HPC For DRM - Operational Flood Management In Urban Environment

A regional flood warning system based on a combination of data processing, modelling and communication tools is proposed. The system is founded on a common framework MIKE CUSTOMISED, giving great flexibility in tailoring solutions as needed. Where more conventional flood warning systems focus mainly on discharge predictions in the main rivers the proposed system considers the whole catchment area – flood plain, as well as tributaries. Local floods on smaller streams and tributaries may cause high damages, particular in urban areas. Such cases require timely reasonably accurate forecasts for proper decision making. The MIKE SHE modelling system is used to ensure simulation of flood danger map. It also provides runoff hydrographs for the detailed hydrodynamic models of the river channel network and flood plains based on 1D / 2D approximations (MIKE FLOOD). Generated flood maps are post processed and ported to the required forms and delivered via communication channels to users. Dissemination of results is done through web pages automatically maintained by the system. Whole forecast simulation should be run in frequency of tens of minutes, which is computational power and transmission capacity demanding. Adaptation of whole system to High Performance Computer solutions (HPC) is on-going issue. This allows also parallel variant computation and real-time probabilistic assessment. IT4Innovations National Supercomputing Centre is a research institute at the VŠB – Technical University of Ostrava. This centre provides perfect platform for applications of HPC for Disaster Risk Management (DRM) and its new dynamic development for real life applications, utilised for life protection and damage minimizing. The paper contributes both to the theory of application of HPC for standard hydrodynamic modelling but also to a real life application. A pilot operational Flood Risk Mapping project is developed for the capitol city of the Czech Republic