Improved Transferability of Data-Driven Damage Models Through Sample Selection Bias Correction

Disaster risk and artificial intelligence: A framework to characterize conceptual synergies and future opportunities

Artificial intelligence (AI) methods have revolutionized and redefined the landscape of data analysis in business, healthcare, and technology. These methods have innovated the applied mathematics, computer science, and engineering fields and are showing considerable potential for risk science, especially in the disaster risk domain. The disaster risk field has yet to define itself as a necessary application domain for AI implementation by defining how to responsibly balance AI and disaster risk. (1) How is AI being used for disaster risk applications; and how are these applications addressing the principles and assumptions of risk science, (2) What are the benefits of AI being used for risk applications; and what are the benefits of applying risk principles and assumptions for AI-based applications, (3) What are the synergies between AI and risk science applications, and (4) What are the characteristics of effective use of fundamental risk principles and assumptions for AI-based applications? This study develops and disseminates an online survey questionnaire that leverages expertise from risk and AI professionals to identify the most important characteristics related to AI and risk, then presents a framework for gauging how AI and disaster risk can be balanced. This study is the first to develop a classification system for applying risk principles for AI-based applications. This classification contributes to understanding of AI and risk by exploring how AI can be used to manage risk, how AI methods introduce new or additional risk, and whether fundamental risk principles and assumptions are sufficient for AI-based applications.

Multi-Hazard and Spatial Transferability of a CNN for Automated Building Damage Assessment

Automated classification of building damage in remote sensing images enables the rapid and spatially extensive assessment of the impact of natural hazards, thus speeding up emergency response efforts. Convolutional neural networks (CNNs) can reach good performance on such a task in experimental settings. How CNNs perform when applied under operational emergency conditions, with unseen data and time constraints, is not well studied. This study focuses on the applicability of a CNN-based model in such scenarios. We performed experiments on 13 disasters that differ in natural hazard type, geographical location, and image parameters. The types of natural hazards were hurricanes, tornadoes, floods, tsunamis, and volcanic eruptions, which struck across North America, Central America, and Asia. We used 175,289 buildings from the xBD dataset, which contains human-annotated multiclass damage labels on high-resolution satellite imagery with red, green, and blue (RGB) bands. First, our experiments showed that the performance in terms of area under the curve does not correlate with the type of natural hazard, geographical region, and satellite parameters such as the off-nadir angle. Second, while performance differed highly between occurrences of disasters, our model still reached a high level of performance without using any labeled data of the test disaster during training. This provides the first evidence that such a model can be effectively applied under operational conditions, where labeled damage data of the disaster cannot be available timely and thus model (re-)training is not an option.