HeatGuard

Inspiration

Singapore is getting warmer, but heat risk is not equal everywhere. Some neighbourhoods may face high heat while also having many elderly residents and limited access to nearby cooling or community support.

This inspired HeatGuard to focus not only on where it is hottest, but on where communities are least equipped to cope with the heat.

What it does

HeatGuard combines real-time weather conditions, elderly population data and existing cooling/support infrastructure to calculate a Heat-Resilience Gap Score for each area.

The platform highlights Low, Medium and High-risk neighbourhoods, explains the key drivers behind each score, and suggests actions such as senior check-ins, heat advisories, extended community-centre hours or temporary cooling support.

How we built it

HeatGuard is designed as an end-to-end Databricks pipeline:

NEA / HDB / MOH / NParks → Lakeflow → Delta Bronze/Silver/Gold → Heat-Resilience Gap Score → AI/BI Dashboard + Genie → Databricks App

We use:

• NEA temperature and humidity data for live heat stress
• HDB elderly population data for vulnerability
• Active Ageing Centre locations and NParks green spaces for cooling/support capacity

Challenges we ran into

The biggest challenge is combining datasets with very different formats, geographic scales and update frequencies. Live weather readings are station-based, while population and infrastructure data are area-based and updated less frequently.

Another challenge is designing a risk score that is useful while remaining transparent and explainable.

Accomplishments that we're proud of

We developed a clear framework that turns separate environmental, demographic and infrastructure datasets into a single heat-resilience gap concept.

Instead of producing another heat map, HeatGuard connects risk directly to whether vulnerable communities have enough support to cope.

What we learned

We learned that heat risk is more complex than temperature alone. Vulnerability depends on who is exposed and what resources already exist around them.

We also learned the importance of designing an explainable model so users can understand why an area is classified as high risk.

What's next for HeatGuard

Next, we aim to build the full live dashboard, validate the scoring methodology and add before-and-after intervention simulation.

Future extensions could include next-day heat-risk forecasting, finer-grained green-space analysis and measuring whether local interventions reduce the Heat-Resilience Gap Score over time.

Built With

• data.gov.sg
• databrick