极地研究

极地研究 ›› 2016, Vol. 28 ›› Issue (1): 103-112.DOI: 10.13679/j.jdyj.2016.1.103

• 研究论文 • 上一篇    下一篇

基于Radarsat-2双极化数据的南极半岛冰盖冻融探测研究

王蒙1, 2 李新武1 梁雷1, 2 陆万雨3   

  1. 1中国科学院遥感与数字地球研究所, 北京 100094； 2中国科学院大学, 北京 100049；3国研信息科技有限公司, 北京 100010
  • 收稿日期:2014-12-25 修回日期:2015-02-08 出版日期:2016-03-30 发布日期:2016-03-30
  • 基金资助:

    国家自然科学基金（41076129）资助

STUDY OF SNOWMELT DETECTION ON THE ANTARCTIC PENINSULA ICE SHEET DERIVED FROM RADARSAT-2 DUAL-POL DATA

Wang Meng1,2, Li Xinwu1, Liang Lei1,2, Lu Wanyu3   

  1. 1Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China; 2University of Chinese Academy of Sciences, Beijing 100049, China; 3National Research Mdt InfoTech Ltd, Beijing 100010
  • Received:2014-12-25 Revised:2015-02-08 Online:2016-03-30 Published:2016-03-30

PDF (PC)

657

摘要： 南极冰盖的融化对全球海平面上升和气候环境变化具有重要影响, 合成孔径雷达（SAR）用于划分南极冰盖冰川带及冻融探测具有不可替代的作用。本文以南极半岛地区为例, 基于C波段星载SAR影像进行南极冰盖冻融探测方法研究。通过对于南极冰盖干雪带、渗浸带和湿雪带的后向散射特征的分析, 采用基于后向散射因子阈值的决策树分类划分冰盖冰川带。统计分析表明, 冰川带后向散射因子分布并不集中, 尤其是融化强烈时的湿雪带受融化程度影响很大, 与干雪带相近而不能仅从后向散射因子数值区分。为将冰盖的冰川带分类, 引入干雪带分布和海拔高度作为辅助信息, 分别发展了两种决策树分类方法并比较分析, 同时利用微波辐射计冰盖冻融探测结果和自动气象站数据做验证。结果表明利用双极化SAR数据的后向散射因子基于两种决策树分类都能够有效地划分冰川带并区分冻融状态, 实现高分辨率的冰盖冻融探测。

关键词: Radarsat-2, 双极化, ScanSAR, 南极, 冰盖冻融, 决策树分类

Abstract: Snowmelt in Antarctica has considerable impact on sea level rise and climate change. We investigated the detection of snowmelt on the Antarctic Peninsula using C-band spaceborne synthetic aperture radar imagery. Based on an analysis of the backscatter characteristics of dry, percolation, and wet snow, we used a decision tree classification to divide the ice sheet into zones. The statistical analysis demonstrated that the backscatter coefficients of snow zones, especially the wet snow zone, depend mainly upon melt level and do not have a centralized distribution. The wet snow zone in drastic melt is too similar to the dry snow zone to be distinguished using the backscatter coefficient alone. Therefore, we introduced the dry snow distribution and elevation into the classification, and compared the two decision tree methods. We verified the detection results using microwave radiometer and automatic weather station data. The results showed that the two presented decision tree classifications, derived from Radarsat-2 dual-pol data, were both efficient in determining glacier zone division and distinguishing snowmelt status and thus, are shown capable of achieving high-resolution snowmelt detection in Antarctica.

Key words: Radarsat-2, dual-pol, ScanSAR, Antarctica, snowmelt, decision tree classification