An Extreme Learning Machine Application in Geophysical Modeling for Scatterometer
Scatterometer observations can provide accurate and spatially wind information near the sea surface which have proven significant for the forecasting of dynamical weather,such as tropical cyclones.One important step for the wind retrieval based on scatterometer observations is using the geophysical model function (GMF) which depicts the relationship between the normalized radar cross-section measurements and the sea surface wind speed,wind direction,radar parameters and environmental parameters.Since the establishment of strict theoretical geophysical model function is very difficult,at present most of the models are empirical models which are built by statistical methods.In this paper,we use a method called Extreme Learning Machine (ELM) to develop a unified GMF respectively using the simulated training data-set generated by the empirical GMF-CMOD5.N and the wind data retrieved from the advanced scatterometer (ASCAT).And we compare our method with the neural network approach.The results indicate that the ELM method shows a good inversion result with higher accuracy and faster training speed compared with the neural network technique.This new method could provide a novel feasible way for sea surface wind inversion.
ELM scatterometer neural network geophysical model function
Boheng Duan Weimin Zhang Chengzhang Zhu
College of Computer,National University of Defense Technology,Changsha,China,410073
国内会议
济南
英文
1-11
2014-10-16（万方平台首次上网日期，不代表论文的发表时间）