基于SMAP卫星雷达资料的海冰密集度反演技术研究

Sea ice concentration retrieval based on SMAP radar data

SMAP是美国于2015年初发射的一颗卫星,搭载了L波段的雷达。它采用圆锥扫描方式,具有固定的入射角、较大的幅宽和千米级的分辨率,在海冰监测方面具有独特的优势。本文利用SMAP卫星雷达资料分别与德国Bremen大学海冰密集度产品和美国国家冰雪数据中心(NSIDC)海冰密集度产品建立3.125km和25km匹配数据集,分析了L波段雷达后向散射系数、极化比和归一化极化差与海冰密集度之间相关性,建立基于人工神经网络的海冰密集度反演算法。为了验证SMAP卫星雷达资料反演海冰密集度的精度,本文选择德国Bremen大学和美国冰雪数据中心发布的海冰密集度产品分别与SMAP海冰密集度产品进行对比分析,SMAP海冰密集度与Bremen海冰密集度的偏差为0.07、均方根误差为0.14;与NSIDC海冰密集度的偏差为0.04、均方根误差为0.18,这表明SMAP海冰密集度产品与现有业务化海冰密集度产品具有很好的一致性。

The Soil Moisture Active Passive（SMAP）was launched by National Aeronautics and Space Administration（NASA）in early 2015,which carried an L-band Radar with a conical scanning antenna.It has the constant incidence angle,the wide swath and the km-scale resolution.And thus it has a significant potential for observing sea ice.The paper establishes the 3.125 km matchup data set by using the University of Bremen sea ice concentration and SMAP radar data,and the 25 km matchup data set by using the National Snow and Ice Data Center（NSIDC）sea ice concentration and SMAP radar data.A sea ice concentration algorithm based on artificial neural network is proposed by the relation between sea ice concentration and HH-polarized NRCS,HH/HV polarized ratio,HHpolarized difference.The retrieved sea ice concentration is validated by using sea ice concentrations from NSDIC and the University of Bremen.The root mean square differences（RMSDs）and biases of sea ice concentration between SMAP Radar and the University of Bremen are 0.14 and 0.07,respectively.The RMSDs and biased of SMAP sea ice concentration versus NSDIC ones are 0.18 and 0.04,respectively.The results show that SMAP sea ice concentration is basically consistent with the operational sea ice concentration.