Python surface plotting

Question

I have following table data(Please see the image) for which I want to have a surface plot in python. Using surface plotting from matplotlib,

import matplotlib.pyplot as plt
from matplotlib import cm
from matplotlib.ticker import LinearLocator, FormatStrFormatter
import numpy as np

fig = plt.figure()
ax = fig.gca(projection='3d')
X=[2,3,5,8,20,30,50,80,100,150,175,200,250,300]
Y=[2,3,4,5,10,15,20,30,40,50,80,100,125,150,175,200]
Y,X=np.meshgrid(Y,X)
Z=np.array([
[0.2885307,0.269452,0.259193,0.2548041,0.2731868,0.4801551,0.7992361,1.7577641,3.2611327,5.428839,19.647976,37.59729,78.0871,152.21466,268.14572,0],
[0.2677955,0.2538363,0.2380033,0.2306999,0.4779794,0.9251045,1.5448972,3.508644,6.4968576,11.252151,0,0,0,0,0,0],
[0.2432982,0.2283371,0.2514196,0.3392502,0,0,0,0,0,0,0,0,0,0,0,0],
[0.2342575,0.3158406,0.4770729, 0.6795485,2.353042, 5.260077,9.78172,25.87004,59.52568, 0,0,0,0,0,0,0],
[0.6735384, 1.3873291,2.346506, 3.5654,0,0,0,0,0,0,0,0,0,0,0,0],
[1.3584715, 2.9405127,5.096819,8.155857,0,0,0,0,0,0,0,0,0,0,0,0],
[3.558062,8.216592,15.768077,27.386694,0,0,0,0,0,0,0,0,0,0,0,0],
[9.537899,25.202589,58.20041,0,0,0,0,0,0,0,0,0,0,0,0,0],
[16.083374,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[54.936775,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[89.185974,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]])

my_col = cm.jet(Z/np.amax(Z))
surf = ax.plot_surface(X, Y, Z,cmap=cm.coolwarm,linewidth=0, antialiased=False)
ax.set_zlim(0, 300)
ax.zaxis.set_major_locator(LinearLocator(10))
ax.zaxis.set_major_formatter(FormatStrFormatter('%.02f'))
fig.colorbar(surf, shrink=0.5, aspect=5)
plt.show()

I get a plot like this

which is correct but not very attractive or intuitive. How can I make the visualizations more smooth and clear? Please note that I have many blanks in my data. Should I use 'zero' for the blanks or 'nan'(not a number)? For the same data, excel shows a much better graph. I appreciate your inputs in order to make python plot more visually attractive.

It's OK... I think the error lies in the data. your plot and the expect one look a lot alike. Are you sure all these zeros exist are actualy in the data — Anwarvic, May 12 '20 at 15:10
I didn't understand your comment completely. All these zeros are there. I mean, these data are in the table — Sandeep Parameshwara, May 12 '20 at 15:15
I understand that my table has so many blanks. I am trying to plot a decent graph from these data. I can delete the last 3 rows for example. But is there good options in plotting , colouring style etc? — Sandeep Parameshwara, May 12 '20 at 15:21

tom10 · Accepted Answer · 2020-05-26T03:21:26.803

The difference between the matplotlib and excel plots is that matplotlib is plotting on a linear scale and excel is logarithmic (or something that looks deceptively like a log axis but actually isn't -- see below). Therefore, in the matplotlib the slopes look extremely steep, but in excel the slopes are dramatically stretched out by the log.

Unfortunately, matplotlib doesn't yet have log axes working well in 3D. I'm not sure why this is, but it is a serious shortcoming. You can see a plot similar to Excel though if you take the log10 of your X and Y data before you do the plots. You can also go further to DIY your own log axes, but I've just done a shorthand for that using a tick formatter.

import matplotlib.pyplot as plt
from matplotlib import cm
from matplotlib.ticker import LinearLocator, FormatStrFormatter, FuncFormatter
from mpl_toolkits.mplot3d import axes3d
import numpy as np

def format_log(x, pos=None):
    x1 = 10**x
    s = "%.3f" % x1
    return s[:-4] if s[-3:]=="000" else " "

fig = plt.figure()
ax = fig.gca(projection='3d')
X=[2,3,5,8,20,30,50,80,100,150,175,200,250,300]
Y=[2,3,4,5,10,15,20,30,40,50,80,100,125,150,175,200]
X = np.log10(np.array(X))
Y = np.log10(np.array(Y))
Y,X=np.meshgrid(Y,X)
Z=np.array([
[0.2885307,0.269452,0.259193,0.2548041,0.2731868,0.4801551,0.7992361,1.7577641,3.2611327,5.428839,19.647976,37.59729,78.0871,152.21466,268.14572,0],
[0.2677955,0.2538363,0.2380033,0.2306999,0.4779794,0.9251045,1.5448972,3.508644,6.4968576,11.252151,0,0,0,0,0,0],
[0.2432982,0.2283371,0.2514196,0.3392502,0,0,0,0,0,0,0,0,0,0,0,0],
[0.2342575,0.3158406,0.4770729, 0.6795485,2.353042, 5.260077,9.78172,25.87004,59.52568, 0,0,0,0,0,0,0],
[0.6735384, 1.3873291,2.346506, 3.5654,0,0,0,0,0,0,0,0,0,0,0,0],
[1.3584715, 2.9405127,5.096819,8.155857,0,0,0,0,0,0,0,0,0,0,0,0],
[3.558062,8.216592,15.768077,27.386694,0,0,0,0,0,0,0,0,0,0,0,0],
[9.537899,25.202589,58.20041,0,0,0,0,0,0,0,0,0,0,0,0,0],
[16.083374,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[54.936775,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[89.185974,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]])
my_col = cm.jet(Z/np.amax(Z))
surf = ax.plot_surface(X, Y, Z,cmap=cm.coolwarm)
ax.set_zlim(0, 300)
ax.zaxis.set_major_locator(LinearLocator(10))
ax.zaxis.set_major_formatter(FormatStrFormatter('%.02f'))
ax.xaxis.set_major_formatter(FuncFormatter(format_log))
ax.yaxis.set_major_formatter(FuncFormatter(format_log))
fig.colorbar(surf, shrink=0.5, aspect=5)
plt.show()

Edit:
After coming back to this question, I realize that the Excel plot isn't actually showing a logarithmic axis, but is instead just plotting the X and Y values given with equal spacing along the axis even those values have no clear mathematical progression.

It's critical to note that this isn't a good representation of the data, since it gives the obvious impression that it's logarithmic (for the specific data presented), but it's actually not, although it requires unusually close inspection to see that. Here the gaps between adjacent numbers aren't even monotonic.

So I discourage this representation, but to reproduce that Excel plot, I'd suggest making equally spaced data, but labeling it with different numbers (and just this sentence alone should be enough to discourage this approach). But here's the code and approach:

fig = plt.figure()
ax = fig.gca(projection='3d')
x=[2,3,5,8,20,30,50,80,100,150,175,200,250,300]
y=[2,3,4,5,10,15,20,30,40,50,80,100,125,150,175,200]

Y,X=np.meshgrid(range(len(y)),range(len(x)))
Z=np.array([
[0.2885307,0.269452,0.259193,0.2548041,0.2731868,0.4801551,0.7992361,1.7577641,3.2611327,5.428839,19.647976,37.59729,78.0871,152.21466,268.14572,0],
[0.2677955,0.2538363,0.2380033,0.2306999,0.4779794,0.9251045,1.5448972,3.508644,6.4968576,11.252151,0,0,0,0,0,0],
[0.2432982,0.2283371,0.2514196,0.3392502,0,0,0,0,0,0,0,0,0,0,0,0],
[0.2342575,0.3158406,0.4770729, 0.6795485,2.353042, 5.260077,9.78172,25.87004,59.52568, 0,0,0,0,0,0,0],
[0.6735384, 1.3873291,2.346506, 3.5654,0,0,0,0,0,0,0,0,0,0,0,0],
[1.3584715, 2.9405127,5.096819,8.155857,0,0,0,0,0,0,0,0,0,0,0,0],
[3.558062,8.216592,15.768077,27.386694,0,0,0,0,0,0,0,0,0,0,0,0],
[9.537899,25.202589,58.20041,0,0,0,0,0,0,0,0,0,0,0,0,0],
[16.083374,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[54.936775,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[89.185974,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]])
my_col = cm.jet(Z/np.amax(Z))
surf = ax.plot_surface(X, Y, Z,cmap=cm.coolwarm)
ax.tick_params(axis='both', which='major', labelsize=6)
ax.set_zlim(0, 300)
ax.xaxis.set_major_locator(IndexLocator(1, 0))
ax.xaxis.set_major_formatter(FixedFormatter([repr(v) for v in x]))
ax.yaxis.set_major_locator(IndexLocator(1, 0))
ax.yaxis.set_major_formatter(FixedFormatter([repr(v) for v in y]))
fig.colorbar(surf, shrink=0.5, aspect=5)

If one wanted to show the specific numbers given for X and Y, one solution would be to plot with a logarithmic axis (since the numbers are spaced very approximately in a log way), and then plot the numbers specifically by lines on the axes, or alternatively, don't use these numbers instead of the usual regularly spaced numbers. (But to plot these as axes values, and space them visually at regular intervals, that is a problem.)

Thanks for your answer, I have a small doubt. Is it possible to display all values along X and Y axis instead of just 10 and 100? — Sandeep Parameshwara, May 25 '20 at 09:44
@SandeepParameshwara: Yes, it's possible, but a general solution is difficult. What needs to happen is for matplotlib to do log axes in 3D. (Until then, if you need an answer for a specific plot, it's probably reasonable to try something based on what I have already posted and if you have trouble, come back with a new question with specifics, for example, do you need: to work for all ranges, publication quality placement, nice integral values for the ticks, etc. Maybe someone will do it, but it's too big of a problem to just add to this question and answer.) — tom10, May 25 '20 at 17:02
@SandeepParameshwara: I was looking this over again, and I had made an error in my reading your initially posted plots. I think that error shows exactly why these plots shouldn't be presented in the way they are in your question, and that a true `log10` is the correct approach, but I wanted to correct that error none-the-less, so i have posted it. — tom10, May 26 '20 at 00:32

Python surface plotting

1 Answers1