RAPID: The Changing Roles of Social Media in Disaster Resilience: The Case of Hurricane Harvey

了解社交媒体在灾害事件中的作用和影响的变化，例如德克萨斯州的哈维飓风，将有助于降低脆弱性，提高社区对这些灾害事件的复原力。飓风哈维于2017年8月25日在德克萨斯州罗克波特附近登陆，为4级飓风。它在休斯顿地区挥之不去，带来了超过50英寸的降雨量，在该地区造成了大范围的洪水和破坏。这一史无前例的灾难性事件暴露了许多问题，包括洪水预警不足和机构反应迟缓。与此同时，哈维事件期间出现了一个新现象：当911系统超载且无法连接时，休斯顿地区的许多居民求助于社交媒体，从被洪水淹没的房屋中呼救。社交媒体使用的这种变化标志着哈维是社交媒体在促进快速响应救援任务方面发挥了重要作用的第一批灾难性事件之一。首要的研究问题是：社交媒体通过其在应对和救援中的新角色在增强韧性方面有多有效，我们是否看到可能影响个人和社区结果和韧性的社交媒体使用的地理和社会差异的增加或减少？该项目收集对时间敏感的推特数据和对休斯顿地区受洪水影响社区的个人和组织的在线调查，以便利用这些数据来解决这一关键问题。研究小组收集了五种类型的数据，包括(1)哈维事件期间的Twitter数据；(2)推文中嵌入的相关网页和多媒体；(3)两次时间序列在线调查，以跟踪居民的情绪、适应策略和不确定情况下的去留决定；(4)对机构和居民进行的关于Twitter在救援行动中的新使用的在线调查；(5)用于与其他类型的数据整合的地理信息层，如洪水地图、破坏和社会经济数据。测试了Twitter数据挖掘和社交网络分析的方法。这些及时收集的第一手数据提供了有关个人和组织的关键问题、情绪和适应性行为的信息。这对决策者和急救人员来说是有价值的信息，从而支持制定更好的战略以减少脆弱性和提高复原力的努力。该项目的结果还可以与2012年的飓风事件进行比较，从而进一步了解差距是扩大了还是缩小了，以及在哪里。从计算角度来看，开发更好、更高效的大数据挖掘算法将推进弹性计算和分析。来自该项目的网站、数据库和报告将可用并广泛传播。该项目所吸取的经验教训和使用的方法可用于研究和比较不同区域的不同灾害。

Understanding the changing roles and effects of social media use in disaster events, such as Hurricane Harvey in Texas, will help reduce vulnerability and improve resilience of communities to these disaster events. Hurricane Harvey made landfall on August 25, 2017 near Rockport, Texas as a category-4 hurricane. It lingered over the Houston area and dumped over 50-inches of rainfall, causing widespread flooding and damages in the region. This unprecedented disastrous event reveals many issues, including inadequate flood warning and slow response by agencies. At the same time, a new phenomenon emerged during the Harvey event: many residents in the Houston area resorted to social media to call for rescue from flooded homes when the 911 system was overloaded and could not be connected. This changing use of social media marks Harvey as one of the very first disastrous events in which social media have played an important role in facilitating fast-responding rescue missions. The overarching research question is: how effective is social media in enhancing resilience through its new role in response and rescue, and do we see an increase or decrease in the geographical and social disparities of social media use that may have affected the outcome and the resilience of individuals and communities? This project collects time-sensitive Twitter data and online surveys of individuals and organizations in the flood-affected communities in the Houston region so that they can be used to address this key question. The research team collects five types of data including (1) Twitter data during the Harvey event; (2) associated webpages and multimedia embedded in tweets; (3) two time-series online surveys to track residents' sentiment, adaptation strategies, and decisions under uncertainty to stay or leave; (4) online surveys of agencies and residents regarding the new use of Twitter in rescue operations; (5) geographic information layers such as flood maps, damage, and socioeconomic data used to integrate with the others types of data. Methods for mining Twitter data and social network analysis are tested. These first-hand, timely collected data provide information on key concerns, sentiment, and adaptive behavior of individuals and organizations. This is valuable information for decision makers and first responders, thereby supporting efforts to map out better strategies to reduce vulnerability and improve resilience. Results from this project can also be compared with the hurricane events in 2012, thus gaining further insights into whether disparities have increased or decreased and where. From a computational point of view, developing better and more efficient algorithms for mining big data will advance the computation and analysis of resilience. The websites, databases, and reports derived from this project will be available and widely disseminated. The lessons learnt and the methodology used in this project can be utilized to study and compare different disasters in different regions.