Interpolate gridded matrix row-wise

Hi, I have an 860 by 122 matrix producing a plot below.

I want to interpolate over NaN columns but not all the NaNs in the matrix. I tried using interp function but it gives strange value that are unreasonably wrong. See my code below and attached mat file

figure(1)
pcolor(Y1,Ydepth,sst_up_574); shading flat
set(gca,'Ydir','reverse')
set(gca,'xlim',[bin(1) bin(91)],...
          'ylim',[0 500],'fontsize',8,'xminortick','on','yminortick','on') 
       hold on
      hc = colorbar;
      rmap=jet(20);
      newmap=rmap(2:1:20,:);
      colormap(newmap);
      caxis([4 23]);
      grid on;
      box on
      set(gca,'Fontweight','bold','Fontsize',16);
      set(get(hc,'ylabel'),'string','Water Temperature (^oC)','rotation',270,'verticalalignment','bottom');
      ylabel('Depth (m)');
      xlabel('Date');
      title(['Water Temperature: Upstream SG574'], 'fontweight', 'bold','FontSize',18)

[Stephen23]

@Tumelo Maja: you mention that you tried interpolation, but your code does not mention interpolation anywhere. You attached some data in a .mat file but did not say which variable/s and which values you want to interpolate. Guessing is not an efficient way to write code or provide advice.

[Tumelo Maja]

I need to interpolate sst_up_574. Other variable only positional info.

For the tried interpolation i used : sst_mid_574_interp = naninterp(sst_mid_574);

You can run this in your code to see the difference

[jonas]

"I want to interpolate over NaN columns but not all the NaNs in the matrix."

Very ambiguous. Is this what you are looking for? Perhaps you can sketch what NaNs you want to keep?

[Tumelo Maja]

Yes. This is what i want.

Answer 1

[jonas]

Normally you could achieve this easily by interpolating over rows with a function that does not extrapolate by default. I believe interp1 could do this (?). I used a slighly more messy approach, where I drew a polygon around the area of interest, filled all nans, then removed all points outside of the polygon.

Z = sst_up_574;
X = Y1;
Y = Ydepth;

%Scan rows and find first non-NaN
x = nan(size(X,1))
y = nan(size(Y,1))
for j = 1:size(Z,1)
    [~,c] = find(~isnan(Z(j,:))==true,1,'first');
    if ~isempty(c)
        x(j) = X(j,c);
        y(j) = Y(j,c);
    else
        x(j) = max(X(j,:));
        y(j) = max(Y(j,:));
    end
end
x(isnan(x))=[];
y(isnan(y))=[];

%Build a polygon
x = [x 0];
y = [y min(Y(:))]

%Find points inside of polygon
in = inpolygon(X(:),Y(:),x,y);

%Fill all NaNs
Zq = fillmissing(Z,'linear',2);

%Set values outside of polygon to NaN
Zq(~in) = NaN;

figure(1)
pcolor(X,Y,Zq); shading flat
set(gca,'Ydir','reverse')
set(gca,'xlim',[bin(1) bin(91)],...
    'ylim',[0 500],'fontsize',8,'xminortick','on','yminortick','on')
hold on
hc = colorbar;
rmap=jet(20);
newmap=rmap(2:1:20,:);
colormap(newmap);
caxis([4 23]);
grid on;
box on
set(gca,'Fontweight','bold','Fontsize',16);
set(get(hc,'ylabel'),'string','Water Temperature (^oC)','rotation',270,'verticalalignment','bottom');
ylabel('Depth (m)');
xlabel('Date');
title(['Water Temperature: Upstream SG574'], 'fontweight', 'bold','FontSize',18)