A framework for incorporating rainfall data into a flooding digital twin

The use of digital twins in operational flood management has garnered attention for its potential to enhance real-time flood risk monitoring and the deployment of timely interventions. With increasing access to real-time rainfall data from national meteorological services, low-cost sensors networks, citizen science data and dense observation networks in ‘urban observatories’, there is immense potential for developing detailed digital twins for flooding. However, the useability of these real-time rainfall data sources, including the reliability and quality of rainfall data for real-time digital twin applications has not yet been analysed. This study investigates the suitability of current real-time rainfall data sources for digital twin applications and identifies barriers to operationalising a flooding digital twin. Using the PYRAMID (near) real-time dynamic flood modelling platform and Newcastle upon Tyne (U.K.) as a demonstrator – due to its high data availability and diversity of rainfall data sources – we evaluate the reliability of rain gauge and radar rainfall data from the Environment Agency, U.K. Meteorological Office and additional sources, including citizen science and urban observatory data. The reliability of real-time data is shown to be a major barrier to digital twin deployment due to variability in data quality and intermittent data streams, even with national rainfall data provided by government organisations. To address these challenges, we propose a blending algorithm that adapts to changing data availability in real-time, implemented within the PYRAMID workflow. Flood depths are shown to be sensitive to data blends, with peak rainfall rates varying by a factor of 10, leading to flood depths differing by up to 15%. This paper highlights the critical need for improved rainfall data reliability to enable the operational use of digital twins for flood management.