Crowdsourced Data Assisting Disaster Relief Practices in the Era of Social Sensing

Crowdsourced data collected through mobile devices have received unprecedented attention from disaster relief taskforces in recent years as an effective way to monitor human activities. In this work, patterns of population displacement during mega-fires, the COVID-19 pandemic, and other disasters are investigated using Facebook Disaster Maps. We demonstrate an operational application to retrieve spatial and temporal patterns of population displacement in a timely manner using emerging hot spot analysis as well as crisis population products from Facebook. We pinpoint challenges in the application of crowdsourced data, including 1) the need to conduct a comprehensive evaluation of the data's representativeness, 2) the possibility of employing remote sensing data sources to a fusion method to reduce bias in the data without violating users' digital privacy, and 3) the necessity to develop recommendations for policy-makers regarding the most appropriate type of crowdsourced data to use during an emergency. Lastly, we discuss the importance of better understanding frontline disaster response needs and bridging the gap between scientific research and the deliverables needed in actual decision-making.