Developing a Crisis Informatics Ecosystem for Smart Cities: Geosciences and Remote Sensing

Summary form only given. The complete presentation was not made available for publication as part of the conference proceedings. Crisis informatics has emerged in recent years as an interdisciplinary field which leverages numerous cyberinfrastructure resources to gather data around unforeseen events, particularly preparation, mitigation, response, and recovery phases of disasters. This interdisciplinary field combines numerous facets of computing and social science knowledge discovery. However, it has never been applied to a modern smart city after a major disaster such as hurricane Harvey and Irma. Hurricane Harvey and Irma, offers a unique opportunity to leveraging existing diverse cyberinfrastructure devices as crisis informatics engines to enhance recovery efforts within the city, while providing a national and international framework for such activities in the future. Houston, and Miami are considered two of America's leading smart cities, particularly in terms of the wireless sensors spread throughout the city, with sensors measuring air quality, core infrastructure edifices, such as roads and bridges, as well as many wireless camera systems. Many of these cyberinfrastructure resources were not impacted by the hurricane, which means these assets are fertile repositories for situational data within the context of locating individuals, missing animals, assessing damage, and most importantly providing valuable analysis of the city's infrastructure. Traditionally, in such situations the emergency management community has utilized these tools in limited fashion, and only in an ad-hoc manner, while relying heavily on surveys, first-hand inspections of physical infrastructure. In this paper, we describe a framework which utilizing plan recognition, which treats vast volumes of data aggregated form of smartphone messages, time-series social media, and IoT device data into a comprehensive crisis informatics partial-order plans. In this paper, we describe a framework which utilizing plan recognition, which treats vast volumes of data aggregated form of smartphone messages, time-series social media, and IoT device data into a comprehensive crisis informatics partial-order plans. These partial-order plans are discrete events or actions which can be organized into a plan, without imposing a specific order. In fact, partial-order plans are considered as representative of multiple plan-instances, obtained imposing a specific order in the execution of the actions. Utilizing, machine learning algorithms to automatically recognition of several (partially or fully achieved), can develop data-driven crisis mapping for resource allocation, and spatial clustering that indicate geographical areas for search and rescue operations, which exhibited abnormally high activities during the disaster.