Eyes on the Park — why green parks fail to cool, and how to fix them

The problem, in four numbers

Ten years of hourly thermal-comfort reanalysis at one Dubai neighbourhood park.

54.7 °C

worst-day thermal comfort index (UTCI) in the sun; 49.8 °C even in shade

3.5–4.9 h

of extreme heat stress (UTCI ≥ 46 °C) every day, every summer

+0.4 °C

is all the cooling this 85%-green park gives its neighbourhood

street shade at 3 pm on the walk to the park: no cool route exists

Ten-year hourly UTCI heat calendar — Ten years, hour by hour: extreme heat stress recurs every summer. It is a design condition, not an anomaly.

Greenery is the wrong metric

We screened every OSM-mapped park in Dubai — 145 parks, one method, one afternoon, one laptop. Green share explains only ~3% of the variation in how much a park cools its surroundings. The four strongest coolers are all pond parks. What cools is dense canopy shade, water and scale, not lawn under full sun.

All 145 OSM-mapped Dubai parks: Park Cool Island vs green share — The city-wide census: green share does not predict cooling (R² = 0.04). Al Safa 2 starred: 85% green, +0.47 °C.

Tier 1 — desk analysis: free data, open models, no site visit

Satellite surface temperature, ten years of hourly UTCI, 1-metre radiant-heat maps, an hourly shadow model of every building and tree crown, a shade-seeking pedestrian simulation and a tree asset ledger, chained into one reproducible pipeline that runs for any park on Earth from a laptop. The city census above is this tier working as a triage instrument: screen every park, then send the deep analysis and the retrofit budget where they matter.

Toolkit pipeline diagram — The pipeline: open data in, peer-reviewed models in the middle, decisions and a public interface out.

SOLWEIG 1 m radiant temperature maps — Radiant heat at 1 m: open ground 72 °C vs 49 °C under canopy. Shade is worth 23 °C.

Hours in shade per square metre — Hours of shade per square metre. The white areas are where new shade should go.

Shade-seeking pedestrian simulation — Simulated shade-seeking pedestrians: detours cannot dodge the heat, so shade must go on the direct desire lines.

Walkshed before and after wall openings — Eight feasible wall openings grow the 5-minute walkshed by 24%.

Tier 2 — field validation: one camera, one evening

Models that never meet a measurement are easy to believe and easy to get wrong. A handheld thermal camera (≈US$550) closed the loop: 18 calibrated thermograms across the park, georeferenced and compared against the model. The least-shaded modelled locations are the hottest measured ones. One camera serves every park a city owns.

Model prediction vs field thermography — Predicted shade vs measured surface temperature: sunlit pavement 45.9 °C, canopy zone 35.4 °C.

Tree-gochi — the trees ask for help

The analysis found that one tree cluster provides 59% of the park's shade. Tree-gochi exists to keep those trees alive: a US$25 open-hardware device whose pixel Ghaf tree visibly wilts as soil moisture drains and asks, in words, to be watered. Alert thresholds come from the project's own UTCI analysis. No auto-refill: the tree's fate is the community's responsibility.

Tree-gochi thirsty state — Soil water 14%: canopy dropped, asking for water.

Tree-gochi recovering state — After watering: recovering, "very hot, holding on".

Live demo: physical board, browser digital twin, and the 3D-viewer link working together.

Explore the model yourself

The interactive 3D explorer below is a direct output of the toolkit: 1,336 buildings and 214 tree-canopy patches, with an hourly sun slider and toggles for hours-in-shade and UTCI overlays, all precomputed by the same pipeline.

Open the 3D shade explorer