AI + Optimization + Urban Climate Model Development
From new climate-process code to decision-ready adaptation: a closed workflow for designing rainwater harvesting and roof sprinkling under extreme urban heat.
Research demo only. This page is a visual introduction to the project. Please read the full article below for the complete methodology, assumptions, numerical results, limitations and supporting evidence.
Integrated research pipeline
Model. Learn. Search. Evaluate.
The workflow couples physical process development with a fast AI approximation, enabling multi-objective search without running the full urban climate model for every candidate design.
Evaporation cools the roof and reduces heat gain
What the framework reveals
Three design insights
Trigger timing dominates
The roof-temperature threshold is more influential than tank size or sprinkling intensity across the tested design space.
Bigger is not always better
Increasing storage improves performance, but the marginal cooling-energy benefit diminishes as tank size grows.
Benefits extend beyond energy
Optimized strategies lower extreme temperatures, reduce heatwave days and mitigate extreme surface runoff.
Decision space
The Pareto frontier
There is no single best design. Each point represents a defensible compromise: smaller infrastructure at one end, greater cooling-energy reduction at the other.
The curve's flattening makes the diminishing return visible and helps decision-makers select a solution that matches local constraints.
Conceptual re-rendering based on the study workflow and reported Pareto relationship; not a reproduction of the paper's numerical dataset.
Full research article
Read beyond the demo
Optimizing the Rainwater Harvesting and Roof Sprinkling System to Adapt to Urban Extreme Heat