Storm surge is often the marine disaster that poses the greatest threat to life and property in coastal areas. Accurate and timely issuance of storm surge warnings to take appropriate countermeasures is an important means to reduce storm surge-related losses. Storm surge numerical models are important for storm surge forecasting. To further improve the performance of the storm surge forecast models, we developed a numerical storm surge forecast model based on an unstructured spherical centroidal Voronoi tessellation (SCVT) grid. The model is based on shallow water equations in vector-invariant form, and is discretized by Arakawa C grid. The SCVT grid can not only better describe the coastline information but also avoid rigid transitions, and it has a better global consistency by generating high-resolution grids in the key areas through transition refinement. In addition, the simulation speed of the model is accelerated by using the openACC-based GPU acceleration technology to meet the timeliness requirements of operational ensemble forecast. It only takes 37 seconds to simulate a day in the coastal waters of China. The newly developed storm surge model was applied to simulate typhoon-induced storm surges in the coastal waters of China. The hindcast experiments on the selected representative typhoon-induced storm surge processes indicate that the model can reasonably simulate the distribution characteristics of storm surges. The simulated maximum storm surges and their occurrence times are consistent with the observed data at the representative tide gauge stations, and the mean absolute errors are 3.5 cm and 0.6 h respectively, showing high accuracy and application prospects.
Figure 4. The resolution of the SCVT grid used in the simulation transitions from 20 km offshore to 2 km nearshore.
Figure 3. Six typical typhoons that made landfall in China. The star (★) is the location of the tide gauges.
Figure 5. Maximum storm surges during typhoon-induced storm surge processes.
Figure 6. Storm surge processes at the representative stations. The black scatter shows the observed values, and the green line and red line show the simulated values of ADCIRC model and our model respectively (unit: cm).