CODEWITHSUNDEEP
netcdfnetcdf4nco
skuzmier
skuzmier

Reputation: 23

Change netcdf4 data type

I have a netCDF4 datafile where the time variable is stored as a float (netCDF: 'f8', numpy: float64) and I need to change it to a 32bit int (netCDF: 'i4', numpy: int32). I have tried making the change in python

tds.variables['time'][:] = np.int32(tds.variables['time'][:])

but this hasn't worked. What is the best way to make this change?

Upvotes: 1

Views: 2958

Answers (1)

Bart
Bart

Reputation: 10278

Since you tagged the question with nco, I assume a solution with nco is also acceptable.. This can be done with ncap2 (example with a NetCDF file that I had lying around):

ncdump -h drycblles.default.0000000.nc`:

gives:

netcdf drycblles.default.0000000 {
dimensions: 
    z = 128 ;
    zh = 129 ;
    t = UNLIMITED ; // (37 currently)
variables:  
    double t(t) ;
        t:units = "s" ;
        t:long_name = "Time" ;
.....

Same dump (of modified file) after:

ncap2 -s 't=int(t)' drycblles.default.0000000.nc drycblles.default.0000000_2.nc

gives:

int t(t) ;
    t:long_name = "Time" ;
    t:units = "s" ;

What you are trying in Python won't work since you cast the data of the variable time to int, but still store it as a float (you don't change the variable type in the NetCDF file). I don't see any options to change the data type in place, I guess you could copy the variable time to another name, create a new variable time with type int, copy the data, and remove the old time variable.

For a 2024 solution with xarray:

import xarray as xr
ds = xr.open_dataset('your.nc')
ds['time'] = ds['time'].astype('int')
ds.to_netcdf('your_new.nc')

Upvotes: 7

Related Questions