| Scenario | Days >30°C | Days >35°C | Days >40°C | Days >45°C |
|---|---|---|---|---|
| Historical 1995-2014 | 92 | 88 | 33 | 0 |
| SSP2-4.5 · MRI-AGCM3-2-S | 238(Δ +146) | 182(Δ +94) | 93(Δ +60) | 9(Δ +9) |
| SSP5-8.5 · MPI-ESM1-2-XR | 248(Δ +156) | 176(Δ +88) | 80(Δ +47) | 2(Δ +2) |
Satellite & climate datasets
Each dataset below contributes a distinct role. The atlas intentionally does not blend them into a single composite number; users can cross-reference the table against any specific plate to see which sensor or product is upstream.
| Dataset | Purpose | Resolution | Notes |
|---|---|---|---|
| NASA POWER | Daily T2M, T2M_MAX, T2M_MIN, TS, RH2M at each centroid | ~0.5° (~50 km) | Drives every per-emirate scalar, choropleth, time series, and Recent Extremes table in this MVP. |
| GloUTCI-M | Universal Thermal Climate Index (UTCI), monthly, JJA climatology over the UAE bbox | 1 km native | Yang et al. 2023, Zenodo record 8310513. Renders as the map heatmap when public/data/utci.json is present. Does not feed any tabular metric — the tables stay on POWER. |
| Open-Meteo Climate API | CMIP6 SSP2-4.5 and SSP5-8.5 daily T2M_MAX, 2021–2050 | ~25 km bias-corrected | Feeds the SSP projection rows in Plate 06. One model per scenario (MRI-AGCM3-2-S, MPI-ESM1-2-XR). |
| Natural Earth | Cartographic boundaries | Generalised 10 m admin-1 | Visualisation only — not authoritative cadastral |
| FCSA 2022 | Emirate land areas (km²) for area-weighting | Administrative | Used only to weight the national T2M series. Not a measurement. |
Every metric rendered in the dashboard traces back to one of the rows above. There are no literature seeds or placeholder scalars in app/data.ts: when the bake at app/_data/atlas.json is empty, the UI renders an em-dash and the choropleth fades to neutral grey. Surface-temperature claims at the district/30 m scale require a raster pipeline (Landsat C2 L2 Surface Temperature, ECOSTRESS, MODIS MOD11A2) that is not wired up in this MVP and so no such claim is surfaced.
How surface heat is measured
The US EPA recognises three complementary techniques for characterising urban heat islands. Direct measurement uses fixed weather stations and vehicle-mounted mobile traverses to record near-surface air temperature — the ground truth, but with sparse spatial coverage. Remote sensing uses satellite thermal-infrared bands to measure radiometric land-surface temperature (LST) across the full urban footprint. Modelling couples observations with urban canopy schemes to interpolate, hindcast, and project under future scenarios.
This MVP leans on the direct measurement leg — but with the recognised caveat that POWER is a 0.5° gridded reanalysis, not a station network. NASA POWER supplies daily T2M (mean), T2M_MAX, T2M_MIN, TS (surface skin) and RH2M. The atlas pulls POWER at every cell of a tight UAE bounding box on its native 0.5° × 0.5° (~50 km) grid for the map heatmap, and at each emirate centroid for the tables and trend lines. Because POWER’s ~50 km cell is the floor, the dashboard does not render any 30 m or district-level claim. The remote-sensing and modelling legs — Landsat 9, Sentinel-2, MODIS MOD11A2, ECOSTRESS — are listed as roadmap items, not active sources, until a raster pipeline is wired.
WBGT (simplified). The 5th Atlas toggle is the Australian BOM apparent-temperature formula applied to POWER’s T2M_MAX and RH2M — a station-data heat-stress proxy, not a true ISO 7243 WBGT.
Heat-day thresholds (30, 35, 40, 45 °C) come in two shapes. The Historical row is computed directly from POWER’s daily T2M_MAX over 1995–2014 at each of the 7 emirate centroids, summed per JJA season and averaged across (emirate × year). The SSP rows come from CMIP6 daily T2M_MAX served by Open-Meteo’s Climate API (2021–2050), one downscaled model per scenario (MRI-AGCM3-2-S for SSP2-4.5, MPI-ESM1-2-XR for SSP5-8.5). Both are reduced through the same threshold-count code path so the Δ column is apples-to-apples.
Administrative boundaries are Natural Earth 10 m admin-1, simplified to 0.003°. Natural Earth boundaries are generalised de-facto cartography for public visualisation — they are not authoritative cadastral boundaries, and they are not used for any area-weighted analytic claim that requires official extents.
What this atlas does — and does not — claim
Every metric on this dashboard is reproducible from the two APIs listed above. The trade-off is scope:
- Air, not surface. POWER returns 2 m air temperature and a coarse skin-temperature estimate (TS). It is not a satellite LST retrieval. Anything you’d normally call “land-surface temperature at 30 m” is out of scope for this MVP and not rendered.
- Two heatmap resolutions, one resolution floor for the tables. The map renders the GloUTCI-M JJA UTCI climatology at 1 km when
public/data/utci.jsonis present, and falls back to a dense POWER pull on its native 0.5° (~50 km) grid otherwise. Every tabular metric (per-emirate scalars, trends, days- above counts, SSP rows) still rolls up from POWER at the 0.5° floor; nothing on this dashboard surfaces a district-, neighbourhood- or parcel-level claim. A true 30 m surface-temperature pipeline (Landsat C2 L2 ST, MODIS MOD11A2) is v2 work. - No SUHI. Computing a defensible urban−rural difference at point scale would need at least MODIS MOD11A2 (1 km) plus an urban/rural land-cover mask. Neither is wired up, so neither is rendered — this MVP carries no SUHI claim.
- Single CMIP6 model per SSP. The SSP rows use one downscaled model per scenario, not the multi-model median. Treat them as one bracketing line rather than the consensus projection.
- Boundaries. Natural Earth admin-1 is a generalised de-facto layer suitable for visualisation only. Area weights on the national series come from FCSA, not from the GeoJSON.
- Roadmap. A v2 with district-level LST would integrate Landsat C2 L2 ST via Planetary Computer STAC or USGS M2M, Overture Maps for the built-up mask, and ECOSTRESS for 70 m hotspot snapshots.