Rapid Damage Assessment Following Natural Disasters through Information Integration

Abstract

A rapid damage assessment is essential for practitioners to make timely and informed decisions following a disaster. This research aims to provide such an assessment through integrating multisource information that comprises hazard characteristic, community exposure, community vulnerability, and social media information. To illustrate the reliability of the proposed strategy, supervised learning was employed because its performance highly relies on the quality of information integration. In detail, reference samples were prepared using the information of three recent hurricanes: Harvey, Irma, and Michael. Then two supervised learning models—multiple linear regression and support vector regression—were trained using the reference samples from Hurricanes Harvey and Irma. The trained models were tested using the reference samples from Hurricane Michael to demonstrate the applicability of the proposed approach. Theoretically, this research proves the concept of integrating multisource information for achieving a rapid damage assessment. Practically, this research proposes the whole pipeline from information collection to final prediction for deriving a rapid damage assessment following disasters.