Grouping by the hour angle adjusted day of year

[jarataraj]

I am creating global charts of daily UTCI highs (UTCI is a thermal stress index). Grouping by time.dt.dayofyear would not account for local time, resulting in a chart that is accurate only for longitudes near longitude 0. For other longitudes, the chart would show the highs for a 24-hour period that is not from around midnight to midnight. Rather than mess with time zones, I adjust the time using the longitude's nearest hour angle (using example data file 'utci.nc' from: utci.zip):

import xarray as xr
import numpy as np
import pandas as pd

ds = xr.open_dataset('utci.nc')
hour_angle = (ds.lon/15).round()
utc_hour_angle = xr.where(hour_angle > 12, hour_angle - 24, hour_angle)
utc_hour_angle.values = pd.to_timedelta(utc_hour_angle, 'h')
ds['hour_angle_adjusted_time'] = da.time + utc_hour_angle
ds = ds.set_coords('hour_angle_adjusted_time')
ds

Output:

To get daily highs, I'd like to do something similar to ds.groupby("time.dayofyear").max(dim='time') but the coordinates of ds.groupby('hour_angle_adjusted_time.dayofyear) are "lat" and "stacked_time_lon" as seen here:

next(iter(ds.groupby('hour_angle_adjusted_time.dayofyear')))

Output:

(35,
 <xarray.Dataset>
 Dimensions:                   (lat: 181, stacked_time_lon: 30)
 Coordinates:
   * lat                       (lat) float64 -90.0 -89.0 -88.0 ... 88.0 89.0 90.0
     hour_angle_adjusted_time  (stacked_time_lon) datetime64[ns] 2023-02-04T22...
   * stacked_time_lon          (stacked_time_lon) object MultiIndex
   * time                      (stacked_time_lon) datetime64[ns] 2023-02-05T09...
   * lon                       (stacked_time_lon) float64 188.0 189.0 ... 217.0
 Data variables:
     UTCI                      (lat, stacked_time_lon) float64 -32.12 ... -58.55)

So, I am looking for some guidance on how best to approach creating charts of daily highs that are time-zone-aware or hour-angle-aware data. Perhaps I am on the right track or perhaps my beginner's understanding has led me astray. My thoughts are that I could try shifting the data depending its longitude. Another idea is that pandas appears to be able to deal with time zones, so I could convert to a dataframe, and use a library like pytzwhere to label (lat, lon) coords with a time zone, and then create a new day of the year index using the time and time zone for each lat and lon. Then convert back to xarray for charting. I haven't used pandas much before though so I am not sure if this is feasible.

Thank you for your help!

[jarataraj]

I've since learned about xr.DataArray.unstack and can use group.unstack('stacked_time_lon') on each group (grouped by dayofyear), but I am still unsure if this is the best approach for time zone or hour angle aware grouping.