Maurizio Mazzoleni, Juan Carlos Chacon Hurtado, Leonardo Alfonso, Dimitri Solomatine
Thursday 2 july 2015
8:45 - 9:00h at Oceania Foyer (level 0)
Themes: (T) Flood risk management and adaptation, (ST) Early warning systems
Parallel session: 10L. Floodrisk - Early warning systems
Early warning systems have been used in the last decades to provide accurate flood forecast and reduce flood risk in urbanized areas. In order to reduce the intrinsic model uncertainty and improve the flood forecasting accuracy, different data assimilation techniques have been proposed to update model states and output as response of real-time observations from physical sensors. Traditionally, hydrological observations from such sensors have a well defined structure in terms of frequency, accuracy and implementation in data assimilation approaches. Nowadays, observations from low-cost sensors having variable space and temporal coverage and unpredictable accuracy in time and space are becoming more available. Such observations can be considered as anarchist since they have no rules in the information retrieving, in contrast to the common observations coming from physical sensors. However, the assimilation of anarchist observations has not been properly considered in hydrological application. The objective of this study is to propose an adaptive modelling framework for the assimilation of anarchist observations in hydrological models to improve flood forecasting accuracy. Synthetic experiments are carried out assuming different distributions of observations accuracy and location within the domain of the Brue basin. Models using data assimilation with and without anarchist observations are compared to evaluate the performance of the method. Preliminary results show how anarchist observations can be integrated in hydrological model in addition to standard streamflow observations to improve flood forecasting.