Final SUEWS Community Hackathon Submission

62present dangerous-heat hours in hottest neighbourhood, Jade Gardens

260future dangerous-heat hours in hottest neighbourhood, Jade Gardens

#1risk rank for Kampong Lama in both scenarios

10UDA-city neighbourhoods modelled

Core Answer

The hottest neighbourhood is not the highest-risk neighbourhood. Jade Gardens is hottest by dry-bulb dangerous-heat hours now and under +2.5 C, but Kampong Lama is where heat is most dangerous to people in this socio-economic risk framing because substantial heat hazard coincides with high exposure and high vulnerability.

Question

Across the ten UDA-city neighbourhoods, where is heat most dangerous to people now and under the official +2.5 C hotter-future stress test, and is that the same as where it is simply hottest?

UDA-city SUEWS Configuration And Assumptions

The workflow uses uda-city.yml, the official ten-site synthetic hot-humid city config from UMEP-dev/uda-city-hackathon. All neighbourhoods share the same forcing; differences come from morphology, land cover, and socio-economic sidecars.

Locked physics: NARP net radiation, classic OHM storage heat, QF/emissions off, and SUEWS/SuPy runtime 2026.6.5. The required assess_readiness and validate_config SUEWS-agent calls passed before SUEWS was run.

Official Present And +2.5 C Future Method

The present scenario uses forcing/present_hot_humid/UDA_2024_data_60.txt. The future scenario uses forcing/future_hot_humid/UDA_2024_data_60.txt, the official humidity-preserving +2.5 K pseudo-warming: air temperature is raised uniformly, relative humidity is held constant, longwave down is scaled for a warmer atmosphere, and wind, pressure, shortwave, and rain are unchanged. This is a stress test, not a downscaled climate projection.

Hazard, Exposure, Vulnerability, And Risk Indicator

Hazard

The modelled heat condition from SUEWS. The core hazard is dangerous-heat hours: post-spin-up hours where hourly mean 2 m air temperature, T2, exceeds 35 C in each neighbourhood.

Exposure

Who or what is present in the affected neighbourhoods. Here it is daytime population density from the UDA-city sidecar, scaled across the ten neighbourhoods.

Vulnerability

Why some exposed people may be more at risk. The index combines older age, young children, low air-conditioning access, outdoor work, and deprivation from the synthetic socio-economic sidecar.

Combination rule

Each pillar is min-max scaled to 0-1 and combined as risk_index = (hazard * exposure * vulnerability)^(1/3), then scaled again to 0-1 for ranking. The geometric mean is conservative: weak exposure or vulnerability lowers risk even where hazard is high.

Interactive Risk Explorer

If you change what counts as risk, who needs help first?

Move the levers to compare heat, who is exposed, and who is more vulnerable. The key finding is visible before any numbers: the hottest neighbourhood is not automatically the most dangerous to people.

Core insight: Jade Gardens is the hottest place in the dry-bulb SUEWS results, but Kampong Lama becomes the top people-risk priority once exposure and vulnerability are included.

Quick definitions in everyday language

Dry-bulb T2: The regular 2 m air temperature from SUEWS, like a shaded thermometer reading.
WBGT screen: A heat-stress screening estimate that adds humidity, sun, and wind. It is useful for worker-safety thinking, but it is not a field-measured WBGT instrument reading.
Exposure: How many people are present in the neighbourhood during the day.
Vulnerability: Why exposed people may be less able to cope, such as older age, young children, low air-conditioning access, outdoor work, or deprivation.

ScenarioSwitch between today's simulated heat and the official hotter stress test.

Heat signalChoose what counts as the heat hazard. WBGT adds humidity, sun, and wind to the story.

Heat weight: 1Higher means the model prioritises the physical heat signal more.

People-present weight: 1Higher means crowded daytime neighbourhoods count more.

Coping-difficulty weight: 1Higher means age, low AC access, outdoor work, and deprivation count more.

People-risk priority ranking

Longer bars mean higher relative risk under the current lever settings. Type labels are shown beside colour so the chart is not colour-only.

hotspotcorerefuge

Hottest versus most dangerous

Each dot is a neighbourhood. Right means a stronger heat signal; up means a higher people-risk score.

Time-of-day robustness check

The cards summarise when heat stress appears by neighbourhood type. This keeps the worker-safety and night-recovery story visible without making the main explorer a spreadsheet.

Show the numbers

Technical table for the current lever settings. Risk score is the weighted geometric mean of min-max scaled hazard, exposure, and vulnerability, then rescaled from 0 to 1.
Priority	Neighbourhood	Type	Hazard signal	Scaled hazard	Exposure	Vulnerability	Risk score

Detailed time-of-day heat-stress summary for the selected scenario.
Type	Time of day	Dry-bulb danger hours	WBGT screen hours	Hot sunny low-wind hours	Max T2	Max WBGT proxy

WBGT, HHI, And Time Of Day

WBGT

Wet Bulb Globe Temperature is the stronger occupational heat-stress standard for outdoor workers because it accounts for temperature, humidity, radiant heat, and wind. This page adds a humidity-aware WBGT screening proxy from SUEWS T2, RH2, U10, and Kdown, but it is not a full measured WBGT. A full WBGT assessment would use a field monitor or a validated globe-temperature model, consistent with OSHA guidance.

HHI

The CDC Heat & Health Tracker points toward the right public-health framing: combine heat, population health, environmental conditions, and social vulnerability. A literal Heat and Health Index was not computed because UDA-city is synthetic and not a ZIP-code health dataset. Instead, this analysis uses the HHI logic to make exposure and vulnerability explicit in the socio-economic risk bridge.

Time of day

Present dry-bulb dangerous hours are concentrated in morning and afternoon. Under +2.5 C, daytime danger intensifies, and the WBGT screening proxy indicates reduced evening and night recovery, especially in hotspot neighbourhoods. That strengthens recommendations for work-rest timing, overnight cooling, and evening outreach.

Defended Choices And Caveats

I use hourly mean T2 > 35 C after 14 spin-up days for the core hazard because it is simple, reproducible, and consistent with the UDA-city reference bridge. The WBGT screen and time-of-day tables are robustness checks from the same completed SUEWS outputs, not a replacement for the core threshold. SUEWS gives environmental hazard, not health outcomes. The socio-economic layer is synthetic, min-max scaling is relative to this dataset, neighbourhood averages hide individuals, and QF is off so population affects exposure rather than modelled anthropogenic heat.

Heat Hazard And Risk Result

Plain-English summary of the core dry-bulb hazard and people-risk result. The risk rank combines hazard, exposure, and vulnerability; it is not the same as hottest rank.
Neighbourhood	Type	Present hazard hours	Future hazard hours	Added hours	Risk priority now	Risk priority future	Plain-English read
Jade Gardens	refuge	62	260	198	7	7	Physically hottest, but low people-risk in this relative dataset because exposure is low.
Serendib Rise	refuge	26	205	179	7	7	Substantial future warming, but low priority in this risk index because exposure is low.
Taman Melati	refuge	47	243	196	7	7	Hot refuge area; hazard is high, but the hazard-to-people bridge is weak here.
Kampong Lama	hotspot	42	249	207	1	1	Highest people-risk priority now and in the future.
Dhobi Lines	hotspot	26	217	191	2	3	High exposure and vulnerability turn moderate-to-high hazard into high risk.
Lusitano Square	core	5	129	124	5	5	Moderate exposure but lower vulnerability keeps risk below the hotspot group.
Mlima Moto	hotspot	5	149	144	4	4	Risk grows strongly when future hazard rises over a very vulnerable exposed population.
Victoria Exchange	core	5	120	115	6	6	Core area with lower vulnerability, so it is not a top risk priority in this framing.
Fuzhou Lanes	hotspot	22	212	190	3	2	Becomes the second-highest future risk priority because vulnerability is very high.
Zheng He Towers	core	2	77	75	7	7	Lowest dry-bulb hazard and lowest vulnerability score in this dataset.

hazard_risk_summary.csv risk_present.csv risk_future.csv heat_stress_screen_summary.csv heat_stress_time_of_day_by_type.csv heat_stress_time_of_day.csv

Where The Link Holds And Breaks

Pattern	Evidence	Meaning for decisions
The link holds in hotspot settlements.	Kampong Lama, Fuzhou Lanes, Dhobi Lines, and Mlima Moto combine high exposure, high vulnerability, and increasing future hazard.	People-protection measures should not wait for these areas to be the absolute hottest. They are already high-risk because people and vulnerability are concentrated there.
The link breaks in refuge neighbourhoods.	Jade Gardens is the hottest by dry-bulb dangerous hours in both scenarios, but its exposure score is the minimum in this relative dataset.	Hottest-place maps alone would over-prioritise some physical hot spots and under-prioritise exposed/vulnerable communities.
Future warming strengthens the bridge.	All neighbourhoods gain dangerous-heat hours, but hotspot risks remain highest when those hours meet high exposure and vulnerability.	Adaptation should combine citywide heat reduction with targeted heat-health operations in high-risk neighbourhoods.
Time of day adds a second warning.	Future WBGT screening hours remain high in evening and night even when dry-bulb hours above 35 C are daytime only.	Recommendations must include overnight recovery, not only afternoon shade.
Uncertainty remains in the people layer.	The socio-economic sidecar is synthetic, and AC access, outdoor work, deprivation, and age are represented as neighbourhood averages.	Before implementation, local agencies should replace synthetic vulnerability values with household, labour, health, and infrastructure data.

Policy Recommendations Across Governance Scales

Scale of governance	Actionable recommendation	Why this follows from the findings
Neighbourhood and community groups	In Kampong Lama, Fuzhou Lanes, Dhobi Lines, and Mlima Moto, organise door-to-door heat checks, map residents who need assistance, open shaded water points before the morning heat builds, and keep evening check-ins active when nights stay humid.	The highest-risk areas are not simply the hottest; they are places where exposed and vulnerable people are present during dangerous heat.
Municipal heat-health teams	Use the risk ranking to deploy cooling centres, mobile outreach, transport to cool spaces, public messaging, and alert escalation. Use the hazard map separately for citywide heat-reduction investments.	Hazard and risk answer different questions. The city needs both physical cooling and people-targeted operations.
Labour and occupational safety agencies	Require work-rest-shade-water plans for outdoor work, add on-site WBGT monitoring during alerts, shift heavy work away from the morning-to-afternoon hazard window, and include acclimatisation procedures for new or returning workers.	The WBGT screen shows that humidity, sun, wind, and recovery time matter for workers, not only air temperature.
Urban planning, housing, and utilities	Prioritise cool roofs, shaded pedestrian corridors, tree or shade structures where water is feasible, ventilation paths, reflective surfaces, and reliable electricity for high-risk hotspot areas.	Physical heat exposure and low adaptive capacity overlap most strongly in hotspot neighbourhoods.
Health and social protection systems	Pre-register older adults, households with low AC access, outdoor workers, and medically vulnerable residents for heat-wave calls, clinic triage, medication advice, and emergency cooling support.	The vulnerability pillar explains why the same hazard can produce very different human risk.
Regional and national government	Fund heat-resilient settlement upgrading, local sensor networks, occupational heat standards, heat-health surveillance, and social protection payments during extended heat events.	Local teams can target interventions, but they need finance, standards, and data systems beyond neighbourhood control.

SUEWS-agent Tool-call Log

The required tool-call log is included in the repository at transcripts/suews_agent_tool_log.md. The first two calls were assess_readiness and validate_config; both passed before SUEWS was run. The WBGT and time-of-day additions were post-processing checks from the completed SUEWS outputs and did not rerun SUEWS.

Tool	Why it was called	Short result
`assess_readiness`	To confirm the manifest, official UDA-city configuration, forcing files, socio-economic sidecars, and runtime before modelling.	Passed. The agent found UDA-city, ten neighbourhoods, and SuPy 2026.6.5.
`validate_config`	To validate the canonical configuration and locked physics before allowing any SUEWS run.	Passed. The configuration validated for ten grid cells with NARP radiation, classic OHM, and QF off.
`run_suews_present`	To run the official present hot-humid SUEWS scenario for all ten neighbourhoods.	Completed for ten sites and 26,208 hourly steps per site. Results were saved as present SUEWS outputs.
`run_suews_future`	To run the official +2.5 C humidity-preserving future stress test for all ten neighbourhoods.	Completed for ten sites and 26,208 hourly steps per site. Results were saved as future SUEWS outputs.
`apply_risk_bridge_present`	To translate present dangerous-heat hours into hazard, exposure, vulnerability, and risk ranks.	The present bridge identified Kampong Lama as the highest-risk neighbourhood.
`apply_risk_bridge_future`	To translate future dangerous-heat hours into hazard, exposure, vulnerability, and risk ranks.	The future bridge again identified Kampong Lama as the highest-risk neighbourhood.
`combine_hazard_risk_results`	To compare the hottest neighbourhoods with the highest-risk neighbourhoods across both scenarios.	The combined table showed that hottest and highest-risk are not the same; Spearman rank correlation was weak.

SUEWS Citation And Version Information

Model/runtime used: supy 2026.6.5. Cite SUEWS following the official guidance, including Jarvi, Grimmond & Christen (2011) and Ward, Kotthaus, Jarvi & Grimmond (2016). The SUEWS-style colour palette and logo concept are credited to the SUEWS project.