CDO

Hello,

I am using cdo in python. I would like to collocate ASCAT wind products with ERA5 wind components in space and time.
The code I am using first interpolates ERA5 on to ASCAT grid (cdo remapnn) and then selects times (that are in a list) closer to the full hour of ERA5 in a for loop for each point defined by (lon, lat). This for loop is very slow. Is there a way to interpolate in time avoiding the for loop?
Thank you in advance. Your assistance would be greatly appreciated.

I attach my code below:

1. load era5 lon, lat
   fid = nc.Dataset("./u10_ANT_2016.nc")
   latE = np.array(fid.variables['latitude'][:], dtype=float)
   lonE = np.array(fid.variables['longitude'][:], dtype=float)

fid.close()

x = np.append(np.vstack([np.array(date[i].year) for i in range(365)]),
np.vstack([np.array(date[i].month) for i in range(365)]), axis=1)

x = np.append(x, np.vstack([np.array(date[i].day) for i in range(365)]),
axis=1)

1. merge date time arrays for asc, desc
   d_asc = np.empty(tod_asc.shape, dtype='U32')
   d_desc = np.empty(tod_desc.shape, dtype='U32')
   for i in range(365):
   for iy in range(len(lat)):
   for ix in range(len(lon)):
   if (~np.isnan(tod_asc[i, iy, ix])):
   tm = int(np.round(tod_asc[i, iy, ix]))
   if (tm == 24):
   tm = 0
   d_asc[i, iy, ix] = (str(x[i, 0])+'-'+str(x[i, 1]).zfill(2)
   '-'+str(x[i, 2]).zfill(2)'T'
   str(tm).zfill(2)":00:00")
   if (~np.isnan(tod_desc[i, iy, ix])):
   d_desc[i, iy, ix] = (str(x[i, 0])+'-'+str(x[i, 1]).zfill(2)
   '-'+str(x[i, 2]).zfill(2)'T' +
   str(int(np.round(tod_desc[i, iy,
   ix]))).zfill(2) +
   ":00:00")
   np.save('d_asc', d_asc)
   np.save('d_desc', d_desc)
2. allocate arrays for collocation
   ue_asc = np.empty(tod_asc.shape, dtype=float)*np.nan

1. first convert lon -180:180 to 0:360 and invert latitudes
   cdo.sellonlatbox('0.0,360.0,-90.0,-40.0', input='-invertlat u10_ANT_2016.nc',
   output='u10_ANT_2016_lon360.nc', options='-f nc')
2. generate ascat grid (terminal)
3. cdo griddes ./ascat_surfacewind/DATA/cropped/ascat_2016.nc > ascat_grid.txt
4. remapnn to ascat grid
   cdo.remapnn('ascat_grid.txt',
   input='-sellonlatbox,0.0,360.0,-90.0,-40.0, '
   '-invertlat u10_ANT_2016.nc',
   output='u10_ANT_2016_gASCAT.nc', options='-f nc')
5. collocation loop (veeery slow)
   for (iy, _) in enumerate(lat): #! FIXME
   for (ix, _) in enumerate(lon):
   if (not np.all(np.isnan(u_asc[:, iy, ix]))):
   print(iy, ix)
   k = np.where(~np.isnan(tod_asc[:, iy, ix]))
   tmp = [e for e in d_asc[:, iy, ix].tolist() if e]
   cdo.select("date="+','.join(str(a) for a in tmp),
   input='u10_ANT_2016_gASCAT.nc', output='tmp.nc',
   options='-f nc')
   del tmp
   fid = nc.Dataset("tmp.nc")
   ue_asc[k, iy, ix] = np.squeeze(
   np.array(fid.variables["u10"])[:, iy, ix])
   fid.close()
   os.system('rm tmp.nc')