| 摘要: |
| 本文利用神经网络的技术手段，针对Sentinel-1A二级波模式数据提出一种用于海浪有效波高（Hs）反演的模型——N_N模型。该模型在基于ERS2 SAR波模数据开发的双参数模型的基础上，加入经度、纬度、方位向截断波长（λc）、图像偏斜（skewness，skew）、图像峰度（kurtosis，kurt）、卫星平台距目标物的距离与卫星飞行速度之比（β）等其他参数信息，根据不同输入参数的组合，建立了14个模型用于Hs反演，旨在分析各参数对有效波高反演的影响。通过分析表明，14个N_N模型相关系数都在0.8以上。随着λc、β参数的加入，N_N模型性能均大幅上升，且λc参数对模型性能的改善作用更加明显，相关系数提升0.06左右，均方根误差（Root Mean Squared Error，RMSE）下降0.12m左右。另外，skew与kurt的加入也使N_N模型性能有所改善，RMSE下降0.03m左右，相关系数提升0.01左右。其中，N_N10模型效果最佳且性能最稳定，与欧洲中程天气预测中心（the European Centre for Medium-Range Weather Forecasts，ECMWF）数据对比，相关系数（CORR）达到0.905，散射指数（Scattering Index，SI）与RMSE最低，分别为18.74%、0.502m，与独立测量的浮标数据的相关系数达到了0.894。 |
| 关键词: 神经网络 有效波高 方位向截断波长 归一化雷达后向散射系数 |
| DOI：10.11693/hyhz20190900177 |
| 分类号:P7;TP3 |
| 基金项目:国家自然科学基金项目，41776197号；中国科学院大学优秀青年教师科研能力提升项目，Y95401N号。 |
附件 |
|
| INVERSION OF GLOBAL SIGNIFICANT WAVE HEIGHT BASED ON SENTINEL-1A |
|
MU Shan-Shan1, LI Hai-Yan1, WU Ming-Bo1,2
|
|
1.University of Chinese Academy of Sciences, Beijing 100049, China;2.Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Science, Beijing 100101, China
|
| Abstract: |
| By using the technique of neural network, a new neural network model a new neural network model (N_N model) was proposed for the inversion of wave effective wave height (HS) from Sentinel-1A level-2 wave model data. Based on the two-parameter model developed by ERS2 SAR wave mode data, the model was added with other parameters including longitude, latitude, the azimuth cutoff (λc), skewness, kurtosis, and the ratio (β) of the distance between satellite platform and the target to the satellite flight speed. The influence of each parameter on the inversion of significant wave height was analyzed using different combinations of input parameters, based on 14 models which were established for HS inversion. Results show that the correlation coefficients of all the 14 models were above 0.8. With the addition of λc and β parameters, the performance of the N_N model increased significantly, and the improvement effect of λc on the model performance was more obvious. The correlation coefficient increased by about 0.06, and RMSE decreased by about 0.12m. In addition, the addition of skewness and kurtosis also improved the performance of the N_N model as the RMSE decreased by about 0.03m, and the correlation coefficient increased by about 0.01. Among them, the N_N10 model had the best effect and the most stable performance. Compared with the ECMWF (European Centre for medium range weather forecasts), the correlation coefficient (CORR) was 0.905, and the scattering index (SI) and RMSE were the lowest, being 18.74% and 0.502m, respectively. The correlation coefficient with the independently measured buoy data reached 0.894. |
| Key words: neural network significant wave height azimuth cutoff normalized radar cross-section |
