不同土地利用/覆盖数据在中国北方草地的精度评价

Accuracy evaluation of different land use or land cover data in grassland of northern China

草地是全球分布最广的植被类型之一,精确界定草地边界对于开展草地资源研究和管理具有重要意义。本研究以中国草地分布最集中的北方地区为研究区域,以野外调查获取的草地样点数据为主要评价依据,对来源于7种土地利用/覆盖数据的中国北方草地边界精度进行评价,包括全球30 m地表覆盖数据(GlobeLand30)、1∶1000000中国草地资源图(GLT80s)、中国土地利用现状遥感监测数据(LUCC2010)、中国土地覆盖数据(MICLCover)、全球陆地覆盖数据(Globcover2009)、全球土地覆盖数据(GLC2000)和MODIS土地覆盖数据(MCD12Q1)。结果表明:GLT80s划定的北方草地面积最大,与其他土地利用/覆盖数据划定的草地分布一致区域最多,基于采样点验证的精度也最高,数值为88.32%,而Globcover2009的精度最低,仅为29.31%,MCD12Q1和GLC2000的采样点验证精度分别为83.74%和83.45%,其他草地图的采样点验证精度则介于73.42%～78.97%;分省统计分析显示,在青海省MCD12Q1的精度最高,数值高达94.23%,GLC2000和GlobeLand30的精度在90%以上,而在新疆、西藏、内蒙古、甘肃则均为GLT80s的精度最高;与GLT80s对比分析显示,其他数据源与其分布不一致的主要原因是将GLT80s中的温性荒漠类草地部分划为裸地,面积占分布不一致区域总面积的22.59%～42.87%。建议在开展有关北方草地资源及功能评价时,除青海省优先选用MCD12Q1数据外,其他省份优先选用GLT80s数据,但要特别注意其在温性荒漠类草地区域划分的准确性。

Grassland is one of the most widely distributed vegetation types in the world. A precise definition of the grassland boundary is essential for grassland resources research and management.We assessed the boundary accuracy of grassland in northern China,where was characterized by the most concentrated distribution of grassland in China. Based on the grassland sampling point data obtained from field survey,seven kinds of land use/cover data were examined,including the30-m global-land cover data( GlobeLand30),1 ∶ 1000000 grassland resource map( GLT80s),remote sensing monitoring data of land use status in China( LUCC2010),multi-source integrated Chinese land cover map( MICLCover),GlobCover land cover map( Globcover2009),global land use data( GLC2000),and MODIS land cover data( MCD12Q1). The results showed that GLT80 s had the largest area of grassland in northern China and was the most consistent with the other six land use/cover data. GLT80 s also had the highest accuracy through sampling point verification,with a value of 88. 32%. Globcover2009 had the lowest accuracy,with a value of29.31%. The accuracy of MCD12Q1 and GLC2000 was 83.74% and 83.45%,respectively. The values of other three land use/cover data ranged from 73.42% to 78.97%. Results of statistical analysis for different provinces showed that MCD12Q1 in Qinghai had the highest accuracy( 94.23%),and the accuracy of GLC2000 and GlobeLand30 was also above 90% respectively.However,GLT80 s always had the highest accuracy in other provinces,including Xinjiang,Tibet,Inner Mongolia,and Gansu. The main reason for the difference of grassland distribution between GLT80 s and other data was that the temperate desert grassland in GLT80 s was generally classified as bare ground in other data,which occupied 22.59%-42.87% of the total inconsistent distribution area. Therefore,we suggested that MCD12Q1 should be given preference in Qinghai for assessing the grassland resources and function,while GLT80 s should be selected in other provinces,but special attention should be paid to the accuracy of its division in temperate desert grassland.