Computing, a powerful tool for improving the parameters simulation quality in flood prediction

Abstract

Floods have caused widespread damage throughout the world. Modelling and simulation provide solutions and tools which enable us to forecast and make necessary steps toward prevention. One problem that must be handled by physical systems simulators is the parameters uncertainty and their impact on output results, causing prediction errors. In this paper, we address input parameter uncertainty toward providing a methodology to tune a flood simulator and achieve lower error between simulated and observed results. The tuning methodology, through a parametric simulation technique, implements a first stage to find an adjusted set of critical parameters which will be used to validate the predictive capability of the simulator in order to reduce the disagreement between observed data and simulated results. We concentrate our experiments in three significant monitoring stations, located at the lower basin of the Paraná River in Argentina, and the percentage of improvement over the original simulator values ranges from 33 to 60%.