Keywords: flood modelling, high performance computing, rainfall-runoff calibration, uncertainty modelling, disaster management system.

This paper describes an automatized online flood monitoring and prediction process used in the Floreon+ disaster management system and details modelling and simulation methods executed during the process. Such a system has to provide accurate results for the current state of the modelled area, its hydrologic parameters and meteorologic situation, so it is crucial to automatically calibrate the models in the system. Some of the dynamic inputs of these models still can contain some inacuraccies based on the methods of their acquisition (e.g. weather forecast models, rain gauge density and distribution). These errors are stochastic in nature and have to be processed by probabilistic uncertainty modelling methods. Both calibration and uncertainty methods execute a high number of basic model simulations that take a lot of computational time. This is in conflict with time constraints of the flood prediction process that has to provide its results as soon as possible while they are still relevant for early warning. Therefore we propose a parallel implementation of the calibration and uncertainty methods along with their deployment to an HPC infrastructure. As the Floreon+ system supports several models to lower the impact of their individual weaknesses, an abstraction framework is created to provide a single interface for all supported models that can be used universally by all methods. The functionality and scalability of the proposed HPC framework then is tested in experiments on the catchments in the Moravian-Silesian region in the Czech Republic.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £95 +P&P)