Correct the region Q3 describes

[jiayuasu]

Q3 filters trips against this ring:

POLYGON((-111.9060 34.7347, -111.6160 34.7347, -111.6160 35.0047, -111.9060 35.0047, -111.9060 34.7347))

Everywhere it appears, it is described as a "10km bounding box" which, with the 0.045° buffer, gives a "15km radius". Neither figure is right.

The ring spans 0.290° of longitude by 0.270° of latitude. At its centre latitude (34.8697 °N, which is exactly the Sedona centre the other queries use):

Longitude	0.290° × 111.32 km × cos(34.87°) ≈ 26.5 km east–west
Latitude	0.270° × 110.94 km ≈ 30.0 km north–south
From the centre	13.2 km E–W, 15.0 km N–S
Corners	√(13.2² + 15.0²) ≈ 20.0 km, or 25.0 km once the buffer is added

So the "10km box" label understates the region by about 3×, and the query reaches ~25 km from the centre rather than 15 km.

What changes

Comments and prose only. The polygon literal, the thresholds and the result sets are untouched, so the committed ground-truth answers still apply and nothing needs regenerating.

• spatialbench-queries/print_queries.py — the SQL source of truth: the -- Q3: header and the inline polygon comment
• spatialbench-queries/{geopandas_queries,spatial_polars,pycanopy_queries}.py — the three dataframe implementations repeat the claim in their Q3 docstrings
• notebooks/queries.ipynb — the source of the rendered page
• docs/queries.md, docs/queries.zh.md — heading and prose, kept in sync with the notebook

The buffer's existing ~5km annotation for 0.045° is deliberately left alone. It is accurate in latitude (0.045° ≈ 5.0 km) and follows the same degree-to-kilometre convention the rest of the geometry suite uses for 0.45 "= 50 km" and 0.0045 "≈ 500 m" — revisiting that convention is a separate discussion, and it is the reason the geography suite in #131 exists.

Found while writing the geography counterpart of this query in #131, where the same label had been carried over.

Verification

python3 spatialbench-queries/print_queries.py SedonaDB   # 12 queries, SQL unchanged
mkdocs build --strict

git diff is 13 lines across 7 files; the notebook round-trips with only the two intended strings changed, and docs/queries.md is verified still in sync with the notebook's Q3 cell.