I Never Thought I'd Build Something Around Sustainability: A Heat and Wind Priority Map for Mumbai
I grew up reading about sustainability in textbooks. For the WeMakeDevs x AWS hackathon, I built a tool that ranks Mumbai's 24 wards by heat and ventilation, hosted on AWS Amplify.
Final Year, Bachelors in Physics student.
I never thought I'd build something around sustainability. I had only read about it in books and written about it for academics, so building a real project on it felt out of the box. This environment-focused hackathon pushed me there, and I'm so happy it did.
The problem -
Mumbai is dense, hot and humid. Some wards are hotter than others, and some get far less airflow than others. The worst case is a ward that is both: heat gets trapped, and no breeze moves through to clear it.
Heat maps exist, and wind data exists, but they live in separate places. Planners have no simple way to see the combination. So I asked one question: where should Mumbai plan wind corridors and open space first?
What I built -
Wind Corridor Planner scores all 24 BMC wards on heat and ventilation, then ranks them by priority.
- Priority layer (default): wards that are both hot and poorly ventilated
- Heat (LST) layer: land surface temperature on its own
- Ventilation layer: how well each ward can breathe
- Season switcher: winter, summer and monsoon wind patterns
- Ward popups: localities, land surface temperature and priority rank
- Top 5 table: a ready list for planners
The data -
Everything comes from public sources:
- Wind direction: Open-Meteo
- Built-up density: Sentinel-2 satellite imagery
- Land surface temperature: Landsat 8 and 9, Collection 2 Level-2, through Microsoft's Planetary Computer
For temperature, I used March to May images from 2021 to 2025 and kept only scenes with under 10% cloud cover. From 61 candidate scenes I kept the best 8, masked out cloud, shadow and water pixels, and took the per-pixel median on a 30 m grid. Then I averaged the result for each ward.
One practical lesson: satellite scenes are huge, and my laptop has little RAM and a slow connection. Reading only the small window of pixels each ward needs, instead of downloading whole scenes, made the pipeline possible.
The priority score -
Each ward gets a heat score from 0 (coolest) to 1 (hottest). Priority combines it with the ventilation score:
plaintext
priority = 0.5 × heat_score + 0.5 × (1 − ventilation_score)
A ward ranks high when it is hot and poorly ventilated. The 50/50 weighting is a design choice, not a scientific constant.
What the data shows
Summer land surface temperature ranged from about 36.1 °C in R/C to 41.5 °C in L. The top 5 priority wards:
Rank | Ward | LST (°C) | Priority score |
|---|---|---|---|
1 | B | 41.3 | 0.898 |
2 | L | 41.5 | 0.889 |
3 | C | 40.2 | 0.855 |
4 | E | 39.3 | 0.748 |
5 | K/E | 40.5 | 0.739 |
Ward C shows why combining both layers matters. It is not the hottest ward, but its ventilation score is only 4.4, so it still ranks third. Looking at heat alone would have understated it.
Where AWS fits -
The app is hosted on AWS Amplify. I connected my GitHub repository to Amplify, so every push to
main triggers a new build and deploys the live site automatically. That made iterating fast: I could fix ward locality names, push, and watch the build go green in a few minutes, with no server to manage.Limitations -
- This is land surface temperature, not air temperature. It shows how hot the ground and roofs are, not what people feel on the street.
- The 50/50 weighting is a judgment call.
- Building height is not included yet, and tall buildings change airflow a lot.
- Locality names are indicative. Ward boundaries follow BMC's 24-ward GeoJSON.
- The data is a March to May snapshot, not the whole year.
What's next -
- Add building height data to improve the ventilation score.
- Add air temperature from weather stations.
- Let planners adjust the heat and ventilation weights in the app.
- Extend the same pipeline to other cities.
Links -
- Demo video: https://youtube.com/shorts/BPHI9MOtOWg
Built for the WeMakeDevs x AWS Hackathon.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article