Providing guidance on efficient flash flood documentation: an application based approach

Flash flood research crucially relies on historical event information for hypothesis testing. However, flash flood science suffers from data scarcity. Due to the high effort for event data collection, few long-term and comprehensive flash flood datasets exist. Yet, to advance flash flood research, scientists should spend time on creating event datasets. Therefore, to reduce the data collection and preparation effort, this paper takes a first step towards structuring the flash flood documentation procedure. In this context, we first investigate the current documentation approaches by reviewing 11 published flash flood datasets. We found great differences between the flash flood datasets regarding temporal and spatial scope, information density, flash flood definition, and usability by others. Based on the review findings and the cross-industry standard process for data mining, we propose a structured 4-step approach for flash flood documentation. We provide recommendations on efficient flash flood documentation and exemplify a possible implementation based on a German flash flood dataset, starting from flash flood definition through the identification of sources to the schematic event documentation. The key feature of our approach is a documentation scheme specifying what event information to report and how. Within the documentation scheme, it proved particularly helpful to use fixed categories for attribute description, to rate information quality, and to document events separated by source. Following our approach, we were able to create a comprehensive event dataset composed of a variety of sources. In addition to flash flood events, our dataset also includes surface runoff events triggered by heavy rain, adding up to nearly 23,800 German events. Scientists can employ this approach to document flash floods more efficiently in the future.