Sea Winds散射计海面风场神经网络建模研究  

Research on Ocean Surface Wind Field Modeling Using Neural Network for Sea Winds Scatterometer

作　　者：解学通[1] 方裕[1] 陈克海[2] 黄舟[1] 陈斌[1] 

机构地区：[1]北京大学地球与空间科学学院,北京100871 [2]中山大学遥感与地理信息工程系,广东广州510275

出　　处：《地理与地理信息科学》2007年第2期12-17,共6页Geography and Geo-Information Science

基　　金：国家863计划项目(2002AA134100);985工程项目(105203200400006)

摘　　要：根据Sea Winds散射计只有两个入射角和两种极化方式的特点,利用其L2A数据和F291海上浮标数据,针对传统建模方法的不足和限制,借助神经网络建立了一个两种极化方式下统一的神经网络地球物理模型函数。该模型的主要特点是建模风矢量全部取自海上浮标测量数据,因而所用风矢量更加客观准确。通过与Qscat-1模型的比较和L2B与浮标风速之间的偏差统计分析,证明了该神经网络模型的有效性,并发现Qscat-1模型存在一定的系统性偏差。SeaWinds is the most advanced spacebome scatterometer at present, and it adopts conically scanning operation mode for the first time. For SeaWinds there are only two incidence angles with different polarization respectively. Based on this characteristic of SeaWinds, using some Level 2A data of SeaWinds and corresponding F291 buoy data, according to the shortage and limitation of the traditional modeling method, an unified neural network geophysical model function for both two polarizations is established in this paper. The main feature of this model is that wind vectors used in modeling are all from buoys,so the wind vectors are more objective and accurate. By the comparison made between neural network and Qscat - 1 model, and by the bias statistics made between L2B and buoy wind vectors, the established model is validated and analyzed. The results demonstrate the validity of neural network model and discover the possible systematic bias of Qscat-1 model.

关 键 词：SEAWINDS散射计 神经网络 地球物理模型函数(GMF) 后向散射系数 

分 类 号：P733[天文地球—物理海洋学] TP183[天文地球—海洋科学]