冰川冻土 ›› 2016, Vol. 38 ›› Issue (1): 145-158.doi: 10.7522/j.issn.1000-0240.2016.0017

• 寒区科学与技术 • 上一篇    下一篇


沙依然·外力1,2, 毛炜峄3   

  1. 1. 南京信息工程大学 应用气象学院, 江苏 南京 210044;
    2. 新疆维吾尔自治区气候中心, 新疆 乌鲁木齐 830002;
    3. 中国气象局 乌鲁木齐沙漠气象研究所, 新疆 乌鲁木齐 830002
  • 收稿日期:2015-11-08 修回日期:2016-01-09 出版日期:2016-02-25 发布日期:2016-05-30
  • 作者简介:沙依然·外力(1966-),男,新疆阿勒泰人,高级工程师,南京信息工程大学在职博士研究生,主要从事积雪遥感及地表参数反演研究
  • 基金资助:

A research on the method of deriving high precision snow parameters from AMSR2 passive microwave remote sensing data

Sayran Wayli1,2, MAO Weiyi3   

  1. 1. School of Atmospheric Physics, Nanjing University of Information Science & Technology, Nanjing 210044, China;
    2. Climatec Center of Xinjiang Uygur Autonomous Region, Ürümqi 830002, China;
    3. Institute of Desert Meteorology, China Meteorological Administration, Ürümqi 830002, China
  • Received:2015-11-08 Revised:2016-01-09 Online:2016-02-25 Published:2016-05-30

摘要: 以新疆为研究区域建立了被动微波遥感积雪深度高精度反演模型,采用高空间和时间分辨率AMSR2被动微波遥感数据(2012年11月-2015年3月逐日数据),结合研究区域海拔高度、坡度、坡向、沙漠,荒漠和地表粗糙度等地形、地貌特征,考虑冰川、水体、林地等地表覆盖类型和不同季节的新雪、干雪和湿雪等积雪属性的微波辐射特征,以决策树阈值法为基础,通过采集样本分类建立起多种雪深判识阈值,在此基础上建立AMSR2高精度积雪深度反演综合模型,分类分析不稳定积雪和冰川信息,从而实现雪深在60cm以内的积雪深度AMSR2反演的主要原理、思路及方法,并对模型的反演结果跟台站实测或者野外观测积雪值以时间和空间角度进行检验.结果表明:该综合模型能够定量判识研究区域复杂地形地貌条件下的1~60cm积雪厚度,检验的复相关系数为0.74~0.88,均方根误差为2.92~6.14cm,平均绝对偏差指数为3~4cm,雪深误差<5cm的精度为91%~94%,雪深误差<2.5cm的精度为81%~87%.

关键词: AMSR2, 积雪辐射特性, 雪深判识模型, 复杂地形地貌

Abstract: This study aims to establish a model of deriving high-precision snow parameters from passive microwave remote sensing data, taken Xinjiang Uygur Autonomous Region as a case. In this paper, firstly high-resolution spatial and temporal data in the research region by means of AMSR2 passive microwave remote sensing were collected from November, 2010 to March, 2015. Then terrain features, such as altitude, slope, aspect, desert and surface roughness, had also been investigated. Other factors, such as variation of surfaces ranging from glacier, paddy field and forest, and microwave radiations of new snow, dry snow and wet snow in various seasons had also been taken into account. After that, a variety of threshold values to assess snow depth were worked out by classifying the collected samples. Finally, a deriving model of snow depth from AMSR2 high-precision sensing data was set up on the basis of decision tree threshold method. The model was then adopted to analyze and classify unstable snow cover and glaciers to establish a deriving method from AMSR2 data targeting at the snow depth of no more than 60 cm. The result of the test was reexamined with those data which were collected from stations or expeditions. The research approves the workability of the new model in identifying the snow depth ranging from 1 to 60 cm in the areas with complicated landforms, with multiple correlation coefficients ranging from 0.74 to 0.88, root mean square error ranging from 2.92 to 6.14 cm, the mean absolute deviation index ranging from 3 to 4 cm. When the snow depth error less than 5 cm, the accuracy of the test ranges from 91% to 94%; when the snow depth error less than 2.5 cm, the accuracy ranges from 81% to 87%.

Key words: AMSR2, snow radiation, snow depth assessing model, varied topography


  • P426.63+5