Speaker
Description
Background
Global cities face unprecedented crises ranging from escalating climate threats to entrenched social disparities. Rapid growth-oriented urbanization over recent decades has led to environmental depletion and magnified socioeconomic marginalization. As these intertwined challenges intensify, there is an urgent need to re-evaluate the nexus of development goals, ecological health, and equitable spatial development. Advances in artificial intelligence (AI) and multi-source data now enable precise and real-time urban hazard risk assessment. Fine-grained spatial data, integrated with temporal dynamics, provide critical insights for quantifying the socio-spatial fairness of disaster exposure, while deep learning and transfer learning techniques further facilitate more accurate forecasting of future hazards.
Purpose
Focusing on two major urban hazards—extreme heat and urban flooding—this study aims to assess compound spatial risks in Shenzhen and predict future dual-hazard exposure under various CMIP6 emission scenarios. By integrating spatial evaluation with robust forecasting, we seek to bolster urban and regional resilience to environmental threats while ensuring socio-spatial equity.
Methods
Using Shenzhen as a case study, a typical Chinese south metropolitan with serious extreme heat and urban flooding, we developed an AI-driven, fair-disaster-exposure assessment and prediction framework. First, multi-sourced data—including temperature, precipitation, topography, 3D urban morphology, vegetation cover, river networks, and socioeconomic attributes such as population age structure, income levels, and healthcare facilities—were acquired from remote sensing satellite imagery, social media platforms, and official statistics. Through a robust multi-phase spatiotemporal alignment approach combined with high-performance computing resources, we achieved precise data fusion at fine granularities. We then constructed a spatial model to evaluate the dual-disaster exposure risks from the perspective of social equity, identifying high-vulnerability and unequal regions. Moreover, we utilized deep learning and transfer learning to predict future compound hazard exposures under different carbon emission scenarios, offering critical data-driven evidence for building resilient cities.
Results
Our findings reveal a significant spatial overlap of extreme heat and flood risks in Shenzhen, emphasizing the intricate interplay between urban form and infrastructure, population density, and climate change. Legacy districts with aging infrastructure face amplified vulnerabilities, often due to limited drainage capacity, suboptimal building materials, and older population groups. Meanwhile, some newly developed urban zones, albeit more modern in appearance, also experience high dual exposure due to planning processes did not fully integrate robust flood mitigation measures and heat management strategies. Notably, areas with large concentrations of low-income families and seniors—populations typically less capable of managing abrupt environmental challenges—demonstrate disproportionately high vulnerabilities and socio-spatial inequalities. These cumulative factors highlight the need to prioritize integrated adaptation and mitigation policies. In addition, our self-adaptive data-driven deep transfer learning model consistently yielded robust predictive insights, accurately identifying locations at greatest future risk under multiple climate scenarios drawn from CMIP6. The successful application of this advanced modeling approach underscores the potential for more comprehensive and equitable resilience planning that not only addresses immediate physical exposures but also considers the long-term social implications. Ultimately, our results underscore the pressing necessity for strategic, well-informed interventions that can effectively reduce multifaceted urban climate risks while preserving social equity across Shenzhen’s diverse neighborhoods.
Conclusion
This research underscores the importance of bridging AI-driven analytics with social equity principles to reconcile environmental imperatives and inclusive urban development. By offering a transformative pathway for policymakers, urban planners, and stakeholders, our framework advocates prioritizing collective well-being in climate-resilient strategies. In doing so, deliberate attention to fairness, adaptability, and innovation will enable urban and regional systems to flourish amid unprecedented global challenges.
References
Eyring, V., Bony, S., Meehl, G.A., Senior, C.A., Stevens, B., Stouffer, R.J. and Taylor, K.E., 2016. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development, 9(5), pp.1937-1958.
Caggiano, H., Kocakuşak, D., Kumar, P. and Tier, M.O., 2023. US cities’ integration and evaluation of equity considerations into climate action plans. npj Urban Sustainability, 3(1), p.50.
Elmqvist, T., Andersson, E., Frantzeskaki, N., McPhearson, T., Olsson, P., Gaffney, O., Takeuchi, K. and Folke, C., 2019. Sustainability and resilience for transformation in the urban century. Nature sustainability, 2(4), pp.267-273.
He, Z., Wu, Z., Herzog, O., Hei, J., Li, L. and Li, X., 2025. Compound health effects and risk assessment of extreme heat and ozone air pollution under climate change: A case study of 731 urban areas in China. Sustainable Cities and Society, 119, p.106084.
| Keywords | AI-driven hazard modeling; Extreme heat and flooding; Socio-spatial equity; CMIP6 |
|---|---|
| Best Congress Paper Award | Yes |
