An investigation into the role of crowdsourcing in generating information for flood risk management

Flooding is a major global hazard whose management relies on an accurate understanding of its risks. Crowdsourcing represents a major opportunity for supporting flood risk management as members of the public are highly capable of producing useful flood information. This thesis explores a wide range of issues related to flood crowdsourcing using an interdisciplinary approach. Through an examination of 31 different projects a flood crowdsourcing typology was developed. This identified five key types of flood crowdsourcing: i) Incident Reporting, ii) Media Engagement, iii) Collaborative Mapping, iv) Online Volunteering and v) Passive VGI. These represent a wide range of initiatives with radically different aims, objectives, datasets and relationships with volunteers. Online Volunteering was explored in greater detail using Tomnod as a case study. This is a micro-tasking platform in which volunteers analyse satellite imagery to support disaster response. Volunteer motivations for participating on Tomnod were found to be largely altruistic. Demographics of participants were significant, with retirement, disability or long-term health problems identified as major drivers for participation. Many participants emphasised that effective communication between volunteers and the site owner is strongly linked to their appreciation of the platform. In addition, the feedback on the quality and impact of their contributions was found to be crucial in maintaining interest. Through an examination of their contributions, volunteers were found to be able to ascertain with a higher degree of accuracy, many features in satellite imagery which supervised image classification struggled to identify. This was more pronounced in poorer quality imagery where image classification had a very low accuracy. However, supervised classification was found to be far more systematic and succeeded in identifying impacts in many regions which were missed by volunteers. The efficacy of using crowdsourcing for flood risk management was explored further through the iterative development of a Collaborative Mapping web-platform called Floodcrowd. Through interviews and focus groups, stakeholders from the public and private sector expressed an interest in crowdsourcing as a tool for supporting flood risk management. Types of data which stakeholders are particularly interested in with regards to crowdsourcing differ between organisations. Yet, they typically include flood depths, photos, timeframes of events and historical background information. Through engagement activities, many citizens were found to be able and motivated to share such observations. Yet, motivations were strongly affected by the level of attention their contributions receive from authorities. This presents many opportunities as well as challenges for ensuring that the future of flood crowdsourcing improves flood risk management and does not damage stakeholder relationships with participants.