Plan

Season

Eight weeks of demand across four sites, with the events that cross them and the weather that moves them. Every block wears its prediction interval, because extrusion reads as fact.

Horizon56 days2026-09-23 → 2026-11-17

Marina Walk on 09-231,157947–1368 at 80%

Events in the horizon16curated, crossing every site

Least certain day10-28DXB-MAR · ±287

no WebGL heresimulated

The terrain answers when — four site lanes across 56 days, so the shape of the season is what you see first. The map answers where, with the horizon collapsed to one scrubbed day and each site’s footprint scaled by its outdoor share. Neither alone answers a promo question.

Four sites, 56 days. The lighter extension above each bar is the 80% prediction interval; the strip below is evening apparent temperature. Event bands cross every lane.

Day 1 of 56 · 2026-09-23 · 39.1°C evening

Why every block wears a cap

The cap on each block is its 80% prediction interval, so a forecast that is uncertain looks uncertain. Daily intervals combine in quadrature rather than summing: adding 24 hourly bands would assume every hour misses in the same direction at once.

Extrusion reads as fact. A 3D landscape is the most confident-looking chart there is, and without the cap a reader has no way to tell a well-known Tuesday from a guess. The cap runs from the low bound to the high one, so an uncertain day is visibly taller in its uncertainty than in its estimate.

RibbonAssumedthe weather the forecast used, not the observation table

Bands16cross every lane, because an event is not per-site

Frame loopalwaysa still image unless something asks for a frame

Without WebGLSame datafour lanes, intervals and bands, in SVG