Application of machine learning techniques for the estimation of seismic vulnerability in the city of Port-au-Prince (Haiti)

Abstract

High resolution, city-level exposure databases are important tools for risk planning and damage assessment. However, this kind of information is not always available, especially in developing countries where cities have grown rapidly without excessive planning or oversight, and sending a team of engineers to manually generate such a database by traditional means can be prohibitively costly. In recent years, we have seen tremendous growth in the availability of Earth Observation (EO) data, such as multispectral imagery or LiDAR point clouds, and also the processing capabilities of said data, not just in raw computing power but also in novel statistical modeling and analytical techniques known as Machine Learning (ML) and Computer Vision (CV). The aim of this work is to apply Machine Learning methods to aid in the generation of such databases as a case study in Port-au-Prince, Haiti, and to expose these methods in a comprehensible, reproducible fashion, as a series of open-source programming scripts.