【摘要】 针对构建用 SeaWinds 散射计数据反演海面风矢量的神经网络模型所面临的主要问题,结合 SeaWinds 散射计的几何观测特征,提出了适合 SeaWinds 散射计旋转圆锥扫描方式的先风向后风速的两步神经网络反演模型及相应算法,并采用两组不同的 L2A 和相应 L2B 数据及浮标数据对该神经网络反演模型进行了初步验证。实验结果证明了该神经网络反演模型的可行性。与最大似然估计(MLE)反演方法相比,该神经网络反演模型在能够保证反演精度的情况下,运行效率提高了约5倍,从而为海面风矢量的实时反演提供了可能性。更多还原

【Abstract】 Aiming at the main problems in establishing a neural network model for retrieval of the ocean surface wind vector from SeaWinds data,and in consideration of the SeaWinds scatterometer’ s geometry observation characteristics,the paper presents a neural network retrieval model and a corresponding algorithm,which are suitable to the conical scanning mode of the scatterometer.The retrieval of the model consists of two steps——the wind direction retrieval and the wind speed retrieval.Two sets of data,Level 2A and corresponding Level 2B,and buoy data were selected to validate the model pre- liminarily.The experimental results demonstrated the feasibility of the retrieval model.The retrieval efficency of the mod- el was approximately five times higher than that of the maximum likelihood estimation(MLE)method without obviously reducing the retrieval precision.Thereby,the model provides the possibility for real-time retrieval of ocean surface wind vectors.更多还原