Patch plot z-values colorbar not correct
Show older comments
Hello,
I have a 'patch' plot with x, y, and z-values. A 2x2 sublot is desired with a single colorbar on the right. The values of the colorbar should be from 0 to 1. However, the values in the plot are an order of magnitude off. My largest value in the upper left plot, for example, is somewhere around 0.08. But, the figure is plotting the value as 0.8.
I've attached my plotting script (below) and the variables needed (attached).
Any ideas?
Thanks.
fig = figure;
subplot(2, 2, 1)
patch(X, Y, norm_DSVD_sum2019_JER,'HandleVisibility','off');
title('Summer 2019, Las Cruces, NM', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0])
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4])
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2);
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--');
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':');
subplot(2, 2, 2)
patch(X, Y, norm_DSVD_win2019_JER,'HandleVisibility','off');
title('Winter 2019, Las Cruces, NM', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0])
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4])
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2);
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--');
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':');
subplot(2, 2, 3)
patch(X, Y, norm_DSVD_sum2019_SFL,'HandleVisibility','off');
title('Summer 2019, Mesa, AZ', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0])
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4])
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2);
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--');
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':');
subplot(2, 2, 4)
patch(X, Y, norm_DSVD_win2019_SFL,'HandleVisibility','off');
title('Winter 2019, Mesa, AZ', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0])
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4])
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2);
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--');
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':');
lgd = legend([p1(1) p2(1) p3(1)],'Atlas et. al., 1973', 'Atlas and Ulbrich, 1977', 'van Dijk et. al., 2002', 'Position', [.72 .12 .14 .1]);
lgd.FontSize = 6.5;
lgd.Title.String = 'Droplet Size/Velocity Relationships';
sgt = sgtitle('Drop Size and Velocity Distributions');
sgt.FontSize = 20;
ax = axes(fig);
yyaxis(ax, 'left');
han = gca;
han.Visible = 'off';
set(gca, 'ycolor', 'k')
han.XLabel.Visible = 'on';
han.YLabel.Visible = 'on';
xlabel('Raindrop Size [mm]')
ylabel('Raindrop Speed [m/s]')
yyaxis(ax, 'right');
c = colorbar('Position', [0.9214, 0.0935, 0.0165, 0.8147]);
c.Label.String = 'Raindrop Count';
c.Label.FontSize = 12;
colormap('jet')
han.YLabel.Visible = 'on';
Accepted Answer
There are several issues to cover.
1) If you want to use 1 colorbar to represent all subplots, all subplots must have the same scale. That can be achieved using caxis(limits). However, the comment below is not correct.
"The values of the colorbar should be from 0 to 1. However, the values in the plot are an order of magnitude off."
The values that define the color in your subplots are all between ~0 and ~0.534. If you were expecting something different, the data you shared is incorrect.
After each patch() command, you need to set the caxis values to the span of the entire color range using
% These two lines only needs computed once, before plotting is done.
allColorData = [norm_DSVD_sum2019_JER, norm_DSVD_sum2019_SFL,norm_DSVD_win2019_JER,norm_DSVD_win2019_SFL];
colorRange = [min(allColorData), max(allColorData)];
% Repleat this line for each subplot
caxis(colorRange)
You'll notice that the colors now differ. Some plots will not show colors outside of a range of blues. If this is not what you want. you cannot have 1 colorbar for all subplots.
2) The extra invisible axes you're adding to control the colorbar is inefficient and prone to error. It's better to add the colorbar to any of the subplots and specify its position. But before studying this step, see the section below "Use tiledlayout instead of subplot"
To add the colorbar, put this after your last subplot
cb = colorbar('Position', [0.9214, 0.0935, 0.0165, 0.8147]);
You may want to move the subplots a bit to make room for the colorbar.
% sp is a 1x4 vector of subplot handles created by
% sp(1) = subplot(___);
% sp(2) = subplot(___);
% etc....
% Move subplots to the right by 0.05 normalized units
arrayfun(@(i)set(sp(i), 'Position', [sp(i).Position(1)-0.05, sp(i).Position(2:end)]), 1:numel(sp));
The bottom and right super-labels can be added easily using text objects.
Use tiledlayout instead of subplot
tiledlayout was introduced in r2019b and allows you to set x and y labels for all subplots. To convert your code to tiled layout,
1) After creating the figure define the layout
tlay = tiledlayout(2,2);
2) change all of your subplot commands to
sp(1) = nexttile(); % replace 1 with whatever subplot you're creating
3) There are some restrictions with the colorbar. It can only be next to 1 subplot. So, remove the 'position' property.
c = colorbar();
If the position restriction is a big problem, you'll have to add an invisible axis just for the colorbar.
ax = axes('position', [.9,.1,.08,0.75]);
caxis(colorRange)
c = colorbar('Location','East');
c.Position(3) = .015
axis(ax, 'off')
4) Then define the x and y labels
xlabel(tlay, 'Raindrop Size [mm]')
ylabel(tlay, 'Raindrop Speed [m/s]')
and replace the sgtitle command with
title(tlay, 'Normalized Drop Size and Velocity Distributions', 'FontSize', 20)
Notice that colors beyond blue are really only seen in Mesa AZ in the summer. It makes sense that rainfall in the summer will have a higher drop count than in the winter. However, it might be surprising that rainfall isn't similar between Las Cruces and Mesa in the summer on bestplaces.net (I have no idea how valid those data are).
28 Comments
Eric Escoto
on 10 Aug 2020
Edited: Eric Escoto
on 10 Aug 2020
Adam Danz
on 10 Aug 2020
Oops. I replaced the first line of code with the correct line.
Adam Danz
on 10 Aug 2020
If you want the colorbar to be normalized to [0,1] you just need to normalize allColorData to be between 0 and 1.
Eric Escoto
on 11 Aug 2020
Edited: Adam Danz
on 11 Aug 2020
"It seems like I am creating the same exact plots I was before using the 'allColorData' line you've suggested."
The color ranges are much different. All subplots in your atttached figure range from dark blue to yellow or orange. The subplots in my answer (and your new code) are all on the same scale and only one of them gets into the yellow range.
You forgot to use caxis on the invisible axes so your colorbar isn't scalled according to colorRange. See step 3 in the "Use tiledlayout instead of subplot" section of my answer.
Eric Escoto
on 11 Aug 2020
Edited: Adam Danz
on 11 Aug 2020
Adam Danz
on 11 Aug 2020
I've edited your comments to format your code. In furture comments, please use the [>] button in the editor to format code in your comments (more info).
"The single colorbar should represent all values in each subplot no matter what the difference in values are for each plot."
That's what my answer does but the range of values in your vectors do not span from 0 to 1. In my answer, the colors in all of the subplots are defined by the colors in the colorbar.
"I may want to change the colorbar range to better represent the actual data range "
Again, that's exactly what my answer is doing. In the original data you shared, the norm_DSVD.... values all spaned from nearly 0 to about 0.534 and if you look at the colorbar in the plot in my answer, you'll see that it also spans from 0 to 0.534.
So, it's unclear whether you're saying that the plot in my answer, given the data, is not what you're aiming for --or- that you're having trouble adapting your code to my solution.
If my solution is not what you're looking for, then the meaning of the norm_DSCD_... values is not clear.
- Why does it matter that they sum to 1?
- What should the color represent?
- Should each subplot show colors that cover the full range of the colorbar?
Perhaps you want to normalize each vector to span from 0 to 1.
z = (norm_DSVD_sum2019_JER - min(norm_DSVD_sum2019_JER)) / range(norm_DSVD_sum2019_JER);
or maybe you want the cumulative sum
z = cumsum(norm_DSVD_sum2019_JER, 'omitnan');
If my solution is what you're looking for, then here are some changed that are needed to the code in your previous comment.
- nexttile does not have any input arguments. It literally just adds a new subplot according to your tiledlayout. See examples. So, you'll need to remove the patch() commands from the nexttile inputs and put them on a separate line after nexttile(). That's what my step #2 explains in my answer under "Use tiledlayout instead of subplot".
- Then you need to follow step #3 in that section - the 5 lines at the end. Note that I'm placing the invisible axes much differently than you are. Critically, I'm also applying caxis to it.
Eric Escoto
on 11 Aug 2020
" I'd like a single colorbar to go from [0, max(of any one plot whichever that may be)]"
Impossible.
You can't have 1 colorbar for 4 different scales. "Canary Yellow" can only mean 1 thing. "Royal Blue" can only mean 1 thing. That's why the colorbar in my answer spans the entire range of all norm_DSVD... values.
" The value of 0.05 in one plot should be the same color as the value of 0.05 in another plot. some plots would not show a color value greater than 0.05 because they simply dont have that value in the vector."
That's exactly what my answer does. That also differs from the first quoted sentence above. For example, let's say the max of subplot 1 is 0.005, the max of subplot 2 is 0.01, subplot 3 max is 0.05 and subplot 4 max is 0.1; Then the colorbar is my answer will span from 0.005 to 0.1 since those are the global min and max values.
"The color represents that portion of the total drops in each bin accrding to the normalization. "
If you'd like, instead of using caxis(colorRange) on each subplot, you could use caxis([0,1]). However, since your max colorbar value is very low, the result will be that all colors go down to blue since the range is increase to 1. Note that the current version, using caxis(colorRange), is just as fine. The only difference is that the colorbar will cut off at the max value rather than 1. The interpretation is still exactly the same.
"I suppose the colorbar should simply span the range of 0 to the maximum value across all plots"
That's exactly what my answer is doing. Your smallest norm_DSVD... value isn't quite equal to zero (it's very very close). If you wanted exactly 0, replace
colorRange = [min(allColorData), max(allColorData)];
with
colorRange = [0, max(allColorData)];
Again, it's unclear whether the problem is with my answer (which, it could be, if the interpretation of the values differs from my understanding), or if the problem is in your implementation of my answer.
Eric Escoto
on 5 Sep 2020
Edited: Eric Escoto
on 5 Sep 2020
Adam Danz
on 6 Sep 2020
Make sure you're using Matlab r2019b or later. Earlier versions do not support tiled layout.
Could you share the entire error message as well as the line that is generating the error?
Eric Escoto
on 7 Sep 2020
See the end of my comment above (<-- click link). Specifically, the part that starts with "1. nexttile does not have any input arguments".
Eric Escoto
on 8 Sep 2020
I've made changes to the nexttile as per part three. However, the color scale is still not working correctly. The figure that is created is below.
allColorData = [norm_DSVD_sum2019_JER, norm_DSVD_win2019_JER,norm_DSVD_sum2019_SFL,norm_DSVD_win2019_SFL];
colorRange = [min(allColorData), max(allColorData)];
fig = figure; % figure 6
tlay = tiledlayout(2, 2);
% subplot(2, 2, 1) % JER Summer 2019
sp(1) = nexttile; %(patch(X, Y, norm_DSVD_sum2019_JER,'HandleVisibility','off'));
patch(X, Y, norm_DSVD_sum2019_JER,'HandleVisibility','off');
caxis(colorRange)
title('Summer 2019, Las Cruces, NM', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0]) % These ticks delineate changes in class spread.
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4]) % These ticks delineate changes in class spread.
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2); % This plots the Atlas empirical v(D) line from Gunn-Kinzer's data.
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--'); % This plots the Atlas and Ulbrich, 1977 v(D) line.
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':'); % This plots the van Dijk et. al. (2002) v(D) line.
% subplot(2, 2, 2) % JER Winter 2019
sp(2) = nexttile, %(patch(X, Y, norm_DSVD_win2019_JER,'HandleVisibility','off'));
patch(X, Y, norm_DSVD_win2019_JER,'HandleVisibility','off');
caxis(colorRange)
title('Winter 2019, Las Cruces, NM', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0]) % These ticks delineate changes in class spread.
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4]) % These ticks delineate changes in class spread.
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2); % This plots the Atlas empirical v(D) line from Gunn-Kinzer's data.
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--'); % This plots the Atlas and Ulbrich, 1977 v(D) line.
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':'); % This plots the van Dijk et. al. (2002) v(D) line.
% subplot(2, 2, 3) % SFL Summer 2019
sp(3) = nexttile, %(patch(X, Y, norm_DSVD_sum2019_SFL,'HandleVisibility','off'));
patch(X, Y, norm_DSVD_sum2019_SFL,'HandleVisibility','off');
caxis(colorRange)
title('Summer 2019, Mesa, AZ', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0]) % These ticks delineate changes in class spread.
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4]) % These ticks delineate changes in class spread.
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2); % This plots the Atlas empirical v(D) line from Gunn-Kinzer's data.
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--'); % This plots the Atlas and Ulbrich, 1977 v(D) line.
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':'); % This plots the van Dijk et. al. (2002) v(D) line.
% subplot(2, 2, 4) % SFL Winter 2019
sp(4) = nexttile, %(patch(X, Y, norm_DSVD_win2019_SFL,'HandleVisibility','off'));
patch(X, Y, norm_DSVD_win2019_SFL,'HandleVisibility','off');
caxis(colorRange)
title('Winter 2019, Mesa, AZ', 'FontSize', 12)
xlim([0 8])
ylim([0 11])
xticks([0 1.25 2.5 5.0 8.0 10.0 20.0 26.0]) % These ticks delineate changes in class spread.
yticks([0 1.0 2.0 4.0 8.0 11.0 16.0 22.4]) % These ticks delineate changes in class spread.
grid off
hold on
p1 = plot(X, vD_AfGK, 'r', 'LineWidth', 2); % This plots the Atlas empirical v(D) line from Gunn-Kinzer's data.
p2 = plot(X, vD_AU, 'r--', 'Linewidth', 2, 'LineStyle', '--'); % This plots the Atlas and Ulbrich, 1977 v(D) line.
p3 = plot(X, vD_VD, 'r', 'Linewidth', 2, 'LineStyle', ':'); % This plots the van Dijk et. al. (2002) v(D) line.
lgd = legend([p1(1) p2(1) p3(1)],'Atlas et. al., 1973', 'Atlas and Ulbrich, 1977', 'van Dijk et. al., 2002', 'Position', [.72 .12 .14 .1]);
lgd.FontSize = 6.5;
lgd.Title.String = 'Droplet Size/Velocity Relationships';
% sgt = sgtitle('Normalized Drop Size and Velocity Distributions');
% sgt.FontSize = 20;
title(tlay, 'Normalized Drop Size and Velocity Distributions', 'FontSize', 20)
ax = axes(fig);
yyaxis(ax, 'left');
han = gca;
han.Visible = 'off';
set(gca, 'ycolor', 'k')
han.XLabel.Visible = 'on';
han.YLabel.Visible = 'on';
xlabel(tlay, 'Raindrop Size [mm]')
ylabel(tlay, 'Raindrop Speed [m/s]')
yyaxis(ax, 'right');
c = colorbar();
% c = colorbar('Position', [0.9214, 0.0935, 0.0165, 0.8147]);
c.Label.String = 'Normalized Raindrop Count';
c.Label.FontSize = 12;
colormap('jet')
han.YLabel.Visible = 'on';
Adam Danz
on 9 Sep 2020
That's because of this unnecessary axis which I mentioned in step #2 in the upper half of my answer.
ax = axes(fig);
yyaxis(ax, 'left');
han = gca;
han.Visible = 'off';
set(gca, 'ycolor', 'k')
han.XLabel.Visible = 'on';
han.YLabel.Visible = 'on';
xlabel(tlay, 'Raindrop Size [mm]')
ylabel(tlay, 'Raindrop Speed [m/s]')
yyaxis(ax, 'right');
Remove that and follow step #4 in the bottom half of my answer instead.
Eric Escoto
on 9 Sep 2020
Got it! Thanks!
With all that done, I've got a small issue to correct with regards to the plot itself.
There seems to be too much space between the left and right columns of plots, and I'd like to reduce it a little. What callout can help with this? I think reducing that white space by half could work better. Basically a portion of the colorbar label is cut off.
Glad I could help.
Regarding the axis spacing, see the TileSpacing and Padding properties of tiledlayout:
Eric Escoto
on 9 Sep 2020
Edited: Eric Escoto
on 9 Sep 2020
That worked to reduce the space, however, the text on the colorbar is still cropped out.
I used Tilespacing = none, and Padding = normal. Is the colorbar location just too far out to the right?
Here's the figure.
Edit: I just found this question in the community. Didn't seem to be resolved though.
Adam Danz
on 9 Sep 2020
I see. In my version (in my answer) the colorbar ticks only have 1 decimal place so the cb legend isn't pushed out of view. Your have 2 dp and that's probably causing the problem. Have you tried placing the cb ylabel on the left of the cb instead of the right? I wonder if that would fix it.
With subplots, we can adjust the position of the cb but with tiledlayout we cannot, at least in r2020a and earlier. The advice in the answer you linked to is for regular subplots, not tiledlayout plots which were introduced relatively recently.
However, you probably can reposition the cb ylable using
cbYlab = ylabel(cb, 'Normalized....Count'); % where cb is your colorbar handle
cbYlab.Position(1) = ____
Eric Escoto
on 10 Sep 2020
Adam Danz
on 10 Sep 2020
This
cbYlab = ylabel(c, 'Normalized Raindrop Count');
is the same as what you're doing above
c.Label.String = 'Normalized Raindrop Count';
except that my version stores the handle to the y-axis-label.
What problems are you having with that line? After running that line, look at
cbYlab.Position
It will be a 1x3 vector where the first and 2nd values show the horizontal and vertical position of the y-axis label. You can shift the position leftward by subtracting a little from cbYlab.Position(1) using something like
cbYlab.Position(1) = cbYlab.Position(1)*.95; % ie, 95% of its current position
Eric Escoto
on 10 Sep 2020
I'm seeing the label itself move, but the cb itself is too close to the plots. Part of the text is cut out. Here's the figure I'm making. Sorry for the fuzzyniess.
For future refernce, whats the best way to inset a figure here? i usually save the .fig as a .jpg but it comes out either to large or too small.
Well, it looks like you're back to using the invisible overlaying axis.
I still have the same response to that as above
If you put together a minimal working example (MWE) of 4 tiled plots using tiled layout and a colorbar with 2 decimal places along the y axis, I can play around with it. The MWE does not need to contain any data. What's important is that your tiled layout spacing is preserved in the MWE.
Eric Escoto
on 10 Sep 2020
Edited: Eric Escoto
on 10 Sep 2020
Unfortunately I cannot run your script since I don't have the variable definitions.
To make a MWE you just need to eliminate anything in your code that calls your variables and maintain the parts that produce the tiled layout and colorbar with the exact spacing you're using. The colors and other details aren't important. It should be able to run if I copy and paste it into my workspace.
Example (run this):
figure()
tl = tiledlayout(2,2);
nexttile
nexttile
nexttile
nexttile
cb = colorbar();
ylim(cb, [0, .5])
cb.Ticks = [.111:.1:.5]; % 3 decimal places
yl = ylabel(cb, 'Label');
The idea is to vastly reduce the amount of time needed for me (or any other volunteer) to reproduce the problem. Otherwise, there's way too much time needed to sift through a bunch of code that has nothing to do with the spacing problems and to guess what variables mean, etc. Sometimes MWEs take time to put together but I can assure you that the amount of time saved on our ends is much greater than the time it takes to create a MWE. With a functional example that reproduces the problem on my end, this could have been solved hours ago.
Have you thought about moving the legend to the upper axis and then putting the colorbar inside the bottom axis? There's lots of open space on the right side of your plots.
Also, the legend should be in the south east corner to avoid overlapping the small amount of data in the north east corner.
Adam Danz
on 20 Sep 2020
Good news.
As of r2020b you can add a shared colorbar to tiledlayout
Eric Escoto
on 20 Sep 2020
More Answers (0)
Categories
Find more on Axes Appearance in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
