Nearshore wave prediction using Graph Neural Network at Darwin Harbour, Australia

发布日期:2024-05-11 阅读:427

报告人:王小华 教授 新南威尔士大学

邀请人:王云涛 研究员

时    间:5月14日(周二)10:00-11:30

地    点:1号楼1210会议室


卫星海洋环境动力学国家重点实验室为增进学术交流,促进实验室对外开放,组织系列“海星系列学术讲座”,邀请国内外知名专家与青年科研人员分享前沿科学动态,讨论最新学术成果,欢迎大家踊跃参加。

召集人:王云涛 研究员

会议时间:5月14日(周二)10:00-11:30

会议地点:1号楼1210会议室

报告人:王小华 教授(澳大利亚新南威尔士大学)

报告题目:Nearshore wave prediction using Graph Neural Network at Darwin Harbour, Australia

报告人简介:王小华(Wang Xiao Hua)毕业于中国海洋大学,获澳大利亚詹姆斯库克大学海洋学博士学位,澳大利亚新南威尔士大学科学学院终身教授,Estuarine, Coastal and Shelf Science和Limnology and Oceanography: Methods等期刊副主编,主要从事近海海洋观测、卫星遥感和数值模拟、沉积物输运动力学以及人类活动和气候变化影响下的近海动力问题等研究。王小华教授与我室合作多年,已在国际知名海洋学专业杂志上公开发表论文170余篇,与卫星海洋环境动力学国家重点实验室合作发表论文10篇左右。

报告摘要:Darwin Harbour (DH), Australia, is a flood-dominated estuary where the wave substantially influences sediment resuspension and transportation, especially in the outer harbour. Hence, the prediction of waves is crucial for coastal activities and management in DH. This paper presents a graph neural network (GNN) model to forecast wave characteristics in the nearshore zone of DH. The model was assigned for next-frame prediction of wave parameters such as significant wave height, peak period, wavelength, velocity, and wave period. This implies that the model had been configured to project future features by assessing parameters at a specific domain and timeframes. The GNN framework is intended to identify graph dependence via message passing between the nodes. Input of the model is the wave findings from 62-days simulation of the SWAN model. The study was carried out among 7194 nodes, and each node was linked to 5 neighbour nodes to forecast dependencies accurately. The data has been split into 80% and 20% for training and testing purposes. Furthermore, the significance of the number of hours as an input on anticipated outcomes was investigated. The results reveal that GNN model can replicate wave variables from physics-based model with mean-squared errors less than 0.14% and coefficients of r-squared more than 71%. Moreover, it demonstrates that increasing the number of hours for input and time steps for forecasting reduces the model’s performance. As such, the proposed GNN model can be useful for wave prediction and can be integrated with traditional coastal modelling to examine coastal phenomena.

  • 卫星海洋环境动力学国家重点实验室

    © 2021卫星海洋环境动力学国家重点实验室 版权所有.
    浙ICP备10040255号-4   流量统计   

  • 地址:杭州市保俶北路36号
    邮编:310012
  • 总机号码:+86-571-8196 3198
    传真:+86-571-8883 9374