Why reshaping the matrix gives entirely different plots while using pcolor()

Hello,

I was trying to plot 3 variables with two of them representing the X and Y axes while the third one is used to denote the color, using the pcolor function. I however find that whenever I change the dimensions of the matrix by reshaping them and try plotting again, I get entirely different plots. I was wondering why that is happening - after all its the same datapoints which are being plotted against each other everytime (no matter what the dimension of the matrix is) , how come a mere change in the dimensions of the matrix give different results?

Following is the code which I tried (given for example):

M = randi([1 100],13, 21);                % original matrices
N = randi([200 500],13, 21);
P = randi([700 1000],13, 21);

M1 = reshape(M,3,[]);                     % reshaped matrices from the original matrices
N1 = reshape(N,3,[]);
P1 = reshape(P,3,[]);

M2 = reshape(M,39,[]);                   % reshaped matrices from the original matrices
N2 = reshape(N,39,[]);
P2 = reshape(P,39,[]);

figure(1);
pcolor(M,N,P);
figure(2);
pcolor(M1, N1, P1);
figure(3);
pcolor(M2, N2, P2);

Thank you.

Answer 1

[Walter Roberson]

pcolor() is equivalent to surf() followed by view(90) .

surf() does not use the input data directly to compute face color. Instead, surf() uses the input data as indicating the vertices, and does a bilinear interpolation to get the value to use for the face color.

When you reshape() you change the relative order of the nodes, and so change which vertices are combined for a particular face. But because of the interpolation to get the actual face color, that changes the face colors.

[Mathan]

Thank you. Is there any alternative to pcolor which I could use that would give me the same plots regardless of the dimension of the matrix?

The reason is because I am trying to plot the temperature profile of a planet’s atmosphere with altitude (x axis is time, y axis is altitude and the colors are temperature at a particular altitude at a particular time). The data is originally collected and stored as a 120 x 780 matrix, plotting which would give me a temperature profile. But when I reshape the matrix to some other dimensions I sometimes get to see more clear trends and features. These trends and features varies (both with time and altitude) when I change the matrix dimension and I am not sure which one is the correct temperature profile. Sometimes the plots look great and sometimes they dont (depending on the reshaped dimensions).

Thank you for any additional suggestions.

[Walter Roberson]

If you have a regular rectangular grid, then use image() or imagesc() or imshow()

If you do not have a regular rectangular grid, then use warp() to do texture mapping.

[Mathan]

Thank you. But is it possible to use either of these when you wanted to have one data inside the first variable on your x axis ( say the matrix M), data inside the another one on the y axis (say the matrix N) and the data inside the third one to denote color (say P matrix)?

[Walter Roberson]

warp() can handle irregular grids of coordinates and data.

However, if you have rectangular grids of coordinates, equal distance between each x coordinate, equal distance between each y coordinate, then I would recommend using image() or imagesc() or imshow() together with the 'XData' and 'YData' parameters

[Mathan]

So is it something like:

imagesc('XData',M, 'YData', N, 'CData', P)

[Walter Roberson]

imagesc('XData', [M(1,1), M(end,end)], 'YData', [N(1,1),N(end,end)], 'CData', P)

[Mathan]

It still gives out the same result (i.e. different plots whenever reshaping is done). The code is as follows:

M = randi([1 100],13, 21);
N = randi([200 500],13, 21);
P = randi([700 1000],13, 21);
M1 = reshape(M,3,[]);
N1 = reshape(N,3,[]);
P1 = reshape(P,3,[]);
M2 = reshape(M,39,[]);
N2 = reshape(N,39,[]);
P2 = reshape(P,39,[]);
M3 = reshape(M.',1,[]);
N3 = reshape(N,1,[]);
P3 = reshape(P,1,[]);

figure(1);
imagesc('XData', [M(1,1), M(end,end)], 'YData', [N(1,1),N(end,end)], 'CData', P)
figure(2);
imagesc('XData', [M1(1,1), M1(end,end)], 'YData', [N1(1,1),N1(end,end)], 'CData', P1)
figure(3);
imagesc('XData', [M2(1,1), M2(end,end)], 'YData', [N2(1,1),N2(end,end)], 'CData', P2)