The dirty way which is not quick but everyone seems to do it, is to perform 1D interpolation for each lat/lon grid box. That would look like this:
% Given this data:
t = 1:135;
A = rand(180,360,135);
% You want Ai at times ti:
ti = linspace(1,135,147);
% Preallocate Ai to speed up the loop:
Ai = NaN(180,360,147);
% Perform 1D interpolation at each lat/lon grid box:
for lati = 1:180
for loni = 1:360;
Ai(lati,loni,:) = interp1(t,squeeze(A(lati,loni,:)),ti,'linear');
end
end
However, nested loops are dreadful. The loops above call interp1 more than sixty four thousand times. That makes computation slow, and debugging nested loops tends to be a headache. A better way to do this is 3D interpolation. Grid your source data, grid the data points to which you want to interpolate, then call interpn just once:
[LON,LAT,T] = meshgrid(-179.5:179.5,-89.5:89.5,t);
[LONI,LATI,TI] = meshgrid(-179.5:179.5,-89.5:89.5,ti);
Ai2 = interpn(LAT,LON,T,A,LATI,LONI,TI,'linear');
