MODIS LST file concatenation (accessed through stackstac; STAC API)

[arindan]

Hi xarray experts and enthusiasts,

I am having a hard time finding a solution to combine/concatenate MODIS LST layers accessed through stackstac (Planetary Computer STAC API; https://github.com/microsoft/PlanetaryComputerExamples/blob/main/datasets/modis/modis-temperature-example.ipynb).

Below, I provide a reproducible example.

After accessing the data, stackstac returns an xarray.DataArray (lets say ds) with the shape (time: 24, band: 1, y: 1201, x: 1962). Although the ds have 48 timesteps, it has 12 different dates (time), that means in the ds there are two layers/scenes from same date. These two layers are spatially adjacent scenes/layers by longitude (x). So, these two scenes/layers are stacked in the ds along the time dimension, which are those 48 timesteps but from only 12 dates/days. Now, I am unable to merge these two same-time/day layers (from the same date) along the x (longitude) coordinate.

Without merging/combining the scenes/layers for the same date, I am unable to perform other regular operations such as aggregation, clipping, etc., properly.

Reproducible code:

import pystac_client
import stackstac
import planetary_computer
import xarray as xr

catalog = pystac_client.Client.open(
    "https://planetarycomputer.microsoft.com/api/stac/v1",
    modifier=planetary_computer.sign_inplace,
)

time_range = "2000-10-01/2000-12-31"
bbox = [73, 34, 80, 37]
search = catalog.search(collections=["modis-11A2-061"], bbox=bbox, datetime=time_range)
items = search.get_all_items()

ds = stackstac.stack(items, epsg=4326, chunksize='auto', assets=["LST_Day_1km"])

Here is a figure of the first two timesteps (note both are from same date, here it is Julian day no 273 of year 2000, just adjacent geographical area):

For easiness, below is the output of, ds.time, as screenshot:

Later, I tried using as: merged_ds = xr.concat(ds_d, dim='time'), but nothing change in the time dimension, it remains as 24.

Any help and solution would be highly helpful for me.