"griddata" results in much smaller values than original data

I am using "griddata" function to interpolate original data on query points. However, the interpolation value on the query points are much smaller than the original data. Is there any way to improve the interpolation?

[KSSV]

I suspect your lon, lat values are not exactly corresponding to elev. While running the code did you see the warning?

Warning: Duplicate data points have been detected and removed - corresponding values have been averaged. 

I suspect the high values correspond to diplicate lon, lat and have been removed.

When you check

nx = length(unique(lon)) ;
ny = length(unique(lat)) ;

nx*ny is not equal to 579072.

[jie hu]

Yes, when I run 'IC_plot.m' matlab reports the duplicate points and removed them. averaged value is taken. My idea is to detect the duplicate points and take the biggest value on each point. Is that realizable?

Answer 1

[Chunru]

It is due to the duplicate points in data. Griddata is averaging over the duplicate points to have a smaller value than peak. If you remove the duplicate points, then the result will be similar.

websave("manila.zip","https://www.mathworks.com/matlabcentral/answers/uploaded_files/1359608/manila.zip")

ans = '/users/mss.system.Ktq2V0/manila.zip'

unzip("manila.zip")
% plot x y z vector

data=importdata('manila.dat');
lon=data(:,2); 
lat=data(:,3);
elev=data(:,4);

% Remove duplicate points
% [~, I, ~] = unique([lon lat], 'first', 'rows');
% lon = lon(I);
% lat = lat(I);
% elev = elev(I);

% Use peak for duplicated points
c = unique([lon lat], 'first', 'rows');
n = size(c,1)

n = 48256

lon1 = zeros(n, 1);
lat1 = zeros(n, 1);
elev1 = zeros(n, 1);
for i=1:size(c, 1)
    idx = lon==c(i, 1) & lat==c(i,2);
    lon1(i) = c(i, 1); lat1(i) = c(i, 2);
    elev1(i) = max(elev(idx));
end

max(elev(:))

ans = 20.8693

plot3(lon,lat,elev,'.');

% xlin = linspace(min(lon)+0.2,max(lon)-0.2, 51);
% ylin = linspace(min(lat)+0.2,max(lat)-0.2, 201);
xr = (min(lon1):0.01:max(lon1));
yr = (min(lat1):0.01:max(lat1));
[X, Y]= meshgrid(xr,yr);
Z = griddata(lon1,lat1,elev1,X,Y,'natural');
max(Z(:))

ans = 20.8693

mesh(X,Y,Z)

% jump_grid = 1;
% 
% figure
% surf(X(1:jump_grid:end,1:jump_grid:end), ... 
%      Y(1:jump_grid:end,1:jump_grid:end), ... 
%      Z(1:jump_grid:end,1:jump_grid:end));
% shading interp; 

% whos

[jie hu]

thanks very much for your help. It works!

Answer 2

[KSSV]

data=importdata('manila.dat');
lon=data(:,2); 
lat=data(:,3);
elev=data(:,4);

xr = 118.7:0.2:121;
yr = 12.4:0.2:20.7;
[X, Y]= meshgrid(xr,yr);
idx = knnsearch([lon lat],[X(:) Y(:)],'k',10) ;
Z = reshape(max(elev(idx),[],2),size(X)) ; 

plot3(lon(1:100:end),lat(1:100:end),elev(1:100:end),'o');
hold on
surf(X,Y,Z)

[jie hu]

thanks KSSV, very helpful your alternative approach!.