摘要
以甘南州为研究区,参考2004年TM遥感影像目视解译分类图,评估了2004年MODIS土地覆盖产品MCD12Q1的分类精度。研究结果表明:1)MCD12Q1产品的IGBP、UMD、LAI/FPAR、PFT 4种土地覆盖分类图像中,UMD分类图像的总体精度最高,为61.03%,草地分类制图精度达91.61%,用户精度达61.53%。2)该产品分类误差主要为林地漏测误差和草地多测误差,其中LAI/FPAR图像的林地漏测误差和草地多测误差最高,分别达85.04%和38.90%。3)统计计算2001-2005年UMD分类图像的甘南州草地总面积,与甘南州统计年鉴的数据基本一致,能够反映甘南草地时空分布范围及面积的动态变化特征。
Gannan Prefecture was selected as research area to evaluate the classification accuracy of MODIS land cover product（MCD12Q1） by using visual interpretation result of TM image in 2004 as reference data.The results showed that： 1） Out of 4 types of MCD12Q1 product（IGBP,UMD,LAI/FPAR and PFT classification maps）,the UMD classification map had the highest overall accuracy（61.03%）,its producer＇s accuracy of grassland classification reached 91.61%,and the user＇s accuracy reached 61.53%.2） The main source of error in 4 types of land cover classification maps was commission error of forest and omission error of grassland.The commission error of forest and the omission error of grassland for LAI/FPAR classification map was the highest（85.04% and 38.90% respectively）.3） Grassland area calculated from UMD classification maps in 2001 to 2005 was basically close to the local statistical result,which could reflect the dynamic change of temporal and spatial distribution and area of grassland in Gannan.
出处
《草业科学》
CAS
CSCD
北大核心
2010年第9期11-18,共8页
Pratacultural Science
基金
教育部科技创新工程重大项目培育资金(708089)
国家高技术研究发展专项经费(2007AA10Z232)
