🚸 Walkthrough on handling multi-resolution climate data (#91)

Tutorial on loading time-series climate datasets with different spatial resolutions for a super-resolution task. Looking at one of the CMIP6 global coupled ocean-atmosphere general circulation model (GCM) outputs and CarbonPlan's super-resolution (DeepSD) product.

* 🚸 Walkthrough on handling multi-resolution climate data

Initial draft tutorial on loading climate datasets with different spatial resolutions for a super-resolution task. Will be looking at one of the CMIP6 model outputs and CarbonPlan's super-resolution (DeepSD) product.

* ➕ Add xarray-datatree

Hierarchical tree-like data structures for xarray!

* 💫 Plot GCM and GARD-MV projected max temperature over South Asia

Show a side by side comparison of daily maximum near-surface air temperature in Aug 2089 over South Asia from the Global Climate Model (GCM) and GARD-MV dowscaled product. Adapted from Figure 1 of https://carbonplan.org/research/cmip6-downscaling-explainer

Also converted arc degree units to arc minutes and fixed a broken link to the Zarr docs.

* 📝 Writeup section on stacking datasets into a DataTree stucture

Show how to stack xarray.Dataset objects with different spatial resolutions into a nested DataTree structure. Showing the DataPipe graph too for good measure. Also added intersphinx link for xarray-datatree.

* 📝 Writeup section on spatiotemporal subsetting of multires datasets

Show how to select data from a DataTree object that fits within a particular geographical bounding box and time range. Had to convert the low-resolution GCM's longitude coordinates from 0-360 to -180-180 first to match the DeepSD grid. Plotted the subsetted output for the Philippines on Dec 2030, and added a note about how one should approach pixel alignment and reprojection/resampling before performing spatial subsets.

* 📝 Writeup final section on time-series slicing and dataloading

Completing the multi-resolution tutorial hastily. Binning the temperature data in a monthly manner and chucking it into DataLoader2. Hopefully able to do multi-resolution spatial subsetting some day with changes in xbatcher!

* ✏️ Reword lesson outline and fix some typos

Clarify that the climate dataset slicing is only on the time dimension and not the spatial dimensions for now. Also fixed a few typos here and there.

docs/_config.yml

@@ -48,6 +48,9 @@ sphinx:
       datashader:
         - 'https://datashader.org/'
         - null
+      datatree:
+        - 'https://xarray-datatree.readthedocs.io/en/latest/'
+        - null
       geopandas:
         - 'https://geopandas.org/en/latest/'
         - null

docs/_toc.yml

@@ -15,6 +15,8 @@ chapters:
         file: object-detection-boxes
       - title: 🏳️‍🌈 Stacking layers
         file: stacking
+      - title: 📶 Multi-resolution
+        file: multi-resolution
   - title: 📖 API Reference
     file: api
   - title: 📆 Changelog