UDA-City Heat Risk Diagnostic
Where is heat most dangerous to people — and what can we do about it?
(Kampong Lama)
under warming
The City
UDA-city is a synthetic tropical metropolis modelled on Colombo, Sri Lanka (6.93°N, 79.86°E) — hot-humid, rapidly urbanising, with stark inequality between neighbourhoods. Its 10 districts fall into three archetypes: 3 Refuge zones (green, sparse, paradoxically hottest), 4 Hotspot districts (dense informal settlements, highest risk), and 3 Core areas (high-rise CBD, coolest due to roughness-driven ventilation). This is a lower-income city where 300–400 people per hectare live without air conditioning in neighbourhoods where metal roofs amplify outdoor heat indoors.
The hottest neighbourhood is not the highest-risk one. Jade Gardens peaks at 38.1°C but ranks 7th on risk. Dhobi Lines peaks lower (36.9°C) but ranks 1st — because 300 people/ha live there with 10% AC access and 60% outdoor workers.
Heat Hazard: What SUEWS Shows
SUEWS v2026.6.5 was run on all 10 neighbourhoods of UDA-city under shared hot-humid forcing (ERA5 2024, April–May window). Hazard metric: hours where modelled 2m air temperature exceeds 35°C.
| Neighbourhood | Type | Peak T2 | Hours Now | Hours +2.5°C | Change |
|---|---|---|---|---|---|
| Jade Gardens | Refuge | 38.1°C | 75 | 513 | +438 |
| Kampong Lama | Hotspot | 37.4°C | 68 | 507 | +439 |
| Taman Melati | Refuge | 37.7°C | 63 | 467 | +404 |
| Dhobi Lines | Hotspot | 36.9°C | 44 | 424 | +380 |
| Fuzhou Lanes | Hotspot | 36.7°C | 36 | 395 | +359 |
| Serendib Rise | Refuge | 36.8°C | 31 | 374 | +343 |
| Mlima Moto | Hotspot | 35.6°C | 5 | 277 | +272 |
| Lusitano Square | Core | 35.3°C | 4 | 237 | +233 |
| Victoria Exchange | Core | 35.1°C | 3 | 218 | +215 |
| Zheng He Towers | Core | 34.4°C | 0 | 132 | +132 |
The "refuge" neighbourhoods are paradoxically the hottest — their low building density means weak turbulent mixing, so near-surface air warms more under strong radiation. The dense high-rise cores are coolest: tall buildings create roughness that drives convective cooling.
Why the Refuges Are Hottest: The Model Physics
This result is counter-intuitive — greener, lower-density neighbourhoods should be cooler. But SUEWS reveals the mechanism:
Low roughness traps heat near the surface
Jade Gardens has λp = 0.047 and mean building height 5.2m. The Macdonald parameterisation translates this into a small aerodynamic roughness length (z0). Weak z0 means weak turbulent mixing — sensible heat flux (QH) cannot efficiently transport heat away from the surface into the boundary layer. The heat stays where people breathe.
Paved surfaces dominate storage heat
At 59% paved fraction, OHM (Objective Hysteresis Model) storage heat flux is large. Impervious surfaces absorb shortwave radiation during the day and re-radiate longwave at night — suppressing nocturnal cooling. The OHM coefficients for paved surfaces (a1=0.72, a2=0.30) give a strong hysteresis.
Tall buildings ventilate the cores
Zheng He Towers (FAI = 0.39, height 25.5m) and Mlima Moto (FAI = 0.90, height 10.2m) generate large z0 — enhancing turbulent transport. Tall canyon geometry also provides self-shading, reducing the net radiation load at street level. The building fraction (34–44%) displaces paved area that would otherwise dominate OHM storage.
What this means for policy
Urban cooling interventions should not target temperature alone. The refuges are hot but sparsely populated. The dense informal hotspots are moderately hot and packed with vulnerable people who cannot escape the heat. Roughness-enhancing interventions (street trees at scale, vertical structures) would improve ventilation in the low-rise hotspots where it matters most.
From Hazard to Risk: The Bridge
Following the UNDRR framing: Risk = f(Hazard × Exposure × Vulnerability)
| Neighbourhood | Type | Hazard | Exposure | Vulnerability | Risk | Rating |
|---|---|---|---|---|---|---|
| Dhobi Lines | Hotspot | 0.59 | 1.00 | 0.92 | 1.00 | 🔴 Red |
| Fuzhou Lanes | Hotspot | 0.48 | 1.00 | 0.97 | 0.96 | 🔴 Red |
| Serendib Rise | Refuge | 0.41 | 0.77 | 0.27 | 0.53 | 🟠 Amber |
| Mlima Moto | Hotspot | 0.07 | 1.00 | 1.00 | 0.49 | 🟠 Amber |
| Lusitano Square | Core | 0.06 | 0.77 | 0.09 | 0.17 | 🟢 Green |
| Jade Gardens | Refuge | 1.00 | 0.00 | 0.32 | 0.14 | 🟢 Green |
| Zheng He Towers | Core | 0.00 | 0.77 | 0.00 | 0.00 | ⬜ Grey |
The CrunchTest: +2.5°C
Under warming, dangerous-heat hours increase 5–7× everywhere. The critical trajectory signal: Mlima Moto moves from Amber to Red — the only neighbourhood that changes RAGG band. Under +2.5°C, Kampong Lama spends 35% of April–May above 35°C for a population with 8% AC access.
What-If Scenarios: What Can We Do?
We tested three interventions the model can simulate. Each changes a physical parameter in the four hotspot neighbourhoods (Kampong Lama, Dhobi Lines, Mlima Moto, Fuzhou Lanes) and re-runs SUEWS under present-day forcing.
→ See the interactive warming slider: At what warming does your neighbourhood break?
Scenario A: Plant Trees (+12% canopy)
What we changed: Evergreen tree fraction increased from 3% to 15% in hotspot grids, reducing paved fraction by the same amount.
Result: Counter-intuitively, dangerous hours increased slightly in some hotspots. The added canopy at only 12% is insufficient to offset the reduction in paved surface roughness and the change in OHM storage dynamics. SUEWS shows that simply planting small trees without achieving closed canopy is not a guaranteed cooling intervention.
Confidence: 🟠 Moderate — result is physically plausible (low canopy without roughness benefit) but sensitive to LAI and tree height assumptions.
Scenario B: Cool Roofs (albedo 0.2 → 0.5)
What we changed: Surface albedo for paved and building surfaces raised from 0.2 to 0.5 in hotspot grids.
Result: Dramatic reduction — 88–100% fewer dangerous hours in hotspot neighbourhoods. Kampong Lama drops from 68h to 8h; Dhobi Lines from 44h to 4h.
Confidence: 🟢 High for the physics mechanism (reduced net radiation directly reduces surface heating); 🟠 Moderate for real-world magnitude (bulk albedo doesn't capture canyon geometry or reflected radiation onto pedestrians).
Scenario C: Combined (Trees + Cool Roofs)
What we changed: Both interventions applied simultaneously.
Result: 77–88% reduction. Slightly less effective than cool roofs alone because the added trees shade surfaces that would otherwise reflect more radiation at higher albedo.
Confidence: 🟢 High for physics; 🔴 Low for real-world implementation feasibility at this scale.
| Neighbourhood | Type | Baseline (h) | Trees (h) | Cool Roofs (h) | Combined (h) | Best Reduction |
|---|---|---|---|---|---|---|
| Kampong Lama | Hotspot | 68 | 69 | 8 | 16 | −88% (roofs) |
| Dhobi Lines | Hotspot | 44 | 48 | 4 | 5 | −91% (roofs) |
| Mlima Moto | Hotspot | 5 | 16 | 0 | 0 | −100% (roofs) |
| Fuzhou Lanes | Hotspot | 36 | 42 | 4 | 5 | −90% (roofs) |
Key finding: Cool roofs are the dominant intervention in this model. But even with 88% reduction in present-day dangerous hours, under +2.5°C warming Kampong Lama would still face hundreds of dangerous hours. Physical interventions buy time; they don't solve the problem alone.
Recommendations
Based on the model evidence, we recommend five prioritised actions:
- Cool-roof programmes targeting metal-roofed informal housing — model shows 88–91% reduction in dangerous hours. Real-world co-benefit: reduces indoor heat (not modelled but literature-supported). Prioritise Kampong Lama and Dhobi Lines.
- Street-tree planting at scale, not token canopy — model shows 12% canopy is insufficient. Literature suggests 30%+ closed canopy with tall trees (roughness benefit) is needed. Design for ventilation corridors, not scattered plantings.
- Heat early-warning system keyed to outdoor-worker hours — model output directly supports timing (which hours exceed 35°C threshold). 60% of Dhobi Lines residents work outdoors.
- Night-time cooling assessment and shelter access — model shows whether T2 drops below 28°C overnight (recovery threshold). Where it doesn't, provide cooling centres or subsidised night ventilation.
- Do NOT rely on temperature alone for policy — the risk bridge demonstrates that hazard ranking ≠ risk ranking. Policy must combine physical data with exposure and vulnerability data.
Where Physics Ends and Governance Begins
The model shows where and when it's dangerous. It shows what land-cover changes reduce the hazard. But the gap between "the air is 37°C" and "this person is in danger" is filled by governance, infrastructure, and individual capacity — none of which the model holds.
- Indoor heat amplification — metal roofs in informal housing push indoor temperatures 5–10°C above outdoor T2. The model doesn't see indoors.
- Human behaviour — do people stay indoors? Do they hydrate? Do outdoor workers have shade breaks? These are governance decisions.
- Work-hour restrictions — model output can identify "do not work outdoors" windows, but enforcement is policy.
- AC provision and waste-heat feedback — as AC adoption grows, waste heat warms outdoor air (QF not enabled). A vicious cycle the model could capture but doesn't in this configuration.
- Emergency cooling centres — no physical intervention eliminates all dangerous hours under +2.5°C. Residual risk must be managed by shelter.
- Real demographics — socioeconomic data is synthetic. Real census data would change absolute risk scores (but likely not the ranking).
The model is a compass pointing at the problem. The response is governance, infrastructure, and political will. That bridge — from physics to policy — is where lives are saved or lost.
Where the Bridge Holds
- SUEWS correctly differentiates neighbourhoods by land cover and building roughness
- The ranking is robust to threshold choice (33°C, 35°C, or 37°C)
- The UNDRR decomposition reveals that temperature maps alone are insufficient for risk governance
What We Would Change
Given more time or a real-world application, these model improvements would strengthen the assessment:
| Enhancement | Why It Matters | SUEWS Support |
|---|---|---|
| Enable QF (anthropogenic heat) | AC waste heat creates a feedback loop: hotter → more AC → more outdoor heat. As warming drives AC adoption in hotspot neighbourhoods, this becomes significant. | Emissions method available; off by design for comparability |
| SPARTACUS radiation | Multi-layer scheme handles canyon reflections, sky-view factor, wall interactions. Would better differentiate dense low-rise from dense high-rise. | Built into SUEWS; not enabled in this config |
| Humid-heat metric (WBGT) | At 81% RH, 35°C dry-bulb is far more dangerous than in a dry climate. A wet-bulb globe temperature would be more physiologically relevant. | Could be post-processed from T2 + forcing RH + radiation |
| Dynamic OHM | Classic OHM uses fixed coefficients. Dynamic OHM adapts to surface moisture — important in a humid city with frequent rainfall. | Available as physics option |
| Deciduous trees | Current config has 0% deciduous everywhere. Tropical semi-deciduous trees lose LAI in dry/hot season → reduced cooling when most needed. | Supported; would need phenology calibration |
Citations
Järvi, L., Grimmond, C.S.B. & Christen, A. (2011). The Surface Urban Energy and Water Balance Scheme (SUEWS): Evaluation in Los Angeles and Vancouver. J. Hydrol., 411(3–4), 219–237. DOI
Ward, H.C., Kotthaus, S., Järvi, L. & Grimmond, C.S.B. (2016). SUEWS: Development and evaluation at two UK sites. Urban Climate, 18, 1–32. DOI
UNDRR (2025). Extreme Heat Risk Governance Framework and Toolkit. United Nations Office for Disaster Risk Reduction.