基于深度学习算法的大气—海浪—海洋耦合模型风暴潮模拟研究

The South China Sea (SCS) is one of the most serious areas affected by the typhoon. It is of great scientific and practical significance to simulate the characteristics of wave and storm surge under the influence of extreme events such as typhoons in this area. Most of the existing studies use numerical or empirical methods to simulate storm surge. The storm tide forecast model is built based on the WRF, SWAN, FVCOM model and recurrent neural network (RNN) technology. The academic issues of common concern in the areas of coastal ocean disaster prevention are studied. (1) A real time two-way coupled atmosphere-wave-ocean model in the South China Sea will been established. The relationship between typhoon wave and storm surge in near shore and the shallow water area is quantified to improve the simulation accuracy of typhoon wave and storm surge in the near shore area of South China Sea. (2) Based on recurrent neural network, a storm surge forecast model by the coupled atmosphere-wave-ocean model and the deep learning technology will been constructed. The typical sample data from 1949 to date is used for training and testing. (3) In view of the occurred and developing typhoons, the model is used to predict the possible track and intensity of the typhoon rapidly, in real time and in high efficiency. The accurate simulation of the distribution characteristics of typhoon wave and storm surge in the South China Sea provide a scientific basis for disaster prevention and mitigation in the regional ocean engineering.

南中国海是受台风影响最严重的地区之一，模拟该地区台风等极端事件影响下的海浪和风暴潮特征，具有重要的科学价值和现实意义，现有研究多采用数值计算或经验方法进行风暴潮模拟。本项目基于大气模型WRF、海浪模型SWAN、海洋模型FVCOM，结合RNN深度学习技术，构建基于深度学习的大气—海浪—海洋耦合风暴潮预报模型。研究解决海洋工程防灾减灾领域共同关心的学术问题：(1)建立南中国海大气—海浪—海洋实时双向耦合数值模型，量化分析近岸浅水区域台风浪与风暴增水的影响关系，提高南中国海近岸地区台风浪和风暴潮模拟精度；(2)利用建立的耦合模型，结合RNN技术，构建基于深度学习的风暴潮数值预报模型，选取1949年至今典型样本数据进行循环训练、验证；(3)针对已发生和发展中的台风，利用模型对台风可能路径及强度进行快速、实时、高效预报，准确模拟台风浪和风暴潮的分布特性，为地区海洋工程防灾减灾提供科学依据。

本项目针对现有研究对台风影响下大气、海浪、海洋三者之间存在强烈交换的动力过程描述存在缺陷这一科学问题，深入分析台风影响期间大风、海浪和风暴潮等因素的相互作用机制，采用中尺度气象模型、海浪模型和海洋模型，构建了一套大气—海浪—海洋实时双向耦合模型，并成功应用于南中国海风暴潮模拟研究，取得了较好的效果。研究成果对理解极端条件下海洋、海岸动力过程和开发耦合模拟方法具有一定的参考价值。依托该项目，以第一作者或通讯作者身份发表了标注项目资助及编号的学术论文12篇，包括在Ocean Engineering上发表3篇、在Ocean Science发表2篇、在Acta Oceanologica Sinica和《科学通报》各发表1篇。发表的12篇论文中有8篇被SCI论文收录、2篇被EI收录，其中1篇入选ESI高被引论文；受邀参编英文著作1部；授权软件著作权4项；参加国际学术会议1次、国内学术会议3次，参加现场调研6次；在本项目支撑下，博士学位论文获评为全国高等学校水利类专业优秀博士论文和湖南省优秀博士论文。