How to correctly use contourf with logarithmic color scale?

Dear Matlab Community,

I have a 1372 x 4118 (double) matrix I want to plot using contourf() function. The data entries of the matrix vary from 1 to 1e-9.In order to see changes throughout the whole scale I want to use a log scale fo caxis. The following code sets the Colorbar to log scale but the color representation in the plot is not in agreement with the colorbar: ( Due to memory constraints I reduce the matrix to 600 x 600 )

matrix = load('matrix.mat');
matrix = matrix.a;

x= linspace(1,600,600);
y=linspace(-1,-600,600);

contourf(x,y,matrix);
set(gca,'ColorScale','log');

c = colorbar;
c.Label.String = 'energy [keV]';

xlim([0 100]);
ylim([-200,0]);

clim([1e-7,1e0]);

Please find attached an image of the wrong plot with the full matrix, a subset of the matrix , and the matrix itself.

Can someone please help me with this challange?

Best Regards

sethi

[Walter Roberson]

I do not see any problem. With your sample data, the left edge (near x == 0) has a sharp peak to 1, but most of the rest of the data is around roughly 5e-4

[Sebastian]

Dear Mr. Roberson,

thank you for the fast reply. I would expect a more pronounced color gradient than this shown in the attached image (contourf_logscale_2.jpg). In this image, values of 1e-5 have the same greenish color as values of 1e-2. But according to the scale bar there should be a bluish or orange color, respectively

.

Answer 1

[Voss]

The problem is that contourf picks the levels in a linear fashion, so the levels in your contour plot are [0 0.1 0.2 0.3 ... 0.9 1], which obviously don't work very well when the color scale is logarithmic.

One way around this is to use use the log of your data in contourf and then adjust the colorbar tick labels appropriately:

matrix = load('matrix.mat');
matrix = matrix.a;

x = linspace(1,600,600);
y = linspace(-1,-600,600);

figure

contourf(x,y,log10(matrix));

c = colorbar;
c.Label.String = 'energy [keV]';
c.Ticks = -7:0;
c.TickLabels = compose('10^{%d}',c.Ticks);

xlim([0,100]);
ylim([-200,0]);
clim([-7,0]);

Another way is to specify the contour levels yourself:

figure

n_levels = 8;
contourf(x,y,matrix,logspace(-7,0,n_levels));
set(gca,'ColorScale','log');

c = colorbar;
c.Label.String = 'energy [keV]';

xlim([0 100]);
ylim([-200,0]);
clim([1e-7,1e0]);

Surface plot (no contours), for reference:

figure

p = pcolor(x,y,matrix);
p.EdgeColor = 'none';
set(gca,'ColorScale','log');

c = colorbar;
c.Label.String = 'energy [keV]';

xlim([0,100]);
ylim([-200,0]);
clim([1e-7,1e0]);

[Sebastian]

Hello, thank you so much this is exactly what i wanted! This issue is finally solved :)

[Samuel]

Thank you very much! This worked perfectly for a similar issue to OP.

[Voss]

@Samuel: You're welcome! Don't forget to vote for useful answers.

Answer 2

[Nathan Hardenberg]

There are two ways I know of:

1. Define your own (manual) values on which contours should be drawn, or
2. just plot the logarithmic values instead of the linear ones

matrix = load('matrix.mat');
matrix = matrix.a;
x = linspace( 1, 600, 600);
y = linspace(-1,-600, 600);
figure(1);
% using contourf with manual lines
contourf(x, y, matrix, [1 0.1 0.01 0.001 0.0001 0.00001 0]);
set(gca,'ColorScale','log');
c = colorbar;
c.Label.String = 'energy [keV]';
xlim([0 100]); ylim([-200,0]);

figure(2);
contourf(x, y, log10(matrix));  % draw logarithmic values
c = colorbar;
c.Label.String = 'log10(energy) [keV]';
xlim([0 100]); ylim([-200,0]);

[Sebastian]

Thank you for your support!