|
Citation
|
Wang X, Dao F, Ji Y, Qiu S, Zhu X, Dong W, Wang H, Zhang W, Zheng X. Innovation (N Y) 2024; 5(4): e100647. |
|
Abstract
|
Natural disasters, including floods, storms, and tsunamis, pose a great threat to human societies. A recent study highlighted this concern, revealing that billions of people globally were exposed to flood hazards. In 2023, Super Typhoon Doksuri caused devasting floods in Beijing and Hebei areas, resulting in massive casualties and huge economic losses. Therefore, there is a need for a precise understanding of disaster processes, reliable forecasting of disaster effects, and timely warning of risks to prevent and mitigate major disasters.2
Numerical modeling stands as the predominant approach to meet these demands. However, the predictive accuracy of such numerical models could be degraded because of various factors: oversimplification of real processes, computational errors, fluctuations of complex environments (e.g., terrains, precipitations, buildings, and plants), and the influence of human activities (e.g., evacuation and rescue) during disasters.
Improving models with a large number of observational data through methods like machine learning or data assimilation is widely recognized as an effective approach. Nevertheless, the lack of disaster-related data and the practical difficulty encountered in data collection, particularly during emergencies, have become major challenges in disaster forecasting. In addition, even after data collection, there is also a research gap in terms of identifying model drawbacks and acquiring improvement solutions based on limited datasets. Crowdsourcing is a rising approach employing crowds to complete various tasks.3
By soliciting insights from people both physically located in disaster zones and virtually connected online, it holds potential to obtain disaster data and pave the way for model improvements. Yet, the question of how to effectively apply crowdsourcing to improve disaster forecasts remains unexplored.
Language: en |