郁凡. 2003. 多光谱卫星图像降水强度场的分析[J]. 气象学报, 61(3):334-345, doi:10.11676/qxxb2003.032
多光谱卫星图像降水强度场的分析
ANALYSIS OF RAINFALL INTENSITY FIELD BY USING MULTI-SPECTRAL GMS IMAGERY
投稿时间:2002-07-16  修订日期:2003-04-11
DOI:10.11676/qxxb2003.032
中文关键词:  多光谱卫星图像  光谱特征空间  降水强度类属矩阵
英文关键词:Multi-spectral satellite imagery  Spectral feature scale  Rainfall intensity category matrix
基金项目:国家自然科学基金项目(41075007);国家重点基础研究发展规划项目:中国暴雨试验研究(G1998040910)
作者单位
郁凡 南京大学大气科学系, 南京, 210093 
摘要点击次数: 2224
全文下载次数: 2524
中文摘要:
      文中将单位特征空间归类方法应用于多光谱GMS卫星图像的降水强度场分析,该方法可交互式地进行多光谱卫星信息和地面实测降水的协同分析,准确划分各强度样本集群的光谱特征空间分布,为可靠确定各波段卫星测值与小时降水量之间的统计关系提供了一条可行的途径。为尽量减少分界点附近数据可能造成的不确定性和误差,文中首先按模糊集合论的隶属度原则,建立了模糊划分公式。按所在降水强度等级,通过对多维光谱空间的各单位特征空间内计算和比较晴空、多云(无雨)、小雨、中雨、大雨和暴雨6种情况发生的概率,经归一化处理后,分别建立相应的降水强度类属矩阵,为多光谱卫星图像降水强度场的分析确定了判识依据。就IR1-VIS降水强度类属矩阵而言,经500余个实测有雨样本的检验,其对暴雨、大雨、中雨和小雨等各强度等级有雨样本的实测命中率均在70%左右。近1400个有云样本降水强度等级判识的总准确率为73.88%。把样本仅分成无雨、中小雨和大、暴雨3个等级进行分析,实测命中率和分析成功率都显著提高,总准确率达到84.49%。IR1-WV,TIR1-IR2降水强度类属矩阵,各项指标虽然均略低于IR1-VIS降水强度类属矩阵,但对无雨、中小雨和大到暴雨的3个级判识,总准确率也能够达到75%。
英文摘要:
      In this paper, Unit Feature Space (UFS) Classification method is applied to analyse rainfall intensity field by using multi-spectral GMS satellite imagery. This method can interactively carry out coordinated analysis between measured rainfall amount per hour and multi-spectral satellite measuring information and can exactly segment every rainfall intensity sample cluster in spectral feature space, thus providing a feasible way to establish statistical relationship between satellite measurements and rainfall intensity grade. To reduce as much indetermi nateness and error that may occur as possible in classifying data around the division, an fuzzy separation formula was given on the membership principles of fuzzy set theory. A rainfall intensity category matrix is formulated based on the calculation and comparison of occurrence probabilities of clear sky, cloudy (rainless), light, moderate, heavy and hard rain in each UFS and that is just the decision rule for rainfall intensity field analysis by using multi-spectral GMS satellite imagery. Tests of more than 500 samples of rainfall occurrence show that the IRVIS rainf all int ensity category matrix yielded the accuracy of 70% or so for all categories of hard, heavy, moderate and light rain samples and total precision of 73.88% for measured rainfall intensity categ orization from nearly 1400 samples with cloud cover available. After the separation of the samples into three broad kinds, i.e., no rainfall, light-moderate and heavy-torrential rainf all, the measurement hit rate and analysis success ratio were considerably raised and total accuracy arrives at the precision of 84.49%. Though the accuracy of IR-WV or IR1-TIR2-IR1 rainfall int ensity category matrix is lower in a sort of way than that of IR-VIS rainf all intensity category matrix, the total precision is also up to 75% for three grade analyses of no rainfall, light-moderate and heavy-hard rainfall.
HTML   查看全文   查看/发表评论  下载PDF阅读器
分享按钮