Remote sensing · Machine learning · Urban planning
Urban Heat Island
Mitigation
A decision-support pipeline that turns Landsat 8 thermal imagery into a ranked, costed plan for where to plant, shade and re-surface a city. Built and validated on Guwahati; it now builds any city from open satellite data, with no account and no credentials.
Guwahati, Assam — the validated case
Heat dashboard
The grid as a continuous thermal surface. Select an area to view the post-mitigation planning scenario. Switch between Guwahati and six more cities, download any city's data, or drop in a grid you built yourself.
Open dashboard → Any cityBuild your own
One command builds a costed cooling plan for any city on Earth from open satellite data — no account, no API key. How the grid ids turned out to be a global lattice, and what does not transfer between cities.
Read the guide → SourceHow it works
Earth Engine workflow, the regression and tiering scripts, the cost engine, and the data contract they share.
View on GitHub →Recommended programme
₹9.99 Cr
249 highest-ranked cells · capped at ₹10 Cr
The ₹167.5 Cr figure is not the recommendation. It is the upper-bound cost of treating all 4,157 eligible cells. The displayed cooling figures are planning assumptions: −0.99 °C applies to treated cells, while −0.51 °C is averaged across the full grid, including cells that receive no action.
Read the caveats: interventions are gated on real ESA WorldCover land cover, so nothing is proposed on water, wetland or existing tree cover. But the costs and the cooling figures are planning assumptions, not measurements, and the model’s honest score is R² 0.51 under a spatial-block split — not the 0.90 a random split reports. Full limitations →
On the other cities: their temperatures are measured the same way, but they are hot-season composites while Guwahati is an annual median, so the figures are not comparable city to city. Heat-risk bounds are per city, so priority tiers are not comparable either, and some cities cover a whole administrative outline while others cover a window inside one. And the rupee rates are Indian municipal rates — Phoenix is included as the worked example of a cost figure that computes cleanly and means nothing. How the multi-city build works →