The issue with Dimensions while saving the bias corrected data output?

[abhilashsinghimd]

The plots seems fine but the NetCDF are not being written/exported properly.

The dimensions should have been time, lat, long (365, 34, 65)
But somehow there are 34 bands i.e. latitude somehow became band i.e number of timesteps

#!/usr/bin/env python

coding: utf-8

In[1]:

from bias_correction import XBiasCorrection
from dask.diagnostics import ProgressBar
import numpy as np
import xarray as xr
import pandas as pd
import matplotlib.pyplot as plt
import os

In[2]:

obs_data = np.random.randn(2365, 34, 65)
model_data = np.random.randn(2365, 34, 65)
sce_data = np.random.randn(365, 34, 65)

lat = range(34)
lon = range(65)

obs_data = xr.DataArray(obs_data, dims=['time','lat','lon'], coords=[pd.date_range('2010-01-01', '2011-12-31', freq='D'), lat, lon])
model_data = xr.DataArray(model_data, dims=['time','lat','lon'], coords=[pd.date_range('2010-01-01', '2011-12-31', freq='D'), lat, lon])
sce_data = xr.DataArray(sce_data, dims=['time','lat','lon'], coords=[pd.date_range('2021-01-01', '2021-12-31', freq='D'), lat, lon])

ds = xr.Dataset({'model_data':model_data,'obs_data':obs_data, 'sce_data':sce_data})
ds['sce_data']

bc = XBiasCorrection(ds['obs_data'], ds['model_data'], ds['sce_data'])
df3 = bc.correct(method='modified_quantile')

In[3]:

bc.sce_data.sel(time='2021-12-15').plot(figsize=(14, 5), robust=True)
plt.show()

In[4]:

df3.sel(time='2021-12-15').plot(figsize=(14, 5), robust=True)
plt.show()

In[5]:

########### convert dataarray to Dataset #################
out_var_name = "tasmax"
dataset_out = df3.to_dataset(name=out_var_name)
dataset_out

The time dimension switches with latitude somehow while saving this dataset as NetCDF.

In[6]:

for year in range (2021,2022):
out_file_dir = r"F:\CMIP6\Raw_CMIP_Data\Bias_Correction\Bias_Corrected\Tmax\"
if not os.path.exists(out_file_dir):
os.makedirs(out_file_dir)
out_file_name = out_file_dir + str(year)+"_Biases_Corrected_Tmax.nc"
print(out_file_name)
yearly_ds = df3.where(((dataset_out['time.year'] == year)), drop=True)
write_job = yearly_ds.to_netcdf(out_file_name, compute=False)
with ProgressBar():
print(f"Writing to {year}")
write_job.compute()

In[7 ]:

Details are here at this link.

https://stackoverflow.com/questions/73196955/dimensions-position-error-in-converting-xarray-dataarray-to-xarray-dataset

How to rectify it?

#The NetCDF is attached here:
2021_Biases_Corrected_Tmax.zip