Hurricane Irma Simulation at South Florida Using the Parallel CEST Model
Hurricane Irma Simulation at South Florida Using the Parallel CEST Model
复制标题
使用并行 CEST 模型模拟南佛罗里达州的飓风艾尔玛
作者：
Yuepeng Li;Qiang Chen;D. Kelly;Keqi Zhang
本研究建立了海岸与河口风暴潮（CEST）模型的平行扩展，并应用该模型模拟了2017年飓风Irma在南佛罗里达州引发的风暴潮潮。对CEST中现有的平流算法进行了改进。这是通过引入高阶单调半拉格朗日平流来实现的。分布式内存并行化是通过消息传递接口（MPI）库实现的。因此，并行CEST模型可以在从多核笔记本电脑到大规模高性能计算（HPC）系统的各种机器上高效运行。能够在多个核心上运行CEST模型的主要优势是，相对较低的运行时间可以在高分辨率的网格上进行真实世界的风暴潮模拟，特别是在飓风登陆的地方。计算时间对于风暴潮模式预报来说至关重要，要在30分钟内完成模拟，并在下一次警报到来之前向用户提供结果。本研究对Irma飓风引发的风暴潮进行了模拟，模拟时间约为22 min，模拟时间为4天，并通过现场实测验证了模拟结果。进一步的效率分析表明，并行CEST模型在处理器数量不是很大的情况下可以实现线性加速。
In this study, a parallel extension of the Coastal and Estuarine Storm Tide (CEST) model is developed and applied to simulate the storm surge tide at South Florida induced by hurricane Irma occurred in 2017. An improvement is also made to the existing advection algorithm in CEST. This is achieved through the introduction of high-order, monotone Semi-Lagrangian advection. Distributed memory parallelization is developed via the Message Passing Interface (MPI) library. The parallel CEST model can therefore be run efficiently on machines ranging from multicore laptops to massively High Performance Computing (HPC) system. The principle advantage of being able to run the CEST model on multiple cores is that relatively low run-time is possible for real world storm surge simulations on grids with high resolution, especially in the locality where the hurricane makes landfall. The computational time is critical for storm surge model forecast to finish simulations in 30 min, and results are available to users before the arrival of the next advisory. In this study, simulation of hurricane Irma induced storm surge was approximately 22 min for 4 day simulation, with the results validated by field measurements. Further efficiency analysis reveals that the parallel CEST model can achieve linear speedup when the number of processors is not very large.