摘要
海面风速是海洋环境的重要参数,微波辐射计是卫星监测海面风速的重要手段。通过微波辐射计SSM/I(Special Sensor Microwave/Imager)亮温与浮标实测风速建立的匹配数据集,利用人工神经网络构建海面风速反演模型。比较不同模型的反演效果,得出七通道单参数神经网络模型SANN(Single-parameter Artificial Neural Network)反演的效果和浮标实测风速较为接近,均方根误差RMSE(Root Mean Square Error)为1.40m/s。因此选择该模型反演全球的月平均风速,并将反演结果与NOAA产品风速比较。结果表明:两者在整体分布和纬度平均上非常接近,均方根误差为1.03m/s。可见,该算法用于海面风速反演还是可行的。研究亮点:提出用微波辐射计SSM/I刈辐型亮温与浮标实测数据进行匹配,可以减小人为误差。比较不同神经网络模型,发现七通道单参数模型效果要优于其他结构模型,然后选择该模型反演全球风速,效果较好,因此在不需要深究海面微波发射和传输的微观机制的前提下,可以为以后的微波辐射计反演海面风速提供参考。
The sea surface wind speed is an important parameter of marine environment and satellite microwave radiometer is an important tool to monitor this parameter. In this paper, a model for retrieving the sea surface wind speed is developed using the artificial neural network （ANN）, through the data sets generated between the microwave radiometer SSM/I brightness temperatures and the in-situ buoy measurements. By comparing the retrieval results of different models, it is concluded that the result of the seven-channel SANN retrieval model is closer to the buoy measured wind speed with the root mean square error （RMSE） of 1.40m/s. Therefore, this model is chosen to retrieve the global monthly-average wind speed, and the retrieval results are compared with the NOAA products. The results show that, both are very close in the overall and latitude-average distribution with the RMSE of 1.03 m/s. It can be seen that the algorithm for the sea surface wind speed retrieval is feasible.
出处
《上海海洋大学学报》
CAS
CSCD
北大核心
2012年第1期124-131,共8页
Journal of Shanghai Ocean University
基金
上海市高校选拔培养优秀青年教师科研专项基金(SSC10008)
关键词
SSM/I
反演
人工神经网络
应用
sea surface wind speed
SSM/I
retrieval
artificial neural network
application
