time issue in netcdf

[FuncUny]

Hi! I am really excited to use the package. My data is a large set of high-resolusion model data (a sub-region dataset of llc4320) stored in separate netCDF files for each day (I take 10 days for test, 20120210 to 20120219), with each file containing variables like u, v, w, salt, temp, etc in 24 timesteps. I run the package successfully, but I find that the output file (here is LF_2012021X.nc, see codes below) have an error in the 'time' coordinate. Besides, a warningabout it (says the output time data won't make sense, see the last line in my main WARNING information) was also showed when running this package. Here is my codes and warning I've got:

---- My codes: ----

fdir= f'/data/llc4320_2012021X/'
fnm = f'LLC4320_pre-SWOT_ACC_SMST_2012021*.nc'
grid_dir = '/data/llc4320_2012021X/LLC4320_pre-SWOT_ACC_SMST_20120210.nc'

filenames = {
    "U": {"lat": grid_dir, "lon": grid_dir, "depth": grid_dir, "data": fdir+fnm},
    "V": {"lat": grid_dir, "lon": grid_dir, "depth": grid_dir, "data": fdir+fnm},
    "W": {"lat": grid_dir, "lon": grid_dir, "depth": grid_dir, "data": fdir+fnm},
}
variables = {"U":"U", "V":"V", "W":"W"} #,"P": "PHIHYD", "RHO": "RHOAnoma"}
dimensions = {
  "U":   {"time": "time", "depth": "Z", "lat": "YC", "lon": "XC"},
  "V":   {"time": "time", "depth": "Z", "lat": "YC", "lon": "XC"},
  "W":   {"time": "time", "depth": "Z", "lat": "YC", "lon": "XC"},
}
indices = {"depth": [36]}

f = filtering.LagrangeFilter(
    "LF_2012021X", filenames, variables, dimensions, sample_variables=["W"],
    mesh="spherical", c_grid=True, indices=indices, window_size=timedelta(days=2).total_seconds(),
)
f()

---- My main WARNING information: ----

WARNING: dt or runtime are zero, or endtime is equal to Particle.time. The kernels will be executed once, without incrementing time
WARNING:parcels.tools.loggers:dt or runtime are zero, or endtime is equal to Particle.time. The kernels will be executed once, without incrementing time
INFO: Temporary output files are stored in /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5.
INFO:parcels.tools.loggers:Temporary output files are stored in /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5.
INFO: You can use "parcels_convert_npydir_to_netcdf /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5" to convert these to a NetCDF file during the run.
INFO:parcels.tools.loggers:You can use "parcels_convert_npydir_to_netcdf /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5" to convert these to a NetCDF file during the run.
100% (172800.0 of 172800.0) |############| Elapsed Time: 0:02:14 Time:  0:02:14
INFO: Temporary output files are stored in /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5.
INFO:parcels.tools.loggers:Temporary output files are stored in /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5.
INFO: You can use "parcels_convert_npydir_to_netcdf /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5" to convert these to a NetCDF file during the run.
INFO:parcels.tools.loggers:You can use "parcels_convert_npydir_to_netcdf /data/home/zhangrd/AIswot/tmpz1k8zdo8.h5" to convert these to a NetCDF file during the run.
100% (172800.0 of 172800.0) |############| Elapsed Time: 0:02:16 Time:  0:02:16
WARNING:root:data file for field 'U' has only one of 'units' and 'calendar' attributes, the output time data won't make sense

---- Here is more information that I can give ----

1. The 'time' in my dataset is an xarray.DataArray:

array(['2012-02-10T00:00:00.000000000', '2012-02-10T01:00:00.000000000',
       '2012-02-10T02:00:00.000000000', ..., '2012-02-19T21:00:00.000000000',
       '2012-02-19T22:00:00.000000000', '2012-02-19T23:00:00.000000000'],
      dtype='datetime64[ns]'), 

with attributes:

Longname :center time of snapshots
axis :T
coverage_content_type :coordinate
standard_name :time

2. The 'time' in output dataset (LF_2012021X.nc) is like:

 array([1.329005e+18, 1.329008e+18, 1.329012e+18, …,1.329516e+18, 1.329520e+18], dtype=float32),

with attributes:

Longname :center time of snapshots
axis :T
coverage_content_type :coordinate
standard_name :time
units :hours since 2011-01-01 00:00:00

How to solve this problem? Also, I’d like to know if this issue will affect the the filtering results? (Maybe not but im not sure, because I set ±2 days window and I got 144 hours (central 6 days in the total 10 days) output in LF_2012021X.nc)
I appreciate your help!

[angus-g]

Thanks for the input! I don't actually know which component the warning is coming from, but the issue is that your time axis doesn't have a calendar attribute. However, I think that because your data only spans a relatively short time, that might not manifest as any real issue. To get rid of the warning, you should just be able to add a CF-compatible calendar attribute.

Also, I’d like to know if this issue will affect the the filtering results

As long as Parcels recognises that your time dimension represents hours (which I think it does...), it should be fine, because an absolute calendar shouldn't be needed for this!