Invalid dimensions

[helena434]

Hi,
when trying to compute the SPI for my precipitation data, I get an error about invalid dimensions. If I interpret that correctly it seems that ('time', 'latitude', 'longitude') is an incorrect order.

2021-03-03 14:09:31 INFO Computing 6-month SPI/Pearson
2021-03-03 14:09:31 ERROR Invalid dimensions for variable 'tp': ('time', 'latitude', 'longitude')
2021-03-03 14:09:31 ERROR Failed to complete
Traceback (most recent call last):
File "c:\python38\lib\site-packages\climate_indices_main_.py", line 1684, in main
compute_write_index(kwrgs)
File "c:\python38\lib\site-packages\climate_indices_main.py", line 798, in _compute_write_index
drop_data_into_shared_arrays_grid(dataset,
File "c:\python38\lib\site-packages\climate_indices_main.py", line 597, in _drop_data_into_shared_arrays_grid
raise ValueError(message)

I have now tried to reorder the dimensions, by parsing the dataset into an xarray and using transpose.

import pandas as pd
import xarray as xr

input_format = "tp_in_mm.nc"
output_format = "tp_in_mm_reordered.nc"

ds = xr.open_dataset("C:\Users\helena\precipitation\%s" % input_format,
decode_times=False)
reference_date = "1981-01-01 00:00:00"
ds['time'] = pd.date_range(start=reference_date, periods=ds.sizes['time'], freq='MS')
ds["tp"].attrs["units"] = "mm"
ds["tp"].transpose("latitude", "longitude", "time")
ds.to_netcdf(path="C:\Users\helena\precipitation\%s" % output_format)

But doing that doesn't work. If I then try to compute the SPI, I get a ton of warnings that the input contains negative values, which get clipped to zero, which shouldn't exist. I assume that somehow coordinates then get interpreted as precipitation values, but I am currently at a loss on how to solve the issue.

[Amarounek]

Hello Helena
You are probably not assigning changed object to variable...
data = data.transpose("lat", "lon", "time")

So I would say in your case:
ds["tp"] = ds["tp"].transpose("latitude", "longitude", "time")

Can you try this?

[helena434]

Ah yes, that did the trick. Thank you very much, I somehow missed that. I'm still curious how forgetting to assign the object to a variable still made some difference and changed the dataset, so that the calculation didn't cause an error.

[Amarounek]

Well, that's the basic of Python... Variables are just the names. The assignment does not copy a value. It just attaches a name to an object (data). Think about it that Variable is a sticky note... So you need to assign a new sticky note to your changed object (even if it has the same name)