Machine Learning for Resilient Cities: Mapping of Food Accessibility and Shock Vulnerability in Metropolis Hong Kong ^†

Food accessibility and resilience in urban environments are critical components of sustainable development, particularly in densely populated metropolises. However, as global cities are growing rapidly and vertically, both above and underground, the contemporary 2D spatial paradigm in assessing food accessibility faces increasing limitations in being informative and suffers from hidden data gaps with vertical urbanisation. Therefore, a remapping of urban food accessibility with a multi-dimensional paradigm is pressing and critical to establish a new datum for sustainable and resilient city planning, a consideration often overlooked in previous studies.

This transdisciplinary study combines participatory GIS (PGIS) and 3D spatial analytics to map out food deserts and identify proximity-based fresh food accessibility (such as supermarkets, wet markets, grocery stores, online delivery hubs, food kitchens and local farms), with a focus on lower-income districts in Hong Kong. Furthermore, with only 1.2% arable land and 90% food import dependence, climate trends, economic shocks, and supply chain disruptions are layered into machine learning (ML) models to assess urban food system vulnerabilities under crises (e.g., extreme weather and pandemics) to optimise the spatial ecology of urban food systems.

Preliminary results reveal not only spatial disparities but also systemic food accessibility gaps in the 100% urbanised Hong Kong. The high-rise urban form, steep topography, and climate vulnerabilities like typhoons and flooding reveal how these landscape factors create uneven food security outcomes, particularly in ageing neighbourhoods and underprivileged communities such as Kwun Tong and Shum Shui Po districts. The study provides valuable empirical evidence to address systemic gaps in urban food resilience, advocating for data-driven interventions in infrastructure planning, emergency response, and equitable food distribution.

The methodology is scalable to other high-density urban areas, offering a replicable framework for enhancing food system robustness in the face of growing environmental and socioeconomic uncertainties. By treating food infrastructure as critical landscape elements, the study provides practical pathways for cities like Hong Kong to achieve SDG 11 (sustainable cities) while addressing the acute housing–land–food nexus. The findings underscore that urban resilience requires reimagining food access through both multi-dimensional technological innovation and ecological place-making.