S.S. Kosukhin – Senior Research Scientist, National Research University of Information Technologies, Mechanics and Optics
S.V. Kovalchuk – Ph. D. (Eng.), Senior Research Scientist, National Research University of Information Technologies, Mechanics and Optics
A.V. Boukhanovsky – Dr. Sc. (Eng.), Professor, National Research University of Information Technologies, Mechanics and Optics
Storm surges are caused primarily by atmospheric pressure and are characterized by no periodicity and significant rise of water level. The strongest and most frequent storm surges in Russia are observed in St. Petersburg (the eastern part of the Gulf of Finland). To protect St. Petersburg from surge floods a flood protection barrier was built in 2011. Flood Warning Systems (FWS) was developed to support the management of barrier system in the cases of flood danger and during the flood situation. Its main objective is to predict the changes of water level in the Gulf of Finland. The maintenance of the FWS involves regular analysis of the quality and correctness of its parts (mathematical models, data sources, etc.) which implies the necessity of organizing and conducting simulation experiments to process and analyze various operating scenarios of FWS. Thereby, special software tools to perform are required. An experience in development of the problem-solving environment (PSE) for simulation experiment management in the storm surges prediction domain is presented in the article.
The meaningful part of the PSE are computing services which are based on numerical hydrometeorological models and data processing packages which are installed on the high-performance resources managed by the cloud computing platform CLAVIRE. The services are provided with the use of the description of the applied software in a special language EasyPackage, which allows to specify an abstract package calling and to generate a web-based user interface automatically. To support the study of special (non-general) scenarios of flood situations and to allow the user to implement own calculation algorithms, the solvers are included in the PSE. External data sources (e.g. water level measurements, external atmospheric model results, reanalysis data sets) are also available in the form of services. The user also has an opportunity to create own data sets based on the results of processing and simulations. A separate group of the PSE services are software blocks which reproduce the basic units of the FWS. Their purpose is to emulate the operational mode of the FWS in order to measure its performance, to analyze its possible failures and to evaluate its various upgrade scenarios. The PSE also provides an opportunity to use various specialized technical equipment (e.g. touch table, tablet PCs, smart phones etc.) which offers more convenient way to analyze and interpret the results of calculations.
The advantages of the described PSE for simulation experiments management are illustrated on the basis of the following specific problems: multi-model ensemble forecasting, uncertainty and sensitivity analysis of water level forecasts and barrier gates maneuvering plans, analysis of the city inundations scenarios. In this case, the feasibility of the presented PSE is due not only to its functionality and ergonomics, but also to its computational efficiency. The analysis showed that the overhead costs caused by the work of CLAVIRE platform, are insignificant in comparison with the time to solve these computing tasks.
Thus, we have demonstrated the possibility to organize the infrastructure for computational experiments required to support the simulation and forecasting of complex systems (for example, FWS in St. Petersburg) in the form of PSEs using CLAVIRE platform.