|
Over the past centuries, a number of severe coastal floods have destructed many places in the world. The mechanism leading potentially to coastal floods is well understood, given the configuration of the coastline and the bathymetry, the severity of the storm surge depends primarily on wind speed, wind direction and duration. The meteorological conditions are affected by the path and the velocity of the depression systems, moving across the Sea. When winds push water towards the coast, it tends to accumulate into what is commonly referred to as storm surge. If a particular high surge occurs together with a tidal maximum, both effects accumulate and serious flooding can result, depending on the coastal structure and their protection.
In the Netherlands, the accurate forecasting of storm surges is very important since large areas of the land lie below sea level. The warning system is also improved in this area to better prevention against flooding by the sea. The storm surge forecasting and warnings are made by the Dutch storm surge warning service (SVSD) in close cooperation with the Royal Dutch Meteorological Institute (KNMI). The forecasts for at least 6 hours ahead are required for proper closure of the movable storm surge barriers. Since the mid-1980s, these forecasts are based on a numerical hydrodynamic model called the Dutch Continental Shelf Model (DCSM). This model uses the forecasts from the meteorological high-resolution limited area model (HiRLAM) as the inputs. In early 1990s, Kalman filter was added to this system to improve the accuracy of the forecasts further by incorporating recent observations from tidal gauges. A number of significant improvements to increase the storm surge model performance have been implemented and tested, include: refining computational grids, calibrating the model, using a better numerical scheme and implementing data assimilation techniques (3D/4DVar and Kalman filter). However, some extended improvements are still needed. One extension would be concerning the development of an ensemble model framework for combining the storm surge forecasts of the European physically-based storm surge models for the North Sea from various institutions in the Netherlands, Denmark, UK, Norwegian, Belgium and Germany. In this work, data-driven models are utilized to ensemble their forecasts with the inclusion of the expert judgments. This allows for creating a "super" model with a high accuracy of forecasts. The other task is to build the univariate and multivariate chaotic models and other data-driven models based on computational intelligence techniques aimed at predicting storm surges. These models can be utilized to complement the existing European storm surge models resulting in a hybrid model. A number of open issues should be solved in order to improve the accuracy of the chaotic model, such as: finding true dynamical neighbors, optimizing the chaotic model parameters, assimilating the chaotic model with new observations. Such hybrid model is considerably able to improve the accuracy of the storm surge forecasts. The last task concentrates on developing the uncertainty prediction models that can predict the uncertainty (accuracy) of the forecasts produced by the European physically-based storm surge models for the North Sea, chaotic and data-driven models and their ensemble. Keywords Coastal flooding, storm surge forecasting, data-driven model, hybrid model, ensemble model, neural networks, nonlinear dynamics, chaos theory, uncertainty |
|