docs: move & improve STAC notebook

docs/_toc.yml

@@ -15,11 +15,11 @@ parts:
     - file: content/02/05_00_S1_SurfMI
     - file: content/02/06_00_S1_Coherence
     - file: content/02/07_00_Copernicus_DEM
+    - file: content/02/08_00_STAC_data
   - caption: How to...
     chapters:
     - file: content/03/01_00_Override_Params
     - file: content/03/02_00_Dask_Dashboard
     - file: content/03/03_00_Clip_to_vec
     - file: content/03/04_00_Spyndex
     - file: content/03/05_00_Count_valid
-    - file: content/03/06_00_STAC_data

docs/content/03/06_00_STAC_data.ipynb → docs/content/02/08_00_STAC_data.ipynb

@@ -4,30 +4,32 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# ...load data from remote STAC Catalogs?"
+    "# Remote STAC Catalogs"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "In order to load data products from remote [SpatioTemporal Asset Catalogs (STAC)](https://stacspec.org/en/), we can make use of the `load_from_stac` function provided by the `sdc-tools` package. Currently, this function supports loading data products hosted by [Microsoft Planetary Computer (MPC)](https://planetarycomputer.microsoft.com/catalog) and [Digital Earth Africa (DEA)](https://explorer.digitalearth.africa/).\n"
+    "In order to load data products from remote [SpatioTemporal Asset Catalogs (STAC)](https://stacspec.org/en/), we can make use of the `load_from_stac`-function provided by the `sdc-tools` package. Currently, this function supports loading data products hosted by [Microsoft Planetary Computer (MPC)](https://planetarycomputer.microsoft.com/catalog) and [Digital Earth Africa (DEA)](https://explorer.digitalearth.africa/).\n"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
     "```{warning}\n",
-    "Please be aware that working with remote data products might be quite inefficient. This is especially true, if the data is loaded with inappropriatly chosen parameters. Before loading a data product, you should get to know its basic characteristics. If you know the answer to at least the following questions, you are good to go:\n",
+    "Up until now, we have worked with data products that are hosted on our local file servers. The loading of these is optimized by the `sdc-tools` package. In case of remote data products, **you are responsible** for choosing the right parameters for loading the data. An inappropriate choice can potentially lead to inefficient loading times and high memory usage, so please be aware of this. Create an issue or contact me directly if you have any questions.\n",
+    "\n",
+    "Before loading a remote data product, you should get to know some of its basic characteristics. If you know the answer to at least the following questions, you are good to go:\n",
     "- **What is the pixel spacing / resolution of the data?** \n",
     "    - Override the default `resolution` parameter if necessary.\n",
-    "- **Is the data categorical or continuous?** E.g., land cover is categorical, while spectral bands are continuous.\n",
-    "    - If the data is categorical you should override the default `resampling` method to `'nearest'`.\n",
-    "- **In which datatype is the data stored and are there differences between the bands?** Common types are `uint8`, `uint16` and `float32`, for example. \n",
-    "    - If there are differences in datatype between the bands you're interested in, it's probably best to load these separately by specifiying the `bands` parameter and using the appropriate `dtype` for each band.\n",
+    "- **Is the data categorical/discrete or continuous?** E.g., land cover is categorical, while spectral bands are continuous.\n",
+    "    - If the data is categorical you should override the default `resampling` method to `'nearest'`. [Here](https://gisgeography.com/raster-resampling) you can find a short summary of a few common resampling methods.\n",
+    "- **In which data type is the product stored and are there differences between the bands?** Common types are `uint8`, `uint16` and `float32`, for example. \n",
+    "    - If there are differences in data types between the bands you're interested in, it's probably best to load these separately by specifiying the `bands` parameter and using the appropriate `dtype` for each band.\n",
     "\n",
-    "You should get an idea of how to handle these cases by having a look at the examples below. If something is unclear, please let me know!\n",
+    "You should get an idea of how to handle these cases by having a look at the examples below. I also recommend you to read the guide on how to {ref}`override-defaults` if you haven't already.\n",
     "```"
    ]
   },
@@ -36,7 +38,9 @@
    "metadata": {},
    "source": [
     "```{note}\n",
-    "In both examples we will use the bounding box of an entire SALDi site as an example. If you have a specific area of interest, you can replace the bounding box with your own. E.g., by using the utility function `sdc.vec.get_vec_bounds`. In general it is recommended to try things out on a small subset first, before scaling up to larger areas and time periods.\n",
+    "In both examples we will use the bounding box of an entire SALDi site as an example. If you have a specific area of interest, you can replace the bounding box with your own. E.g., by using the utility function `sdc.vec.get_vec_bounds` to generate a bounding box from a vector file.\n",
+    "\n",
+    "In general it is recommended to test on a small subset first, before scaling up to larger areas and time periods.\n",
     "```"
    ]
   },
@@ -618,10 +622,12 @@
     "\n",
     "bounds = get_site_bounds(site=\"site06\")\n",
     "time_range = (\"2018\", \"2023\")\n",
-    "override_defaults = {'crs': 'EPSG:4326', # this is already the default, but just to be explicit I wanted to show it here\n",
-    "                     'resolution': 0.005, # equal to approx. 500 m pixel spacing, similar to the original data\n",
-    "                     'resampling': 'nearest', # the data is categorical, so `nearest` resampling is appropriate!\n",
-    "                     'chunks': {'time': 'auto', 'y': 'auto', 'x': 'auto'}} # if you're not sure, you can set all to 'auto'\n",
+    "override_defaults = {\n",
+    "    'crs': 'EPSG:4326', # this is already the default, but just to be explicit I wanted to show it here\n",
+    "    'resolution': 0.005, # equal to approx. 500 m pixel spacing, similar to the original data\n",
+    "    'resampling': 'nearest', # the data is categorical, so `nearest` resampling is appropriate!\n",
+    "    'chunks': {'time': 'auto', 'y': 'auto', 'x': 'auto'} # if you're not sure, you can set all to 'auto'\n",
+    "    } \n",
     "\n",
     "modis_burned = load_from_stac(\n",
     "    stac_endpoint='pc',\n",
@@ -1775,10 +1781,12 @@
     "\n",
     "bounds = get_site_bounds(site=\"site04\")\n",
     "time_range = (\"2019\", \"2019\") # only a mask for the year 2019 is available\n",
-    "override_defaults = {'crs': 'EPSG:4326', # this is already the default, but just to be explicit I wanted to show it here\n",
-    "                     'resolution': 0.0001, # equal to approx. 10 m pixel spacing, similar to the original data\n",
-    "                     'resampling': 'nearest', # the data is categorical, so `nearest` resampling is appropriate!\n",
-    "                     'chunks': {'time': -1, 'y': -1, 'x': -1}} # it's a single time slice and the dtype is uint8 (\"smaller\" data), so we can load it all into one chunk\n",
+    "override_defaults = {\n",
+    "    'crs': 'EPSG:4326', # this is already the default, but just to be explicit I wanted to show it here\n",
+    "    'resolution': 0.0001, # equal to approx. 10 m pixel spacing, similar to the original data\n",
+    "    'resampling': 'nearest', # the data is categorical, so `nearest` resampling is appropriate!\n",
+    "    'chunks': {'time': -1, 'y': -1, 'x': -1} # single time slice and the small dtype (uint8), so loading it into one chunk should be fine\n",
+    "    } \n",
     "\n",
     "crop_2019 = load_from_stac(\n",
     "    stac_endpoint='deafrica',\n",
@@ -2320,13 +2328,6 @@
    "source": [
     "crop_2019.mask.plot()"
    ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": []
   }
  ],
  "metadata": {

docs/content/03/01_00_Override_Params.md

@@ -1,4 +1,5 @@
 
+(override-defaults)=
 # ...use other loading parameters with `load_product`?
 
 ```{warning}