Sync shared docs from grid-doctor

landingpage/shared-sources.toml

@@ -11,7 +11,7 @@
 [source.shared]
 repo = "freva-org/grid-doctor"
 ref = "main"
-commit = "6d891de1176a075c426f41a20776ec5d94ffd3eb"
+commit = "a96e532b799df46e86542c1491a96f607cddc1e7"
 source_dir = "docs/shared"
 target_dir = "shared"
 

landingpage/shared.lock

@@ -4,8 +4,8 @@
 # tell a hand edit from a sync without going to the network.
 
 [shared]
-commit = "6d891de1176a075c426f41a20776ec5d94ffd3eb"
+commit = "a96e532b799df46e86542c1491a96f607cddc1e7"
 
 [shared.files]
-"high-level-access.md" = "0ab36941e00f9fca3f36c2e0ca286cc760bff9910d8fa55c21e1e2b5cce6bc1d"
-"technical-decisions.md" = "e2ef22035790186b2bb22177275f7f8d329e73fe8e65231b2e1ec33010411757"
+"high-level-access.md" = "9dda1b30221272bd71247b61f0a420763d81502176feec4c2028e9c008d02469"
+"technical-decisions.md" = "415aab02d8dda8ba1636f48c4ad06f916c5cc1db91c956157eeb495572e5931c"

landingpage/shared/high-level-access.md

@@ -1,10 +1,10 @@
-<!-- Vendored from freva-org/grid-doctor@6d891de1 (docs/shared/high-level-access.md).
+<!-- Vendored from freva-org/grid-doctor@a96e532b (docs/shared/high-level-access.md).
      Do not edit here: changes belong upstream and will be overwritten
      by scripts/sync_shared.py. -->
 
 # Accessing High-Resolution Regional Data
 
-Some datasets in the hub cover a small area (a city, an island, a
+Some datasets cover a small area (a city, an island, a
 catchment) at very high HEALPix levels (16 is roughly 100 m, 20 roughly
 6 m cells).  These stores look different from the global pyramids in
 two ways, and both change how you load them:
@@ -143,6 +143,17 @@ coordinates and the global HEALPix indices in `berlin["cell"]`:
 berlin["t2m"].mean()
 ```
 
+!!! note "Fields with missing values"
+    For variables with NaNs (ocean-only fields, observation gaps), cells
+    on coarser levels can be only partly valid, and a plain mean over
+    cells is biased.  If the store has a `<name>_valid_fraction`
+    variable, use it as the weight:
+
+    ```python
+    ds.sst.weighted(ds.sst_valid_fraction.fillna(0)).mean("cell")
+    ```
+
+
 ### Circular regions
 
 Select a bounding box around the centre as above, then mask by
@@ -206,15 +217,15 @@ era5_region = era5_level9.isel(
 ```
 
 No interpolation, no index join.  The hierarchies coincide by
-construction, which is one of the reasons the hub uses nested ordering
+construction, which is one of the reasons to use nested ordering
 everywhere.
 
 ## Coarser overview levels
 
 High-level stores are still pyramids.  For a quick overview, open a
 coarse level of the *same* store.  Levels at or below the coordinate
 threshold are written with materialised coordinates and behave exactly
-like every other dataset in the hub, including in the browser viewer:
+like every other global pyramid, including in the browser viewer:
 
 ```python
 overview = xr.open_zarr(".../city-example.zarr/level_8.zarr", chunks=None)