Plotting 3D surface contours of a variable with respect to x, y and z

I am trying to plot a 3D surface contour plot a varaible (dust mixing ratio) with respect to lat (x-axis), long (y-axis) and levels (z axis) (attached csv file). But I am unable to plot that. While making the meshgrid I am getting an error Requested 248040x248040x2 (916.8GB) array exceeds maximum array size preference. How to plot this?

[Rik]

What code did you try? You can put it right in your question and run the code to show us exactly how to reproduce the error.

And what do you want exactly? A surface plot or a contour plot? Both rely on a single value for the height given an x-y coordinate pair. Your error suggests you have a list of coordinates that should be reshaped instead of using meshgrid.

[shravani banerjee]

I want a 3D contour plot with x as lat, y as lon, and z as levels and the legend colour will represent my variable. I tried to created a mesh to make a 3D contour plot.

T=readtable('MIXING_RATIOe.csv');
lat = T.lat;
lev = T.levels;
lon = T.lon;
values = T.dust_mixing_ratio;

y = lat;
m = lon;
z = lev;
[Y,M,Z] = meshgrid(y,m,z);

Answer 1

[Rik]

It sounds like you want a 3D pointcloud where the color represents a 4th variable.

As you can see below, this might not actually be a very informative plot.

t = tic;%keep track of time
unzip('MIXING_RATIOe.zip')
T=readtable('MIXING_RATIOe.csv');
lat = T.lat;
lev = T.levels;
lon = T.lon;
values = T.dust_mixing_ratio;

cmap = colormap('parula');
N_colors = size(cmap,1);
unique_values = unique(values);
C_ax = [min(values) , max(values)];
N = numel(unique_values);
%for n=1:N
for n=randperm(N) % this is to make sure stopping halfway doesn't distort the plotted colors
    L = values == unique_values(n);
    col_idx = interp1(...
        linspace(C_ax(1),C_ax(2),N_colors),... % data range
        1:N_colors,... % colormap indices available
        unique_values(n)); % query current value
    plot3(lat(L),lon(L),lev(L),...
        '.','color',ind2rgb_private(col_idx,cmap))
    if n==1,hold('on'),elseif n==N,hold('off'),end
    
    % for the online environment, exit after 20 seconds of work
    if toc(t)>20,break,end,hold('on')
end
colorbar,caxis(C_ax) % add colorbar and adjust limits to match the plotted results

function col = ind2rgb_private(col_idx,cmap)
% extend ind2rgb to allow decimal indices
N_colors = size(cmap,1);
col = [...
        interp1(1:N_colors,cmap(:,1),col_idx),...
        interp1(1:N_colors,cmap(:,2),col_idx),...
        interp1(1:N_colors,cmap(:,3),col_idx)];
end

[shravani banerjee]

Can't I plot a surface instead of point plot using this data and reduce the opacity of the plot such that maximum information is visible?

[Rik]

You could (and you can use a similar idea to this code, but I doubt that will give you more visible information than the current plot. You need to consider what the maximum amount of usefull information is.

It might be a better idea to interpolate your data to a solid cube. Then you can use 3D imaging tools to scroll through your data in different planes.

[shravani banerjee]

Here is an example of the plot I wanted to make.

[Rik]

I suspect you could get something like this if you set the Alpha property of a patch object below a certain threshold (and scale that with the value as well).

I don't know how fine your slices need to be before you don't notice them anymore. I suggest you give it a try. I'm not aware of any way to create this using solid voxels in Matlab.

Answer 2

[shravani banerjee]

I got the required plot. It can be plotted through pcolor3. https://in.mathworks.com/matlabcentral/fileexchange/49985-pcolor3

[Rik]

Can you provide a link? I can't find it on Google, so I doubt it will be easy to find for future readers of this thread.