IndexError when including fields on the T-grid #1968
-
QuestionQuestionHello, I've been trying to update my scripts to use them with v3.1.2. Previously, it all worked with v2.3.1. While executing, I get the error message I hope you can help me with that, thank you! Supporting code/error messagesdef fieldset_defintions(
list_of_filenames_U, list_of_filenames_V,
list_of_filenames_W, list_of_filenames_T,
mesh_mask
):
filenames = {'U': {'lon': (mesh_mask), 'lat': (mesh_mask), 'depth': list_of_filenames_W[0],
'data': list_of_filenames_U},
'V': {'lon': (mesh_mask), 'lat': (mesh_mask), 'depth': list_of_filenames_W[0],
'data': list_of_filenames_V},
'W': {'lon': (mesh_mask), 'lat': (mesh_mask), 'depth': list_of_filenames_W[0],
'data': list_of_filenames_W},
'T': {'lon': (mesh_mask), 'lat': (mesh_mask), 'depth': list_of_filenames_T[0],
'data': list_of_filenames_T},
}
variables = {'U': 'vozocrtx',
'V': 'vomecrty',
'W': 'vovecrtz',
'T': 'votemper',
}
dimensions = {'U': {'lon': 'glamf', 'lat': 'gphif', 'depth': 'depthw',
'time': 'time_counter'},
'V': {'lon': 'glamf', 'lat': 'gphif', 'depth': 'depthw',
'time': 'time_counter'},
'W': {'lon': 'glamf', 'lat': 'gphif', 'depth': 'depthw',
'time': 'time_counter'},
'T': {'lon': 'glamt', 'lat': 'gphit', 'depth': 'deptht',
'time': 'time_counter'},
}
return FieldSet.from_nemo(
filenames, variables, dimensions,
chunksize=False,
mesh='spherical',
tracer_interp_method='cgrid_tracer',
allow_time_extrapolation=True,
)
def create_fieldset(
data_path=data_path, experiment_name=experiment_name,
fname_U=fname_U, fname_V=fname_V, fname_W=fname_W, fname_T=fname_T,
mesh_mask = mesh_mask
):
files_U = list(sorted((data_path / 'output').glob(fname_U)))
files_V = list(sorted((data_path / 'output').glob(fname_V)))
files_W = list(sorted((data_path / 'output').glob(fname_W)))
files_T = list(sorted((data_path / 'output').glob(fname_T)))
fieldset = fieldset_defintions(
files_U, files_V,
files_W, files_T, mesh_mask)
return fieldset
OW_eddy = Field.from_netcdf(
filenames=OW_path,
variable=("OW_eddy", "type"),
dimensions={"lon": "nav_lon", "lat": "nav_lat", "depth": "nav_lev", "time": "time_counter"},
interp_method="nearest",
)
fieldset.add_field(OW_eddy)
OW_area = Field.from_netcdf(
filenames=OW_path,
variable=("OW_area", "area"),
dimensions={"lon": "nav_lon", "lat": "nav_lat", "depth": "nav_lev", "time": "time_counter"},
interp_method="nearest",
)
fieldset.add_field(OW_area)
OW_scale = Field.from_netcdf(
filenames=OW_path,
variable=("OW_scale", "scale"),
dimensions={"lon": "nav_lon", "lat": "nav_lat", "depth": "nav_lev", "time": "time_counter"},
interp_method="nearest",
)
fieldset.add_field(OW_scale)
fieldset = create_fieldset()
fieldset.computeTimeChunk(0,1) |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 4 replies
-
|
I think I should've shown everything above... I also read in MLD and SSH (the same way I did for T), but they of course don't have a depth dimension. So this is working now. For the OW-variables, the depth axis is different (48 instead of the 120 levels which the others have) and the order of dimensions in the ncs is different. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for raising this problem, @LeonMock. Do you use the released conda version 3.1.2, or the latest I haven't seen this problem before, but it could be that the issue is because your files contain multiple timeslices. What happens if you try to load only one of your files? Does that work? Also, in your code above I'm wondering how you can first add the OW fields to a FieldSet, and only create the fieldset itself later. Is this really the code you use? Since the details might matter, please use an actually working/breaking code for us to look at. Finally, would it be possible to share some of the files (perhaps a subset in space?) via a file transfer facility, so that we can have a look ourselves and see what's going wrong? |
Beta Was this translation helpful? Give feedback.
-
|
Hi, I am using the conda version 3.1.2. Here is a link to the beginning of the script, the environment and some cut-outs of the data. I hope they're small enough. https://cloud.geomar.de/s/FkoqeKqWZJZqzHA As said above, I now managed to set up the initial FieldSet (SSH and MLD without depth dim) but adding the OW fields still does not work (of course I add them after creating, I mixed that up above). I am aware that the depth axis of the OW fields is not the same, is that the issue? |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for sharing this file, @LeonMock. It helped me dig a bit deeper into what's going on. And I think/hope I found a working solution for you... If you explicitly set the Add a dictionary with indices = {
'U': {'depth': range(0, 119)},
'V': {'depth': range(0, 119)},
'W': {'depth': range(0, 119)},
'T': {'depth': range(0, 119)},
'S': {'depth': range(0, 119)},
'SSH': {},
'MLD': {},
'OW_eddy': {'depth': range(0, 24)},
'OW_area': {'depth': range(0, 24)},
'OW_scale': {'depth': range(0, 24)},
}And then add Can you check whether this works for you too? |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for tackling this, this works, but I am afraid this is not the final solution yet. It somehow seems to mix up the time and depth axes. For the OW-files, the depth axis is 48 long. So if I set the indices like you showed me (range(0,24)), it works, but then I don't have the whole depth axis, right? Also, the length of the time axis is different in some files, so I also couldn't slice that in the indices dict. If that is the issue... |
Beta Was this translation helpful? Give feedback.
-
|
Hi @LeonMock, sorry for the confusion; I didn't realise indeed that depth and time were swapped; so the What could be a proper solution, though, is to swap Note that this time of problem is exactly why we want to move to stronger leverage of |
Beta Was this translation helpful? Give feedback.
-
|
Thank you @erikvansebille for your help! |
Beta Was this translation helpful? Give feedback.
Hi @LeonMock, sorry for the confusion; I didn't realise indeed that depth and time were swapped; so the
dimensionsfix isn't a real fix, indeed.What could be a proper solution, though, is to swap
timeanddepthin your OW files. If dods = ds.transpose("time_counter", "z", "y", "x")and then save the new dataset, the code seems to work. Could you try too?Note that this time of problem is exactly why we want to move to stronger leverage of
xarrayin the new Parcels v4...