Flooding is the most common natural hazard, leading to property damage, injuries, and death. Despite the potential for major consequences, urban flooding remains difficult to forecast, largely due to a lack of data availability at fine spatial scales and associated predictive capabilities. Crowdsourcing of public webcams, social media, and citizen science represent potentially important data sources for obtaining fine-scale hydrological data, but also raise novel challenges related to data reliability and consistency. We provide a review of literature and analysis of existing databases regarding the availability and quality of these unconventional sources that then drives a discussion of their potential to support fine-grained urban flood modelling and prediction. Our review and analysis suggest that crowdsourced data are increasingly available in urban contexts and have considerable potential. Integration of crowdsourced data could help ameliorate quality and completeness issues in any one source. Yet, substantial weaknesses and challenges remain to be addressed.

洪水是最常见的自然灾害，会导致财产损失、伤害和死亡。尽管有可能产生重大后果，但城市洪水仍然难以预测，这主要是由于缺乏精细空间尺度的数据可用性和相关的预测能力。公共网络摄像头、社交媒体和公民科学的众包代表了获取精细水文数据的潜在重要数据源，但也提出了与数据可靠性和一致性相关的新挑战。我们提供了关于这些非常规来源的可用性和质量的现有数据库的文献综述和分析，然后推动了对它们支持细粒度城市洪水建模和预测的潜力的讨论。我们的审查和分析表明，众包数据在城市环境中越来越可用，并且具有相当大的潜力。众包数据的整合可以帮助改善任何一个来源的质量和完整性问题。然而，仍有大量弱点和挑战有待解决。