Towards identification of critical rainfall thresholds for urban pluvial flooding prediction based on crowdsourced flood observations

Abstract. Urban drainage systems are challenged by both urbanization and climate change, intensifying urban pluvial flooding impacts. Urban pluvial flooding impacts can be reduced by improving infrastructure and operational urban flood management strategies. This study investigated the relation between heavy rainfall and urban pluvial flooding in Rotterdam with the aim to identify parameters and thresholds that can be used as predictors of urban pluvial flooding. The focus of the investigation was on an area of 16 km². Datasets for this research included historical crowdsourced flooding reports, overflow pumping volumes, open spatial data and rainfall data at different temporal and spatial resolutions. Threshold values, (which can be used as part of early warning systems), were derived from historical flooding data and rainfall depths over sub daily durations. Threshold values of rainfall depth were found to be 6 mm (±3 mm) in 15 min and 11 mm (±6 mm) in 60 min. Surprisingly, the derived thresholds are only approximately half of the theoretical drainage system design capacity. Furthermore, a threshold value of 70 % (±4 %) imperviousness was found above which flooding incidents significantly increase. Results also suggested a strong dependence on spatial aggregation scale, as it highly influences correlation coefficients and parameter density values. Elevation differences did not appear to contribute to urban pluvial flooding, based on a flow path analysis for the study area. Finally, we showed that antecedent rainfall does not explain additional variance in reports, meaning it is not an influential urban pluvial flooding predictor, at least not on a daily temporal resolution.