University of Twente Student Theses


Towards a generalizable Urban Heat Island assessment tool using data-driven models

Dikkers, Marco (2023) Towards a generalizable Urban Heat Island assessment tool using data-driven models.

[img] PDF
Abstract:Cities all over the world are increasingly exposed to the Urban Heat Island (UHI) phenomenon as a result of global warming and urbanization. Data-driven models have gained popularity over the recent years and are used to analyze UHI intensity in different urban settings. These models specifically address the need for an easy-to-use assessment tool that can incorporate UHI concerns in the decision-making of urban planning. Multiple data-driven models that employ location specific data have been developed and successfully validated for Land Surface Temperature (LST) estimation in cities. It would be of great value to be able to reuse the trained UHI models for multiple cities, because this saves the effort to train the model with location specific data for every city in advance. Yet, previous research provides very little insight into the extent to which such data-driven models are generalizable for cities that are different in size, population, and regional climate. This research aims to conduct a comprehensive generalizability study of a Random Forest (RF) regression approach in relation to different levels of similarities between the urban characteristics of the cities used in this research. To this end, five different cities from three different countries were selected to cover a diverse range of (dis-)similarities in urban contexts. The individual models that were developed for each city are shown to be accurate in LST estimation in the cities for which they were trained. However, external cross-validation of the model to data from other urban contexts reveals that the proposed data-driven models have very low generalization capabilities, regardless of the observed (dis-)similarities between the cities. It was concluded that small changes in the feature properties can result in significant variation in the UHI behavior, and therefore the generalization is deficient. The results of this research show that the emergence of UHI is very context-specific, and that implies that implementation of standardized/universal mitigation strategies across cities world wide may be inappropriate. Instead, urban planners should address UHI in a location-specific manner, considering the local UHI mechanism for any given city.
Item Type:Essay (Master)
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering and Management MSc (60026)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page