Sensing climate justice : A multi-hyper graph approach for classifying urban heat and flood vulnerability through street view imagery
(2025) In Sustainable Cities and Society 118.- Abstract
Recognising the increasing complexities posed by climate challenges to urban environments, it is crucial to develop holistic capabilities for urban areas to effectively respond to climate-related risks, forming the backbone of sustainable urban planning strategies and demanding a comprehensive understanding of urban climate justice. It requires a thorough examination of how climate change exacerbates social, economic, and environmental inequalities within urban settings, which requires a series of sophisticated spatial modellings and relies on data collected periodically. This paper introduces a novel dual-GNN approach, Multi-Hyper Graph Neural Network (MHGNN), with street view imagery as input. The proposed model integrates a... (More)
Recognising the increasing complexities posed by climate challenges to urban environments, it is crucial to develop holistic capabilities for urban areas to effectively respond to climate-related risks, forming the backbone of sustainable urban planning strategies and demanding a comprehensive understanding of urban climate justice. It requires a thorough examination of how climate change exacerbates social, economic, and environmental inequalities within urban settings, which requires a series of sophisticated spatial modellings and relies on data collected periodically. This paper introduces a novel dual-GNN approach, Multi-Hyper Graph Neural Network (MHGNN), with street view imagery as input. The proposed model integrates a multigraph and a hypergraph to model intricate spatial patterns for classifying urban climate justice. The multigraph component of the MHGNN captures spatial proximity and pair-wise connections between urban areas to assess climate impacts. Meanwhile, the hypergraph component addresses higher-order dependencies by incorporating hyperedges that connect multiple geographic areas based on their similarities, thus capturing the multi-faceted relationships among areas with comparable geographic characteristics. By harnessing the strengths of both multigraph and hypergraph structures, the MHGNN provides a comprehensive understanding of the spatial dynamics of urban climate justice. It achieves nearly a 24% performance improvement compared to conventional spatial modelling methods, establishing it as a valuable tool for researchers and policymakers in this domain. Codes available at GitHub.1
(Less)
- author
- Liu, Pengyuan ; Lei, Binyu ; Huang, Weiming LU ; Biljecki, Filip ; Wang, Yuan ; Li, Siyu and Stouffs, Rudi
- organization
- publishing date
- 2025-01
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Graph neural network, Hypergraph, Multigraph, Spatial modelling, Urban resilience
- in
- Sustainable Cities and Society
- volume
- 118
- article number
- 106016
- publisher
- Elsevier
- external identifiers
-
- scopus:85211735698
- ISSN
- 2210-6707
- DOI
- 10.1016/j.scs.2024.106016
- language
- English
- LU publication?
- yes
- id
- fff8bf5e-ca63-4ca1-9fe0-ff763707a767
- date added to LUP
- 2025-02-26 11:58:50
- date last changed
- 2025-04-04 15:40:09
@article{fff8bf5e-ca63-4ca1-9fe0-ff763707a767,
abstract = {{<p>Recognising the increasing complexities posed by climate challenges to urban environments, it is crucial to develop holistic capabilities for urban areas to effectively respond to climate-related risks, forming the backbone of sustainable urban planning strategies and demanding a comprehensive understanding of urban climate justice. It requires a thorough examination of how climate change exacerbates social, economic, and environmental inequalities within urban settings, which requires a series of sophisticated spatial modellings and relies on data collected periodically. This paper introduces a novel dual-GNN approach, Multi-Hyper Graph Neural Network (MHGNN), with street view imagery as input. The proposed model integrates a multigraph and a hypergraph to model intricate spatial patterns for classifying urban climate justice. The multigraph component of the MHGNN captures spatial proximity and pair-wise connections between urban areas to assess climate impacts. Meanwhile, the hypergraph component addresses higher-order dependencies by incorporating hyperedges that connect multiple geographic areas based on their similarities, thus capturing the multi-faceted relationships among areas with comparable geographic characteristics. By harnessing the strengths of both multigraph and hypergraph structures, the MHGNN provides a comprehensive understanding of the spatial dynamics of urban climate justice. It achieves nearly a 24% performance improvement compared to conventional spatial modelling methods, establishing it as a valuable tool for researchers and policymakers in this domain. Codes available at GitHub.<sup>1</sup></p>}},
author = {{Liu, Pengyuan and Lei, Binyu and Huang, Weiming and Biljecki, Filip and Wang, Yuan and Li, Siyu and Stouffs, Rudi}},
issn = {{2210-6707}},
keywords = {{Graph neural network; Hypergraph; Multigraph; Spatial modelling; Urban resilience}},
language = {{eng}},
publisher = {{Elsevier}},
series = {{Sustainable Cities and Society}},
title = {{Sensing climate justice : A multi-hyper graph approach for classifying urban heat and flood vulnerability through street view imagery}},
url = {{http://dx.doi.org/10.1016/j.scs.2024.106016}},
doi = {{10.1016/j.scs.2024.106016}},
volume = {{118}},
year = {{2025}},
}