Ingrid2Python

Troubleshooting

Read the pink error messages, starting at the end
These error messages are generated by the open-source code. They were added in by the programmer to be helpful
Search the web, someone else has almost always asked the question before!
<Tab> and <Shift><Tab> give contextual help in a Jupyter Notebook

For the most recent versions of xarray, we can replace an ingrid time grid just by changing its calendar and re-decoding. If xr.decode_df complains: “AttributeError: module ‘cftime’ has no attribute ‘utime’”, then you will need to hard code in the correct time using pd.date_range.

New way:

import xarray as xr
  
# Assuming the time grid is called 'T':
url = 'http://iridl.ldeo.columbia.edu/SOURCES/.NOAA/.NCEP/.CPC/.CAMS_OPI/.v0208/.mean/.prcp/dods'
ds = xr.open_dataset(url,decode_times=False)
ds['T'].attrs['calendar'] = '360_day'  # Fix the calendar
ds = xr.decode_cf(ds)                  # Now we can decode the time grid!

Old way:

import xarray as xr
import pandas as pd
  
url = 'http://iridl.ldeo.columbia.edu/SOURCES/.NOAA/.NCEP/.CPC/.CAMS_OPI/.v0208/.mean/.prcp/dods'
ds = xr.open_dataset(url,decode_times=False)
ds['T'] = pd.date_range('1979-01',periods = len(ds['T']),freq='MS').shift(15, freq='D')

If your dataset, ds, is giving trouble with the time grid, time, (as happens frequently), just replace it!

So here is my crude recipe to replace troublesome standard time grids with time-centered datetime64. Note that there are many variations of frequencies, so please check your time grid after replacement and adjust the recipe accordingly.

# Replace a time grid which starts at the first of the month to one which is centered in the month
url = 'http://kage.ldeo.columbia.edu:81/SOURCES/.LOCAL/.sst.mon.mean.nc/dods'
ds = xr.open_dataset(url)

first_time = str(ds.time.values[0])
  
freq = xr.infer_freq(ds.time)
print(freq)
time = pd.date_range(first_time, periods=len(ds.time), freq=freq)
if 'D' in freq:
    ds['time'] = time.shift(12, freq='H') 
elif 'M' in freq:   
    ds['time'] = time.shift(15, freq='D') 
elif 'A' in freq:
    ds['time'] = time.shift(6, freq='M') 
ds.time

An xarray DataSet contains xarray DataArrays. Each DataArray usually corresponds to what we think of as one variable. A good convention, so you remember which you are dealing with, is to call Datasets: ds_{name} and DataArrays: da_{name}.
You cannot plot a Dataset, only a DataArray. A pink error results:

You can combine DataArrays in one Dataset
You can convert a DataArray into a Dataset - if the DataArray doesn’t know the name of the variable, you can specify it.

For our usual example:

url = 'http://kage.ldeo.columbia.edu:81/SOURCES/.LOCAL/.sst.mon.mean.nc/.sst/dods'
ds = xr.open_dataset(url)
da = ds.sst
ds_new = da.to_dataset(name='sst')

import xarray as xr
# Print version of a package
print(xr.__version__)
  
url = 'http://kage.ldeo.columbia.edu:81/SOURCES/.LOCAL/.sst.mon.mean.nc/.sst/dods'
ds = xr.open_dataset(url)  
# Print version of a package
print(ds.info())
  
# print size of dataset in megabytes
print(ds.nbytes/1e6,'M')
  
# print names and sizes of dimensions
print(ds.dims)

import xarray as xr

url = 'http://iridl.ldeo.columbia.edu/SOURCES/.WCRP/.GCOS/.GPCC/.FDP/.version2018/.1p0/.prcp/dods'

ds = xr.open_dataset(url,decode_times=False,chunks=-1).load()
ds = ds.rename({'T':'time'}).sortby('Y').sel(X=slice(-120,-95),Y=slice(25,40))
dst = ds.mean(['Y','X'])
dst.prcp.plot()

# Switch North to South 
ds = ds.sortby('lat')
  
# Longitude [-180,180] to [0,360]
import numpy as np
ds.coords['lon'] = np.mod(ds['lon'], 360)
ds = ds.reindex({ 'lon' : np.sort(ds['lon'])})

# Longitude [0,360] to [lon_start,360+lon_start] 
ds.coords['lon'] = (ds.coords['lon'] - lon_start) % 360 + lon_start
ds = ds.sortby(ds.lon)

plt.tight_layout()