Fix STAC coordinate system and dimension detection

Corrected a metadata bug where projected coordinates (e.g., in meters
for HRRR) were incorrectly labeled as EPSG:4326 (degrees) in the STAC
Datacube extension. Added explicit CRS extraction via rioxarray and
auto-detection of x/y dimension names to ensure the catalog accurately
reflects the dataset's coordinate system.

Author: rsignell

examples/build_dynamical_catalog_full.ipynb

@@ -222,8 +222,16 @@
     "            # 1. BBox via rioxarray\n",
     "            bbox = extract_spatial_extent_rio(ds)\n",
     "            \n",
-    "            # 2. Temporal\n",
+    "            # 2. Dimensions & CRS\n",
     "            time_dim = next((d for d in [\"time\", \"init_time\"] if d in ds.dims), None)\n",
+    "            x_dim = next((d for d in [\"lon\", \"longitude\", \"x\"] if d in ds.dims), None)\n",
+    "            y_dim = next((d for d in [\"lat\", \"latitude\", \"y\"] if d in ds.dims), None)\n",
+    "            \n",
+    "            try:\n",
+    "                ref_sys = ds.rio.crs.to_epsg() or ds.rio.crs.to_wkt()\n",
+    "            except Exception:\n",
+    "                ref_sys = 4326\n",
+    "\n",
     "            start_dt = end_dt = None\n",
     "            if time_dim:\n",
     "                start_dt = datetime.fromisoformat(str(ds[time_dim].min().values[()]).split(\".\")[0])\n",
@@ -277,7 +285,15 @@
     "            # 6. Datacube Extension via xstac (if available)\n",
     "            try:\n",
     "                import xstac\n",
-    "                item = xstac.xarray_to_stac(ds, item, validate=False)\n",
+    "                item = xstac.xarray_to_stac(\n",
+    "                    ds, \n",
+    "                    item, \n",
+    "                    temporal_dimension=time_dim, \n",
+    "                    x_dimension=x_dim, \n",
+    "                    y_dimension=y_dim, \n",
+    "                    reference_system=ref_sys,\n",
+    "                    validate=False\n",
+    "                )\n",
     "            except ImportError:\n",
     "                cube = build_datacube_extension(ds, start_dt.isoformat() + \"Z\" if start_dt else None, end_dt.isoformat() + \"Z\" if end_dt else None)\n",
     "                item.properties.update(cube)\n",

scripts/build_dynamical_stac.py

@@ -433,6 +433,22 @@ def build_item_for_store(
 
     temporal_dim = detect_temporal_dimension(ds)
 
+    # Detect horizontal dimension names and CRS for xstac
+    x_dim = y_dim = None
+    for x in ["lon", "longitude", "x"]:
+        for y in ["lat", "latitude", "y"]:
+            if x in ds.dims and y in ds.dims:
+                x_dim, y_dim = x, y
+                break
+        if x_dim: break
+
+    ref_sys = 4326
+    try:
+        if hasattr(ds, "rio") and ds.rio.crs:
+            ref_sys = ds.rio.crs.to_epsg() or ds.rio.crs.to_wkt()
+    except Exception:
+        pass
+
     description = entry.get("Description", "").strip()
     description += xarray_open_snippet(item_id, catalog_url)
 
@@ -457,6 +473,9 @@ def build_item_for_store(
             providers=[DYNAMICAL_PROVIDER],
             virtual=False,
             temporal_dimension=temporal_dim,
+            x_dimension=x_dim,
+            y_dimension=y_dim,
+            reference_system=ref_sys,
         )
     except Exception as exc:
         log.warning("  Failed to build STAC item for %s: %s", store_uri, exc)