Regarding Interpolation( vertical)

I have two data sets in which one is 4D data sets and another one is 2D.

1) first data: 120( longitude) x 90(latitude) x 34 (vertical level) x 8 (Time) 2) second data: 15( vertical level) x 469( number of observations)

I would like to interpolate 34 vertical levels to 15 vertical levels. so somebody knows how to fix this problem?

Answer 1

[Sean de Wolski]

A = rand(5,5,34,5);
[ii jj kknew LL] = ndgrid(1:5,1:5,linspace(1,34,15),1:5);
C = interpn(A,ii,jj,kknew,LL);

You could just use interpn directly, thought this will use information from other dimensions.

[Uday]

Thanks for answer but still I have one question ,

A=rand(5,5,34,5), but in my case , A=(120,90,34,8). I always get message that "Matrix dimensions must agree".

and the dimensions A(120 x 90 X 34(1:34) x 8 & (15 (1:15) X 469)

[Sean de Wolski]

I just did that as an example since I didn't want to kill the RAM on my laptop. For your case:

[ii jj kknew LL] = ndgrid(1:120,1:90,linspace(1,34,15),1:8);

C = interpn(A,ii,jj,kknew,LL);

Also, how do you plane on comparing a 2d matrix 15x469, to a 4d matrix? I think bsxfun/repmat might be necessary. Or do you want to append it?

I.e.

cat(3,A,C)? (in which case you don't even need to interpolate down?

Please clarify the goal.

[Uday]

Actually I have two different data-sets 1) 4D data (120(longitude) X 90(latitude) x 34( pressure levels) x 8 (time)), 2) 2D 15( number of vertical levels) x 469 ( number of observations those varies always). My main objective to interpolate vertical levels of two data-sets ( which is 34 & 15). so for this first of all I need to get 15 levels from 4D data. I tried with 4D data-sets the method which you have mentioned earlier, but I got error "Undefined function or method 'colon' for input arguments of type 'double' and attributes 'full 4d real".

[Sean de Wolski]

Show us _exactly_ what you wrote.

Answer 2

[Andrei Bobrov]

In = DataInput; % Array 4D
[a b oldc d] = size(In);
newc = 15;
Out = zeros([a b newc d]);
for j4 = 1:d
    d1 = reshape(permute(In(:,:,:,j4),[3 2 1]),oldc,[]);
    Out(:,:,:,j4) = permute(reshape(cell2mat(arrayfun(@(i1)interp1((1:oldc)',d1(:,i1),linspace(1,oldc,newc)'),1:a*b,'un',0)),newc,b,a),[3 2 1]);
end