logarithmic color scale where one color represents one order of magnitude.

Some example data:

1 9.233444e+05 1.667961e+05 1.000000e+00 1.000000 1.628000e+05 49113.404001 1.297632e+06 1.417425e+08 9.874216e+06 6.604796e+05 244640.784032 2.644814e+06 302680.552383 4.179854e+05 157001.632229
2 1.035594e+06 1.000000e+00 1.000000e+00 1.000000 1.678359e+05 53792.826077 1.546623e+06 1.605406e+08 1.360309e+07 8.274078e+05 387068.818619 3.309055e+06 327772.433710 5.212390e+05 183668.570530
3 1.757615e+06 2.674736e+05 1.000000e+00 21193.964624 2.619884e+05 87328.604686 1.382115e+06 2.223598e+08 1.332000e+07 1.164365e+06 477190.744769 6.037433e+06 574947.946448 9.707206e+05 339096.608121
4 1.568719e+06 2.586641e+05 1.000000e+00 1.000000 2.269648e+05 111880.599266 1.559659e+06 2.214848e+08 1.681052e+07 1.013042e+06 365251.506398 5.524255e+06 498559.533033 8.148302e+05 273630.913931
5 1.836254e+06 3.793271e+05 1.000000e+00 1.000000 1.000000e+00 164330.522936 2.023138e+06 2.859675e+08 1.741852e+07 2.131141e+06 597542.798838 4.079047e+06 347385.192677 8.431494e+05 194838.752708
6 2.384388e+06 1.000000e+00 1.000000e+00 1.000000 4.542583e+05 262315.667507 2.814922e+06 3.351289e+08 1.002064e+07 5.739982e+05 67494.501506 1.013531e+06 965788.960452 7.560733e+04 483717.390066
7 3.025885e+06 1.626263e+06 5.567249e+06 278075.276104 5.024219e+05 510559.593751 2.269265e+06 1.052652e+07 1.000000e+00 1.000000e+00 863660.428694 1.188858e+07 1.000000 2.953666e+06 1.000000
8 3.665986e+06 1.704403e+06 7.424446e+06 173460.962223 9.440780e+05 175174.671543 1.000000e+00 1.864523e+06 1.000000e+00 1.000000e+00 179608.650232 2.177289e+06 1.000000 5.171294e+05 1.000000
9 4.206076e+06 1.971293e+06 8.298467e+06 835034.598981 1.675579e+06 1.000000 1.569502e+06 1.000000e+00 1.000000e+00 1.000000e+00 1.000000 6.342325e+05 1.000000 1.000000e+00 1.000000
10 4.972394e+06 2.009257e+06 7.716523e+06 670068.164179 7.422010e+05 145980.973202 3.615038e+06 1.048249e+06 1.000000e+00 1.000000e+00 1.000000 4.702494e+05 1.000000 1.000000e+00 1.000000


I wrote the following code:

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.colors import LogNorm

data=pd.read_excel('example_data.xlsx')

custom_color_map=sns.color_palette("rocket", as_cmap=False, n_colors=9)
fig, ax = plt.subplots(figsize=(20,10))
ax = sns.heatmap(data, norm=LogNorm(), annot=True, cmap = custom_color_map, linewidths=0.05, linecolor='grey')

plt.tight_layout()

plt.show()



This resulted in the following heatmap:



The lower end of the colorbar is as desired, the first color starts at 100. But then, the second color does not start at 101, but somewhere below 101. This trend goes on for the following colors.

The main problem, however, is that the mapping of values to colors is unexpected, which can be seen looking at the annotated values: In column 0 of the heatmap, there are the values 6.3x105 and 9.2x105 in the same orange color as the following 1x106, etc. Then, the last value in this column, 5x106, is has a different color, despite being in the same order of magnitude as 1x106.

Why is this and how can I adjust the color mapping so that one color matches one order of magnitude?

Continue reading...