Making a square figure with pcolor

Hello,

I have 2 matrices for x and y coordinates, and a matrix of temperatures. From the matrices, I make 2 coordinate vectors of size 99000 x 1 from the 300 x 300 matrices.

When I plot with imagesc and the vectors, I have the linked result. When I plot with pcolor and the coordinates matrices, I have the other figure.

I would like to make the second figure (plotted with pcolor) square like the one plotted with imagesc. How could I achieve that ?

Thanks in advance !

[darova]

Can you show somethign?

[V.D-C]

Ah sorry I didn't get the notification, and I thought I attached the figures. Here you go

[darova]

Can you attach your code? Data?

[V.D-C]

Sorry again, I couldn't answer to your message earlier.

The code looks something like that :

 
%%% Pcolor
h = figure();
    
    pcolor(x(50:250,50:250),y(50:250,50:250),Tair_mean(50:250,50:250));%% I frame the area of interest
    shading flat
    
    colorbar;
    ylabel(colorbar,'Air Temprature [°C]');
    title(['Mean Daily Air Temperature ' date]);
    axis equal
    set(gca, 'Color', 'none'); 
    
    
%%% Imagesc

% I transform my coordinates matrices in vectors in order to make them work with imagesc
x_vect = x(:);
y_vect = y(:);


h = figure();
imagesc(x_vect,y_vect,Tair_mean(50:250,50:250))  %% I frame the area of interest

How could I get a square figure like imagesc but with pcolor ?

Answer 1

[darova]

You can either use pcolor without X and Y vectors

newTair = flipud(Tair_mean(50:250,50:250));     % flip matrix upside-down
pcolor(newTair)

Or you can create rotate X and Y

a = 51;
R = [cosd(a) sind(a);-sind(a) cosd(a)];
V = R*[x(:) y(:)]';
x1 = reshape(V(1,:),size(x));
y1 = reshape(V(2,:),size(x));
pcolor(x1(50:250,50:250),y1(50:250,50:250),Tair_mean(50:250,50:250));%% I frame the area of interest

[V.D-C]

Hello, thank you very much for your answer.

But it is not giving me what I would like.

The output I want is a figure where the archipelago is straight (as in my pcolor figure), but with square window (as the my imagesc figure, or the one given by your code).

[darova]

axis tight

[J. Alex Lee]

darova's code seems to work for me...

rather than axis tight, I think you want

set(gca,'DataAspectRatio',[1 1 1])

darova has already identified that you only need a rotation through 51deg to "rectify" your coordinates, so you woulnd't need to fiddle around with fitgeotrans().

[V.D-C]

I will test it as soon as my matlab works !

One question though, how did you do to find the rotation value ?

[darova]

• One question though, how did you do to find the rotation value ?

Good question. I used my eyes to calculate it. But maybe it's not the best way

dx = x(2) - x(1);
dy = y(2) - y(1);
a = 90 + atan2d(dy,dx)
a =
   50.9309

Answer 2

[J. Alex Lee]

Your coordinate grid matrices x and y are not "aligned" with the axes. You can see this if you just do

plot3(x,y,zeros(size(x)),'.k')
view(2)

So I would say your pcolor code is giving you the correct rendering, and imagesc behavior seems anomalous, since your values of x and y (called x_vect and y_vect in your code), do not follow the requirements of imagesc:

per the documentation, x and y should either be vectors of respective lengths size(Tair_mean), or 2-element vectors specifying only the corner coordinates (implicitly assume on a rectangular canvas aligned with axes), or scalars specifying only 1 corner. As you supply the entire grid in a long vector, it must be assuming you only care about the first and last points, or something like that.

If you want to align with axes, I guess you have 2 options:

• Forget about the actual coordinate values and simply issue

pcolor(Tair_mean(50:250,50:250));

• Transform your original grid (x,y) to a new grid aligned with the axes, using some rotation (and maybe a skew). If you have the image processing toolbox, you can play with https://www.mathworks.com/help/images/ref/fitgeotrans.html

[V.D-C]

Hello, thank you very much for your answer.

I think I will have to forget about the values because to transform my original grid I need a rotation matrix but also to take account of the deformation of the pixels because of the UTM conversion.

I will dig into the image processing toolbox though, as you suggested.