feat(layers): HotSpot layer for the Risico mapset (#919)

[yorickdewid]

Closes Laixer/FunderMaps#919 (frontend half).

What

A HotSpot layer in the Risico mapset: buildings whose highest risk class across the two groundwater-driven models is D or E.

MAX(drystand_risk, dewatering_depth_risk) >= 'd'

Classes order a < b < c < d < e, so "the maximum" is just the worse of the two; A/B/C are excluded. It is a fast visual answer to where do we look first, not a new model.

Note

The issue body is truncated — it cuts off mid-code-fence right after the rule (775 bytes, no comments, opened 2024-11-01). The functional spec that is there is complete enough to build against, but see the open questions below.

Why it is only a JSON file

No data work is needed. maplayer.building_tiles already carries both drystand_risk and dewatering_depth_risk, and maplayer.buildings(z,x,y) emits both at every zoom — including the slimmed z12–13 attribute subset. So the whole layer is a MapLibre filter + paint expression over the existing buildings Martin source: no new column, no tile rebuild, no change to the nightly refresh.

address_count != 0 matches the guard the other risk layers already use.

Styling

Colored by the worst of the two classes, reusing the exact D and E swatches from the existing drystand-risk / dewatering-depth-risk legends, so HotSpot reads as a summary of those two rather than as a separate scale.

worst class	color
D	#ffac33
E	#ff5533

Verification

• Validates clean against @maplibre/maplibre-gl-style-spec v26.
• Compiled the filter and the color expression and evaluated all 108 combinations of (drystand, dewatering) × {null,a,b,c,d,e} × address_count {0,1,4} against the spec table — all match. The absent-attribute cases carry real weight here: 3.6M buildings have no drystand_risk at all, and MVT drops null attributes entirely rather than encoding them.

Scope today: 165,286 of 6,449,590 buildings (2.6%) qualify — 8,817 class E, 156,469 class D. Top municipalities: Rotterdam 34,746 (21%), Haarlem 12,465, Dordrecht 7,658, Leeuwarden 5,752, Deventer 5,654, Schiedam 5,550. The peat-and-pole belt, as expected.

Deploy order

Important

This PR ships first. The matching legend row is FunderMapsWorker#66 (sql/migrate/add_hotspot_layer.sql) — applying that first would put a layer id in layerSet that this repo cannot resolve, giving a dead legend entry and a console error.

Open questions for @DonZandbergen

1. One color or two? This paints D orange and E red. The issue's table only says include/exclude, so a single flat "hotspot" color is an equally valid read — one-line change.
2. Reliability gate? 106,683 of the 165,286 hotspots (65%) are indicative on both models; only 49,831 are established. The issue is silent on it. Painting 100k indicative buildings bright orange as "direct aandacht gewenst" is a fairly loud claim — worth deciding whether to exclude indicative or split the legend.

🤖 Generated with Claude Code