netcdf size problem.....

[mashinde]

Dear frds,

I have following two doubts...

(A)I am using d_climatology.m and c_climatology.m from roms matlab tools to create clim file. I have created a grid with 1/16 spatial resolution with 72 vertical levels. The geographical area of grid is -20W to 37E and 29N to 47N.

I am using R2009b version of matlab. I downloaded netcdftools from John Evans for R2009b. I am using netcdf4 with hdf5.

When i compile d_climatology.m script , then it gives me following error and does not create required clim.nc file. This may happen because i am using large grid and clim.nc file exceeds 4GB size.


The error: I only put here last few lines ......

..
..
..
5 -9.305555555556e-01 -6.872736177292e-01
4 -9.444444444444e-01 -7.408057963027e-01
3 -9.583333333333e-01 -7.985069611433e-01
2 -9.722222222222e-01 -8.607018334734e-01
1 -9.861111111111e-01 -9.277404234711e-01
0 -1.000000000000e+00 -1.000000000000e+00

NetCDF file: m16_Lclm_s2.nc
??? Error using ==> nc_inq at 49
mexnc: ncinquire - error while inquiring file: m16_Lclm_s2.nc

Error in ==> nc_write at 84
nc_inq(fname);

Error in ==> d_climatology at 228
[status]=nc_write(CLMname, 'spherical', 'T');

--------------------------------------------------------------------------------------------------------
(B)when i reduce the grid size and vertical levels...then this error is gone and clim.nc file generated which has size less than 4GB.




Anybody knows why this happening. Can matlab handle netcdf files greater than 4GB?





Thanks in advance.

[kate]

Here's how I got around this once:
http://www.arsc.edu/support/news/HPCnew ... l#article2

[m.hadfield]

This is just a hunch, because I'm not very familiar with the ROMS Matlab scripts, but does the problem arise because the climatology file doesn't have a single, unlimited time dimension (as ROMS output files do) but several fixed-size time dimensions, one for each time-dependent variable?

I ask this because ROMS can create output files well over 2 GiB or 4 GiB or whatever in size, even with the 32-bit classic format.

If my hunch is correct, I'm not sure of the easiest way to tackle it. If all the time-dependent fields in the climatology file share the same times (which they often will) then in principle you can give them all the same time dimension and make it unlimited. How easy this is to do with the ROMS Matlab tools I have no idea.