RAPIDS: Early Warning System for Urban Flooding and Water Quality Hazards
Chen, Albert S.
This paper describes the application of Artificial Neural Networks (ANNs) as Data Driven Models (DDMs) to predict urban flooding in real-time based on weather radar and/or raingauge rainfall data. A time-lagged ANN is configured for prediction of flooding at sewerage nodes and outfalls based on input parameters including rainfall. In the absence of observed flood data, a hydrodynamic simulator may be used to predict flooding surcharge levels at nodes of interest in sewer networks and thus provide the target data for training and testing the ANN. The model, once trained, acts as a rapid surrogate for the hydrodynamic simulator and can thus be used as part of an urban flooding Early Warning System (EWS). Predicted rainfall over the catchment is required as input, to extend prediction times to operationally useful levels. Both flood-level analogue and flood-severity classification schemes are implemented. An initial case study using Keighley, W Yorks, UK demonstrated proof-of-concept. Three further case studies for UK cities of different sizes explore issues of soil-moisture, early operation of pumps as flood-mitigation/prevention strategy and spatially variable rainfall. We investigate the use of ANNs for nowcasting of rainfall based on the relationship between radar data and recorded rainfall history; a feature extraction scheme is described. This would allow the two ANNs to be cascaded to predict flooding in real-time based on current weather radar Quantitative Precipitation Estimates (QPE). We also briefly describe the extension of this methodology to Bathing Water Quality (BWQ) prediction.
Machine Learning in Water Systems symposium: part of AISB Annual Convention 2013, University of Exeter, UK, 3-5 April 2013
Convention organised by the Society for the Study of Artificial Intelligence and Simulation of Behaviour (AISB), www.aisb.org.uk/
Machine Learning in Water Systems symposium, AISB 2013
Place of publication