[GH-2700] Add 03-fire-risk-fusion notebook: raster + vector fusion

[jiayuasu]

Did you read the Contributor Guide?

• Yes, I have read the Contributor Rules and Contributor Development Guide

Is this PR related to a ticket?

• Yes, and the PR name follows the format [GH-XXX] my subject. Closes part of Sedona example notebooks in the docker image are very out of date #2700.

What changes were proposed in this PR?

Continues the docker-image notebook refresh series (issue #2700, milestone 1.9.1). Adds the first notebook in the series that mixes raster algebra, raster→vector zonal aggregation, and vector-on-vector distance joins in one pipeline — the workflow GeoPandas alone can't do.

docs/usecases/03-fire-risk-fusion.ipynb answers:

Given a county's terrain steepness, fuel load, building footprints, and road network, score every building for wildfire risk weighted by distance from the nearest evacuation route.

End-to-end:

1. SedonaContext setup.
2. Synthesize slope.tif + fuel.tif (256×256 single-band float32 tiled GeoTIFFs in /tmp/fire-risk/). Slope highest in the east, fuel highest in the north.
3. Load both with sedona.read.format("raster") (auto-tiling, Add a new raster data source reader that can automatically tile GeoTiffs and bypass the Spark record limit #2672), keep (x, y) tile-index columns, join on those, compute composite risk via two-raster RS_MapAlgebra (0.5 * slope + 0.5 * fuel). The same SQL works for single-tile inputs and for multi-tile DEM-sized scenes.
4. Build a 4×4 grid of building polygons + two bisector LINESTRING roads as Spark DataFrames.
5. Compute each building's distance to its nearest road via MIN(ST_DistanceSpheroid) over the building × road cross product (metres regardless of EPSG:4326 lon/lat units).
6. Score: RS_ZonalStats(composite, footprint, 'mean') × (1 + min(dist_km, 5) / 5). Multiplicative form means a building only ranks high when it has both high terrain risk and poor road access.
7. Rank, write top-5 as GeoParquet 1.1 (auto covering-bbox + projjson), round-trip read back to verify.
8. matplotlib panel: composite risk as basemap, building footprints filled by risk_score (red = high), roads overlaid, top-5 buildings labelled.

Built-in ground truth: slope-east + fuel-north synthesis means corner buildings carry the highest composite risk; the multiplicative evacuation factor then favours buildings far from the bisector roads. B33 (NE corner) should rank top, which the harness confirms.

Notebook is structured as numbered markdown sections (## 1. through ## 7.), matching the convention from the prior notebooks. Notebook intro flags **Requires Sedona ≥ 1.9.0** for the auto-tiling raster reader.

No new data shipped. No network required.

How was this patch tested?

End-to-end through the local mirror of docker/test-notebooks.sh (matched docker stack: Python 3.10, pyspark==4.0.1, apache-sedona==1.9.0, JDK 17, local[*], DRIVER_MEM=4g, Sedona JAR via PYSPARK_SUBMIT_ARGS Maven coords).

PASS  03-fire-risk-fusion  19s elapsed

Output sanity-checked:

bid	mean_risk	dist_km	risk_score
B33	0.8772	5.21	1.7543
B23	0.7511	4.29	1.396
B32	0.7524	3.42	1.2675
B13	0.6218	4.29	1.1558
B31	0.6226	3.42	1.0489
…
B00	0.1244	5.21	0.2488

Top-ranked building B33 (north-east corner) matches the synthesis design; the entire NE quadrant clusters at the top of the ranking; B00 (SW corner) ranks bottom with mean_risk≈0.12 (lowest slope + lowest fuel). GeoParquet top-5 round-trip read back identical rows.

The Docker-build CI workflow (path-filter widening landed in #2889) will run on this PR and execute test-notebooks.sh against the built image, so the in-container PASS line lands directly in CI.

Did this PR include necessary documentation updates?

• The notebook is itself the documentation; intro markdown calls out **Requires Sedona ≥ 1.9.0** and explains both the multi-tile join pattern and the multiplicative score rationale.
• No new data shipped, so no docs/usecases/data/README.md updates.
• No public API changes.

[Copilot]

Pull request overview

This PR adds a new Sedona use-case notebook that demonstrates a raster+vector wildfire-risk workflow in the refreshed Docker example suite. It fits the ongoing notebook modernization effort by showcasing raster algebra, raster-to-vector zonal aggregation, distance-based vector analysis, GeoParquet output, and plotting in one end-to-end example.

Changes:

• Add 03-fire-risk-fusion.ipynb, a synthetic wildfire-risk scoring notebook built on Sedona raster and geometry SQL functions.
• Demonstrate composite-risk raster creation from synthesized slope and fuel GeoTIFFs, then score synthetic building footprints by zonal mean risk and road distance.
• Persist the top-ranked buildings as GeoParquet and visualize the final result with matplotlib.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[jiayuasu]

Pushed d1d895db58.

Review point	Action
Per-row RS_ZonalStats('mean') over buildings × composite produces per-tile means and silently wrong scores when composite has more than one tile	Restructured to a two-step aggregation: per (building, tile) RS_ZonalStats(rast, geom, 'sum') and 'count' gated by RS_Intersects, then SUM(tile_sum) / SUM(tile_cnt) per building. Pixel-count-weighted, multi-tile-correct, single-tile result identical to the prior version.
buildings × roads cartesian for nearest-road won't scale to county OSM	Replaced with JOIN roads ON ST_KNN(b.geom, r.geom, 1, false) — indexed nearest-neighbour join, returns one row per building with the actual ST_DistanceSpheroid computed alongside.
Synthetic geometries had SRID 0, so the GeoParquet 1.1 writer wouldn't auto-populate projjson	Wrapped every ST_GeomFromText(...) with ST_SetSRID(..., 4326) so the writer can derive projjson CRS metadata as the section prose claims.
Intro requirement note pointed at ST_Transform, which the notebook never calls	Reattributed to the actual 1.9-only features the notebook uses: the auto-tiling raster reader (GH-2672) and the GeoParquet 1.1 writer's auto-populated covering-bbox + projjson CRS (GH-2646, GH-2664).

Re-verified end-to-end through the local mirror of docker/test-notebooks.sh after the changes:

PASS  03-fire-risk-fusion  18s elapsed

Ranking unchanged (B33 top with risk_score=1.7543, B00 bottom with 0.2488) — the new SQL is mathematically equivalent to the old one for single-tile inputs but now stays correct under scaling.