FY-2红外云图中强对流云团的短时自动预报算法  

Short-term automatic forecast algorithm of severe convective cloud identification using FY-2 IR images

作　　者：刘延安[1] 魏鸣[2] 高炜[1,3] 李南[4] 

机构地区：[1]华东师范大学地理信息科学教育部重点实验室,上海200062 [2]南京信息工程大学气象灾害省部共建教育部重点实验室,江苏南京210044 [3]美国科罗拉多州立大学自然资源生态实验室,美国柯林斯堡80523 [4]南京信息工程大学大气物理学院,江苏南京210044

出　　处：《遥感学报》2012年第1期79-92,共14页NATIONAL REMOTE SENSING BULLETIN

基　　金：国家重点基础研究发展计划(编号:2010CB951603);上海市科委世博科技专项(编号:10DZ0581600);江苏省普通高校研究生科研创新计划(编号:CX09B_227Z);华东师大优博培育行动计划(编号:PY2011010)~~

摘　　要：目前气象预报业务中,预报员主要借助卫星云图,定性判断云团的移动趋势,缺乏形式化的定量评判方法。本文基于FY-2C与FY-2D的高时间分辨率的近红外影像(10.3—11.3μm),采用亮温和面积阈值方法进行云团识别,然后根据最大相关系数云团匹配技术进行追踪,系统地实现强对流云团的自动临近预测。实验结果表明,本文提出的最大相关系数追踪比传统的交叉相关系数法具有更高的匹配精度和运行效率,而且研究发现云团质心外推明显优于最低亮温外推,平均亮温、面积、圆形度对云团的分裂合并有较好的指示作用,经列联表法检验,本文提出的自动识别追踪技术具有较高预测精度和预测时效,并且为卫星云图业务化应用提供了定量科学依据。The movement of clouds is qualitative analyzed by forecasters with satellite images currently,which is,however,lack of objectivity and quantitativity.In this paper,based on the stationary satellite infrared（IR） channel（10.3—11.3 μm） images of FY-2C and FY-2D with the time resolution of 15 minutes,brightness temperature（BT） and area threshold are selected to identify the severe convective cloud（SCC）.We then use the SCC matching algorithm of maximum correlation to track the shorttime automatic prediction of SCC systematically.The experiment results show that the tracking method proposed in this work has higher matching accuracy and efficiency compared with the traditional cross-correlation approach.The cloud center of gravity（CG） extrapolation is markedly superior to the minimum temperature,and the mean temperature,area and roundness all have better indications to the cloud split and merge.Tested by contingency table,the automatic identification and tracking technology has high prediction accuracy and timeliness.In addition,the research of this paper provides a scientific basis for the objective and quantitative application of satellite images to SCC short-time prediction in operation.

关 键 词：FY-2C和FY-2D 强对流云团 云团阈值识别 最大相关系数追踪算法 自动临近预测技术 

分 类 号：P456.1[天文地球—大气科学及气象学]