Advancing the Utility of Crowdsourced Rainfall Networks for Improved Community Flood Resilience

Floods are the most common natural disaster causing significant economic damage, loss of life, and social impacts each year. With climate change resulting in increases in the frequency and intensity of storms, these flooding impacts will only grow in the coming years. A first step in mitigating flooding impacts to communities is measuring rainfall rates at high resolutions in space and time. While there are government-operated in situ rainfall monitoring networks and remote sensing methods for measuring rainfall, these approaches alone do not produce the resolution and accuracy of rainfall data needed for hyper-resolution, street-scale flood modeling. Crowdsourced rainfall monitoring already available in many communities through the use of Personal Weather Stations (PWSs) offers a potential solution to this problem. Adoption of PWSs has been growing exponentially and, for many populated regions, now provides the ability to observe rainfall at a much finer spatial resolution compared to traditional rainfall monitoring networks. Despite this potential benefit, there are several challenges associated with using PWSs rainfall data in decision making that have yet to be assessed. This dissertation research focuses on three of these challenges through three studies organized under the general topics (1) data trustworthiness, (2) data representativeness, and (3) data usefulness. First, data trustworthiness speaks to rainfall data collected from PWSs that, because they are crowdsourced, are often less trusted compared to more traditional government-maintained rainfall monitoring networks. Next, data representativeness speaks to PWS adoption patterns that can exhibit spatial clusters correlated with socio-economic factors and demographic features. Finally, data usefulness speaks to the ability to use crowdsourced rainfall data to improve our understanding of flood losses. To address these challenges, the first study focused on developing a reputation system approach to assess the trustworthiness of crowdsourced PWSs. The second study focused on exploring the spatial pattern of PWS adoption. The third study focused on analyzing the characteristics of rainfall and tide, as well as the ability of PWS rainfall data in explaining recorded flood losses based on a Federal flood insurance claim dataset. Key findings from this research are (1) a reputation system approach can ensure trustworthy and accurate 15-min rainfall estimates from crowdsourced PWSs, (2) current crowdsourced PWS adoption patterns exhibit spatial biases toward wealthier neighborhoods and flood-prone regions, (3) the spatial density of PWSs provides the ability to better capture localized storms that resulted in flood claims. Across these three topics, this dissertation advances methods that improve the utility of crowdsourced rainfall data available through PWS networks. These advancements contribute to the growing knowledge related to crowdsourcing in general, which can assist in community-scale flood assessment and management.