"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "dataarray.isel(band=0).plot.imshow()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "1d30749b",
+ "metadata": {},
+ "source": [
+ "This shows some flooded rice paddy fields in Cambodia, with 0 being dry land and 1\n",
+ "being flooded water areas. You can see the location on a map by visiting this\n",
+ "[preview page on source.coop](https://source.coop/c2sms/c2smsfloods/ms-dataset-chips/3e71513a-ec54-4e9b-aae6-93a8a3648763/s2/S2A_MSIL1C_20180804T031541_N0206_R118_T48PWU_20180804T061739_09053-06778/LabelWater.tif)."
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "388a8ff8-9945-4214-ae10-586b3f88a0b3",
+ "metadata": {},
+ "source": [
+ "## Concatenate multiple single-band GeoTIFFs\n",
+ "\n",
+ "Let's now try a slightly trickier set up - with bands stored in different files!\n",
+ "\n",
+ "We'll pull the optical satellite bands corresponding to the water mask above.\n",
+ "Specifically, we'll read some Sentinel-2 bands using\n",
+ "[`xr.open_mfdataset`](https://docs.xarray.dev/en/stable/generated/xarray.open_mfdataset.html),\n",
+ "ensuring that the list of URLs/filepaths correspond to the band order we want.\n",
+ "\n",
+ "Note that some parameters are different than before:\n",
+ "- `concat_dim=\"band\"` and `combine=\"nested\"`: ensures that the bands are concatenated\n",
+ " as data_vars along the dimension 'band'\n",
+ "- `device_id=None` (actually the default setting): so that the data is decoded by\n",
+ " whatever GPU is available (usually this is the first CUDA GPU device, unless you are\n",
+ " using dask-cuda with multiple GPUs)."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "id": "0d00d078-9c04-469b-80bb-e93d88c5b463",
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "