Thermal Intelligence

The Thermal Impacts of Urban Densification : Using Deep Learning for climate assessments of urban development scenarios in Rotterdam

Master Thesis (2026)
Author(s)

Y. Yang (TU Delft - Architecture and the Built Environment)

Contributor(s)

D. Maiullari – Mentor (TU Delft - Architecture and the Built Environment)

A. Ersoy – Mentor (TU Delft - Architecture and the Built Environment)

Faculty
Architecture and the Built Environment
More Info
expand_more
Publication Year
2026
Language
English
Graduation Date
06-07-2026
Awarding Institution
Delft University of Technology
Programme
Architecture, Urbanism and Building Sciences, Urbanism
Faculty
Architecture and the Built Environment
Page Views
40
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Faced with a growing population, the Dutch government is promoting development plans to increase built density and accommodate the demand of new homes within existing urban areas. However, high-density developments are widely considered to exacerbate heat risks, causing health risks for vulnerable populations. Despite the known heat-related impacts on urban livability, effective methods are currently lacking at the municipal level for adopting heat-sensitive densification strategies while the high computational cost of existing simulation tools makes rapid assessment of multiple planning scenarios difficult. To address these challenges, the study addresses a densification case in Rotterdam and develops an AI-based methodology to quantify the thermal impact of increasing density within an existing urban area. The U-Net approach is used to estimate Tmrt and PET at a resolution of 1 m based on SOLWEIG-calculated data. This model reaches very high levels of precision (R2 ≈ 0.95). It also allows rapid assessment of different scenarios while maintaining high resolution.

Nine densification scenarios with different floor space index and morphological configurations were tested through the U-Net model, showing that it can generate high speed prediction with high accuracy. The modelling results show that the morphological characteristics of the new buildings and street canyons, and the distribution of building masses lead to different levels of diurnal and nocturnal increase of temperature and thermal stress. On wider streets, densification through horizontal development contributes to higher Tmrt and PET compared to vertical development, as high-rise buildings cast more shade especially during the day. Besides, vertical development also exhibits less heat retention and foster better ventilation than horizontal development. In narrow streets, horizontal development has better shading and ventilation performance compared to vertical development. Although horizontal development slightly increases heat retention, the effect is minimal and can be reduced by the guiding airflow movements.

Files

A3_Report_Yilin_Yang.pdf
(pdf | 281 Mb)
License info not available
Presentation.pdf
(pdf | 144 Mb)
License info not available