At a glance  

This research addresses critical gaps in the availability of a nationally consistent, geo-tagged database of unreinforced masonry (URM) buildings across Australia. URM structures are widely recognised as potentially vulnerable during earthquake events and can pose significant risks to life safety, emergency response operations, and post-disaster recovery. This project is developing a comprehensive and accessible national inventory that will enable emergency management organisations to strengthen risk mitigation strategies.

Existing databases are fragmented and rely heavily on time-consuming and costly manual surveys that are inconsistent across jurisdictions. This research responds to the needs identified by key emergency management stakeholders, including the Queensland Fire Department, South Australia Police, Department of Fire and Emergency Services Western Australia, and Fire and Rescue NSW. It leverages artificial intelligence (AI) and computer vision techniques to detect and classify URM buildings at scale. The resulting geospatial database will support improved preparedness of emergency management agencies across Australia, particularly in scenarios involving large buildings and resource constraints.

Organisation, sector and geographical location involved

The project is funded by Natural Hazards Research Australia (NHRA) and led by Queensland University of Technology (QUT) in collaboration with the University of Newcastle and emergency management organisations across Australia.  

Through ongoing engagement with stakeholders and utilisation activities, the research supports continuous improvement in earthquake risk mitigation primarily by strengthening emergency planning and preparedness.

Key findings or outcomes of the research

At time of writing this case study, the project is approximately 45 per cent complete and has developed an automated pipeline for detecting and geo-tagging URM buildings across Australia using AI and computer vision techniques. The methodology integrates satellite imagery, large geospatial datasets, and street-level imagery to identify and classify buildings that may exhibit characteristics associated with URM construction.

Preliminary results indicate that, within the practical limitations of current AI and computer vision technologies, automated identification of URM building types is both feasible and cost-effective when compared with traditional field surveys. This highlights the potential for scalable national assessments of vulnerable building stock.

As the project is ongoing, the full set of outcomes and operational applications for emergency management stakeholders is still being developed. Remaining work will focus on completing the datasets and demonstrating how the resulting database can support mitigation planning and disaster preparedness activities.

What this means for the disaster management sector

This project delivers the first nationally consistent geospatial inventory of potentially vulnerable URM buildings in Australia. Once all URM buildings are identified, the standardised classification across jurisdictions allows the disaster management sector to adopt a more coordinated and comparable approach to earthquake risk assessment, preparedness, and mitigation planning. Emergency management stakeholders access, contribute to, and learn from shared datasets and mitigation strategies, supporting collaboration and knowledge exchange nationally.

The database is delivered in a GIS-compatible format, enabling spatial analysis of URM building distributions, surrounding infrastructure, and access routes. This information integrates with existing or emerging earthquake monitoring and alert systems to help identify areas where building damage and community impacts are more likely to occur.  

The dataset supports preparedness activities, such as developing realistic training scenarios for Urban Search and Rescue (USAR) teams. Beyond emergency response, the database provides valuable exposure information for government agencies, planners, and the insurance sector to support earthquake scenarios development and proactive mitigation strategies. Collectively, these capabilities support evidence-based planning, improve situational awareness, and disaster resilience across Australia.

Entities relevant to this research include universities, state and local governments, emergency management organisations, planning authorities, and the insurance sector. These groups play key roles in identifying, assessing, and managing risks associated with vulnerable building stock and contribute to the development and utilisation of geospatial datasets disaster risk management.

The study primarily focuses on the earthquake hazard, with particular attention to the vulnerability of URM buildings. It also considers the implications of earthquake impacts for urban environments, critical infrastructure, and emergency response operations.

Thematically, the work focuses on collaboration and coordination, continuous improvement, risk management, preparedness planning, and resilience.  It also highlights the role of AI, computer vision, geospatial analysis, and exposure databases in enabling large-scale risk identification and coordinated mitigation strategies across jurisdictions.

Key search words

Unreinforced masonry (URM) buildings; Earthquake risk mitigation; Seismic vulnerability; Building exposure database; Artificial intelligence (AI); Computer vision; Convolutional neural networks (CNN); Geospatial analysis; Geographic Information Systems (GIS); Satellite imagery; Street-level imagery; Automated building detection; Disaster risk reduction; Emergency management; Urban search and rescue (USAR); Earthquake preparedness

Introduction

Legacy URM buildings present significant earthquake risk in Australia. There are approximately 1,100 such buildings in Queensland, mostly concentrated in the central business districts. Built prior to seismic design codes, these structures were designed mainly for gravity loads, with no consideration of earthquake forces. Australia experiences one to two magnitude-5 (M5) earthquakes annually, with one of these earthquakes occurring in Queensland every five years. Earthquakes of this size can damage URM buildings, as demonstrated by the 1989 Newcastle earthquake (M5.6).

Research problem  

In the event of an earthquake affecting a major Australian town, damage to URM buildings may present significant challenges for emergency management organisations. Concentrated damage in heritage precincts, may delay search and rescue operations due to factors including road blockages from debris, restricted access to damaged structures, and high logistical support demands that may exceed available capacity.

Effective preparation and training require a comprehensive and accurate understanding of URM buildings. However, a national database of these buildings is currently unavailable, or exists only in a limited and fragmented form across certain regions of Australia. This gap constrains the development of a coordinated and consistent framework for training exercises and emergency response planning among stakeholders across Australian states.

Research aim

The project seeks to document the presence of URM buildings nationwide and classify them into seismic vulnerability groups. The primary output will be GIS maps containing URM building footprints. The project also includes utilisation activities to strengthen emergency management organisations to use the datasets for improved risk mitigation and preparedness.  

Significance  

This project makes a significant contribution to seismic risk reduction across Australia, particularly given the disproportionately high earthquake risk associated with URM buildings and the concentration of these structures within populated areas.

Methodology

By integrating AI, computer vision, and geospatial technologies, the project aims to automatically detect URM buildings using big data, satellite imagery and publicly available street view images. Images of known URM buildings are being used to train a class of computer algorithms known as Convolutional Neural Networks (CNNs). Once trained, the CNN will automatically detect URM buildings from street-level imagery. This approach removes the need for extensive foot surveys, substantially reducing associated costs.  

The workflow comprises three stages:  

1. Training data collection (completed as of March 2026) 
2. CNN model training (commenced) 
3. Automated building detection (to start June 2026).

Much of this process involves automated retrieval of building images and footprints from platforms such as Google, Bing, and Microsoft, along with computer coding to detect and classify images.  

Results and discussion

The results to date include:

a) Establishing definitions and classifications for vulnerable groups, in consultation with stakeholders
b) Development of a full pipeline to collect street view images from multiple sources (e.g. Microsoft, Bing and Google)  
c) Automated image collection for known URM buildings for CNN training
d) Annotation of the building images using specialised software for CNN training.  

The buildings include previously surveyed URM buildings in Queensland and many URM buildings from other Australian cities, including Sydney, Adelaide and Melbourne. 

Conclusion (implications, impact & insights)

The research demonstrates the value of AI and computer vision for systematically detecting and classifying URM buildings, enabling faster and more economical seismic risk mitigation strategies. It also provides a proof-of-concept for this methodology that may be applicable to other hazard building types.  

The study delivers a national GIS database of URM buildings, offering emergency management agencies a structured tool for mitigation planning, preparedness exercises, and coordinated response strategies.  

The project strengthens cross-agency collaboration and supports interoperable building data, contributing to improved earthquake preparedness and resilience. Integrating these insights into disaster risk reduction frameworks helps ensure risk-informed planning and mitigation becomes standard practice across the sector.

What the results mean for disaster management in Queensland

From Queensland Fire Department’s perspective, the creation of a national database of URM buildings represents a significant advancement in delivering hazard and risk intelligence for earthquakes. By harnessing geospatial data, AI, and computer vision, the project provides a practical and scalable solution to enhance earthquake preparedness, response, and recovery.  

The resulting GIS maps of URM building locations help emergency services anticipate and mitigate seismic risks, particularly in urban areas with dense concentrations of vulnerable structures. This supports more precise and effective urban search and rescue operations, rapid damage assessments, ultimately helping to safeguard communities and minimise impacts. The project also sets a benchmark for modern, technology enabled risk-informed decision-making, demonstrating how emerging capabilities can be integrated into contemporary disaster management practice.

Next steps

Next steps include training the CNN using the developed pipeline to automatically detect URM buildings and presenting the results in a GIS-based database for emergency management stakeholders. Engagement meetings and workshops will explore how the dataset can support mitigation planning, preparedness exercises, and integration with alert systems. These activities will ensure the database is practical, accessible, and supports evidence-based risk reduction, enabling stakeholders to enhance earthquake preparedness and resilience across jurisdictions.

References  

Project profile including link to presentations in the funding agency website:  

Natural Hazards Research Australia (NHRA) website

Questions?

Please email all questions and photos to: research@igem.qld.gov.au 
Research Connect case studies webpage  
 

NHRA Project Manager

Nicola Moore; Node Research Manager Qld & NT (nicola.moore@naturalhazards.com.au)
 

Researchers

Dr Hossein Derakhshan; Senior Lecturer (QUT) and NHRA Principal Researcher (hossein.derakhshan@qut.edu.au)

Dr Alan Woodley; Senior Lecturer (QUT) and NHRA Chief Investigator (a.woodley@qut.edu.au)

Prof Mark Masia; Professor (University of Newcastle) and NHRA Chief Investigator (m.masia@newcastle.edu.au)

Dr Abdullah Nazib; Postdoctoral Research Associate (QUT and NHRA; a.nazib@qut.edu.au)

Dr Nouman Khattak; Postdoctoral Research Associate (QUT and NHRA; n.khattak@qut.edu.au)
 

Qld key stakeholders  

Matthew Dyer; Queensland Fire Department (matthew.dyer@fire.qld.gov.au)

Dr Jane Sexton; Queensland Fire Department (jane.sexton@fire.qld.gov.au)