interpolate 3 dimensional arrays (in time and space)

Hi all,

I have 3D matrix 180x90x120 which is equal to lonxlatxtime. In space, the data has 1 degree resolution (e.g. longitude ranges from 1 to 180), while the time interval is 3 hours (e.g. 3h, 6h, 9h.....).

Now I want to interpolate data to get a finer resolution of 0.5 degree and with 1-hour interval, so the new matrix would be 360x180x360.

Could you please help?

Thanks

[KL]

You might want to read griddedinterpolant

[wave_buoys]

Hi, I see I can use griddata to interpolate in spatial space, so now my concern is only how to interpolate data in time. Lets simplify, assuming I have 2 dimensional matrices at 1:00AM A1=[1 1; 1 1], at 2:00AM A2=[2 2; 2 2], at 3:00AM A3= [3 3; 3 3] what I need to do, for instance, to get additional data at 1:30AM it should be A1.5 = [1.5 1.5; 1.5 1.5] and at 2:30AM A2.5 = [2.5 2.5; 2.5 2.5]. To this end, I will have data at 1:00 AM, 1:30AM; 2:30AM; 3:00AM...etc.

I believe it could be simple for many of you, but I am still stuck in it. Could you please help.

Thanks

[Matt J]

No, do not use griddata for this. KL's suggestion of griddedInterpolant was appropriate. See also my second answer.

[Matt J]

Now I want to interpolate data to get a finer resolution of 0.5 degree and with 1-hour interval, so the new matrix would be 360x180x360.

Those will not be the new dimensions. If you sample from 1 to 180 at intervals of 0.5, you will obtain 359 points, not 360. Similarly for the other dimensions.

Answer 1

[Matt J]

 F=griddedInterpolant(yourData);
  qlon=(1:.5:180);
  qlat=(1:.5:90);
  qtime=(1:1/3:120);

 newData=F({qlon,qlat,qtime});

[wave_buoys]

THIS WORKS FOR ME!!!!!!!! THANKS MATT

[wave_buoys]

Dear Matt J,

Sorry. It works but the outcomes are not as I expected. Particularly the new results look weird. Perhaps I also misunderstood about the input matrix format. To be corrected, the 3D matrices I mentioned, for instance, 180x90x120 contains 120 2D arrays size (180x90) for monthly wind speeds in Longitude and Latitude direction. These 120 matrices correspond to a 1D time matrix (which is separated): 0h 6h 12h 18h.....720h..

So when I use:

v = winddata % (size: 180x90x120)
F= griddedInterpolant(v)
qlon=(1:.5:180)
qlat=(1:.5:90);
qtime=(1:1/3:120);
newwinddata=F({qlon,qlat,qtime});

it does not work properly because time does not present in griddedInterpolant(v).

May you help more? Thanks

Answer 2

[Matt J]

You could download IMRESIZEN, and then do things like,

>> A=rand(180,90,120); B=imresizen(A,[2,2,3]); whos B

Name   Size            Kilobytes     Class    Attributes

B      360x180x360        182250     double

[wave_buoys]

Thanks Matt, but it does not for my situation because Imresizen just scales up the data.

Has anyone done something similar to this? May you help?

Thanks

[Matt J]

How is that different from what you are trying to do?