Smart modelling of urban pluvial flooding: crowdsourced data and rapid simulation

Cloudbursts and extreme rainstorms pose an escalating threat to urban areas, requiring high-resolution rainfall data for accurate flood prediction. While official rain gauges are precise but sparse, and radar suffers from signal attenuation, crowdsourced data from private weather stations (PWS) offers dense spatial coverage. The first part of this Dissertation evaluates PWS performance across two flood events in Oslo. Although private sensors underestimate cumulative rainfall by approximately 25% on average, they demonstrate excellent event detection. Results show that bias-corrected PWS data generates more accurate inundation maps than official rain gauges, proving PWS a vital resource for urban flood mitigation. Building on the potential of crowdsourced data, the second part of this Dissertation addresses the need for efficient urban flood modelling. While advanced hydrodynamic models require extensive setup, simplified conceptual models offer a practical alternative. This research utilizes Safer_RAIN, a fast-processing DEM-based model founded on Hierarchical Filling-and-Spilling Algorithms (HFSAs) and assesses its applicability beyond its original scope by evaluating point-source inundation caused by drainage system overflows, ensuring reliable flood simulation for at-risk urban infrastructure. To address the limitations of HFSAs in predicting inundation extent and flow-path flooding, the third part introduces Kinematic Safer_RAIN. This enhanced model incorporates a kinematic runoff travel time distribution and a flow path inundation extension using the Height Above Nearest Drainage (HAND) approach and Manning’s equation. These modifications enable a physics-based representation of flooding that expands beyond the fixed boundaries of depressions and captures inundation along urban flow paths, overcoming the previous limitations where flooding was restricted to depression extents and flow paths remained dry. Tested in the Cottonwood Lake area (USA) and validated during real flood events in Pamplona (Spain), the algorithm shows comparable accuracy to 2D hydrodynamic models. Finally, the Dissertation discusses the model's potential and proposes future research directions for urban pluvial flood modelling.