摘要
针对卫星观测数据难以满足很多统计学方法中要求的正态分布假设而导致误差的问题,以AMSR-E亮温数据为研究对象,对3种常用的正态转换方法的统计性能进行对比分析。首先对亮温数据的时变性进行分析,然后验证了Box-cox转换、Johnson转换和正态分位数转换(QNT)3种正态转换方法可以对亮温数据进行正态转换,最后详细比较了这3种方法转换后数据的正态化通过率和效果。结果表明,Box-cox转换方法是3种转换方法中最有效的。
With the application of statistical methods in satellite data simulation,it is often necessary to normalize the satellite data.The statistical properties of the three normal conversion methods which are applied to AMSR-E brightness temperature data are compared and analyzed.Firstly,the time variability of the bright temperature data is analyzed.Then,three normal conversion methods which are Box-cox conversion,Johnson conversion and normal quantile transform（QNT）are verified,which can be used to carry on the normal conversion of the bright temperature data.Finally,the normal pass rate and the effect of the three methods are compared in detail.The results show that the Box-cox conversion method is the most effective method.
出处
《遥感信息》
CSCD
北大核心
2017年第3期162-166,共5页
Remote Sensing Information
基金
北京市属高校青年拔尖人才培育计划项目(CIT&TCD201504047)
北京市教委科研计划面上项目(KM201410017008)
国家自然科学基金(41101311/D0106)
关键词
正态转换
亮温数据
统计性能
时变性
正态化通过率
normal conversion
brightness temperature data
statistical property
time variant
normalized pass rate
