[Bug]: regrid2 converts time coordinates to int instead of preserving original datetime objects after annual cycle climatology

[lee1043]

What happened?

When proceeding regrid2 after getting an annual cycle, somehow it loses the time coordinate data array (see below log print 4).

What did you expect to happen? Are there are possible answers you came across?

It does not happen when using xesmf regridder. time coordinate has to be preserved (see below log print 5).

Minimal Complete Verifiable Example (MVCE)

import xcdat as xc

# Prepare data
target_grid = xc.create_uniform_grid(-90, 90, 5, 0, 180, 5)
f = "/p/user_pub/PCMDIobs/obs4MIPs/ECMWF/ERA-INT/mon/pr/gn/v20210727/pr_mon_ERA-INT_PCMDI_gn_197901-201903.nc"

# Open data and test
ds = xc.open_dataset(f) # 1

# Immediate interpolation
ds_interp = ds.regridder.horizontal('pr', target_grid, tool='regrid2') # 2

# Interpolation after getting climatology
ds_ac = ds.temporal.climatology('pr', freq="month", weighted=True) # 3, annual cycle
ds_ac_interp_regrid2 = ds_ac.regridder.horizontal('pr', target_grid, tool='regrid2') # 4 interp (regrid2)
ds_ac_interp_xesmf = ds_ac.regridder.horizontal('pr', target_grid, tool='xesmf', method='bilinear')  # 5 interp (xesmf)

print(1, list(ds.coords))
print(2, list(ds_interp.coords))
print(3, list(ds_ac.coords))
print(4, list(ds_ac_interp_regrid2.coords))
print(5, list(ds_ac_interp_xesmf.coords))

Relevant log output

1 ['lat', 'lon', 'time']
2 ['lon', 'lat', 'time']
3 ['lat', 'lon', 'time']
4 ['lon', 'lat']
5 ['time', 'lon', 'lat']

Anything else we need to know?

I guess difference between regrid2 and xesmf output is in time array. From regrid 2, it is a data array with index numbers 0 to 11, but from xesmf, it is cftime object.

• time before interpolation

ds_ac.time

xarray.DataArray'time'time: 12
array([cftime.DatetimeGregorian(1, 1, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 2, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 3, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 4, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 5, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 6, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 7, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 8, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 9, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 10, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 11, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 12, 1, 0, 0, 0, 0, has_year_zero=False)],
      dtype=object)
Coordinates:
time (time) object 0001-01-01 00:00:00 ... 0001-12-...
Indexes: (1)
Attributes: 
bounds : time_bnds
axis : T
long_name : time
standard_name : time

• time after interpolation using regrid2

ds_ac_interp_regrid2.time

xarray.DataArray'time'time: 12
array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11])

• time after interpolation using xesmf

ds_ac_interp_xesmf.time

xarray.DataArray'time'time: 12
array([cftime.DatetimeGregorian(1, 1, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 2, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 3, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 4, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 5, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 6, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 7, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 8, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 9, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 10, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 11, 1, 0, 0, 0, 0, has_year_zero=False),
       cftime.DatetimeGregorian(1, 12, 1, 0, 0, 0, 0, has_year_zero=False)],
      dtype=object)

Environment

Error occurred with xcdat 0.7.1 and 0.7.2

[tomvothecoder]

Hey @lee1043, is this a major disruptive bug that we should prioritize?

[lee1043]

Hey @lee1043, is this a major disruptive bug that we should prioritize?

Hi @tomvothecoder, thank you for checking it with me. I believe so because I think there is a very good chance this will start impacting PMP and obs4MIPs workflow.

[tomvothecoder]

@jasonb5 do you have time to work on this within the next few weeks? If not, I can take a look at it.

[jasonb5]

@tomvothecoder Yea I can look at this in the next few weeks.