Performance‑driven optimization of nature‑based solutions for advancing urban canyon shading

Source abstract

Climate change is affecting environmental quality, especially at the urban canyon level, leading to hazards such as heat stress, and intensified heatwaves. To enhance urban resilience and mitigate rising temperatures, systemic solutions are required. Nature-Based Solutions (NBS) are recognized as promising strategies, yet their implementation often lacks proper planning and quantification of performance. Existing simulation tools remain limited by high computational demands. This study applies Machine Learning (ML) to reduce reliance on simulations, streamline workflows, all while optimizing NBS performance, specifically urban street trees. A comprehensive dataset was developed, integrating urban canyons geometries and trees characteristics representative of typical urban conditions and plant species commonly used in design. A validated real time tool for computational simulations was used to investigate correlations between site morphology and street trees shading performance, which were then generalized across the Italian territory through Machine Learning (ML) approaches. Across more than 300’000 simulated canyon-tree configurations, latitude emerged as the dominant predictor of shading performance, while wider canyons and larger crown radii consistently achieved high scores (TOPSIS Scores > 0.9), identifying single- and double-row layouts as the most effective solutions. This methodology demonstrates that urban trees shading effectiveness can be predicted and generalized thanks to the representativeness of the dataset. These findings support the integration of NBS into early-stage design, providing a scalable framework for improving microclimate regulation in cities.