文献类型：期刊文献

中文题名：中国东部沙地温带稀树草原的分布与制图

英文题名：Distribution and mapping of temperate savanna in the sandy lands of eastern China

作者：李晓雅[1,2,3] 王锋[1,2,3] 段涛[4] 杨凯捷[1,2,3] 杨斌[5] 王春梅[6] 田昕[7] 卢琦[1,2,3]

第一作者：李晓雅

通信作者：Wang, F[1];Lu, Q[1];Wang, F[2];Lu, Q[2];Wang, F[3];Lu, Q[3]

机构：[1]中国林业科学研究院生态保护与修复研究所,北京100091;[2]中国林业科学研究院荒漠化研究所,北京100091;[3]三北工程研究院,巴彦淖尔015200;[4]生态环境部卫星环境应用中心,北京100094;[5]中国科学院微电子研究所,北京100029;[6]中国科学院空天信息创新研究院,北京100094;[7]中国林业科学研究院资源信息研究所,北京100091

年份：2025

卷号：55

期号：1

起止页码：126-140

中文期刊名：中国科学:地球科学

外文期刊名：Scientia Sinica(Terrae)

收录：;EI(收录号：20245117533185);Scopus(收录号：2-s2.0-85211955050);WOS:【SCI-EXPANDED(收录号:WOS:001376273400001)】；北大核心:【北大核心2023】；

基金：We would like to acknowledge assistance from Professor Weiwei CONG from the Shenyang Agricultural University, Associate Professor Rina WU from the Chinese Academy of Forestry, and postgraduate students Yu JIANG, Yuxin ZHANG and Yifei CAI for their support during the field work. We would like to thank Professor Lu SEN from the Chinese Academy of Forestry for his suggestions regarding soil type analysis. This work was supported by the Chinese Academy of Forestry Science Fund (Grant Nos. CAFYBB2020QD002, CAFYBB2021MC002, and CAFYBB2023ZA009), the National Natural Science Foundation of China (Grant No. 32171875), the National Key Research and Development Program of China (Grant No. 2023YFF1304103), and the Open Bidding for Selecting the Best Candidates Project supported by National Forestry and Grassland Administration (Grant No. 202401).

语种：中文

中文关键词：温带稀树草原;无人机;高分卫星;木本植物覆盖度;草本植物覆盖度;机器学习

外文关键词：Temperate savanna; UAV imagery; GF-6; Fractional woody vegetation cover; Fractional herbaceous vegetation cover; Machine learning

分类号：S81

摘要：稀树草原广泛分布于全球各个区域,对调节陆地净初级生产力和陆地碳循环至关重要.由于稀树草原树草混合分布使其具有较高的空间异质性,当前的全球土地覆盖分类图多将其划分为森林或草原.中国温带稀树草原覆盖面积广泛,但缺乏对其空间格局和面积的认识.本研究基于中国温带稀树草原植被结构特征,参考联合国粮食及农业组织对稀树草原的定义,发展了一套融合无人机和卫星两种数据源的温带稀树草原的遥感制图新方法.利用覆盖主要稀树草原类型总面积1818ha的无人机高分辨率图像构建训练和验证数据集,集成分类和回归树算法与随机森林算法,估算了稀树草原上木本和草本植物的覆盖度,绘制了最小识别面积为0.5ha的中国东部科尔沁和浑善达克沙地温带稀树草原分布图.结果表明,新方法对稀树草原上木本和草本植物覆盖度估算均方根误差分别为10.69%和11.67%,决定系数分别为0.66和0.63.中国东部沙地温带稀树草原分布图精度评估Kappa系数为0.70,总体精度为0.88,平均精度为0.85,表明分布图具有较高的识别精度.中国东部沙地温带稀树草原集中分布于科尔沁沙地北部、浑善达克沙地中东部和大兴安岭南麓山地向平原过渡的区域,总面积为3.27×10^(6)ha.区域稀树草原类型主要为榆树稀树草原、栎树/槭树稀树草原和灌丛稀树草原.榆树稀树草原广泛分布于浑善达克沙地、科尔沁沙地东部和北部,呈条带状与流动沙丘相间分布;槭树/栎树稀树草原主要分布于科尔沁沙地北部,生长于低山丘陵;面积最大的温带稀树草原类型为灌丛稀树草原,主要分布于浑善达克沙地西部、科尔沁沙地中部和西部.本研究绘制了中国东部科尔沁、浑善达克沙地温带稀树草原空间分布图,提供了两大沙地温带稀树草原地理分布和面积的精确信息.数据成果可支撑中国温带稀树草原生态系统保护和利用、东部沙地荒漠化防治和生态系统恢复等相关科学研究和保护政策制定.研究方法为全球尺度稀树草原精细分类制图及稀树草原动态变化分析提供了新思路.
Savannas are distributed globally across various regions. They perform a significant socioeconomic role in regulating net primary productivity and the terrestrial carbon cycle. Owing to the high spatial heterogeneity induced by the mixed distribution of trees and grasses, current global land-cover classification maps often classify savannas as forests or grasslands. Temperate savannas in China cover a vast area, yet their spatial patterns and extent remain poorly understood. This study developed a novel remote sensing temperate savannas mapping method by integrating very high-resolution (VHR) unmanned aerial vehicle (UAV) and GF-6 satellite imagery based on the vegetation structure characteristics of temperate savannas using the FAO definition of savannas. Training and validation datasets were collected from VHR UAV imagery covering the main temperate savannah types with a ground area of 1818 ha. The fractional woody vegetation cover (FWVC) and fractional herbaceous vegetation cover (FHVC) in temperate savannas were estimated by integrating Classification and Regression Tree and Random Forest algorithms. The current distribution of temperate savannas in the Horqin and Otindag Sandy Lands of Eastern China was mapped to a unit area of 0.5 ha. The root-mean-square error (RMSE) values for the FWVC and FHVC models were 10.69% and 11.67%, and the R2 values were 0.66 and 0.63, respectively. The assessment indicated that the new temperate savanna map had high identification accuracy, with a Kappa coefficient of 0.70, overall accuracy of 0.88, and average accuracy of 0.85. Temperate savannas in the sandy lands of eastern China were primarily distributed in the northern Horqin Sandy Land, the central and eastern regions of the Otindag Sandy Land, and the transition zone from the southern slopes of the Greater Khingan Mountains to the plains, covering a total area of 3.27x106 ha. The main savannah types were elm, oak/maple, and shrub savannas. Elm savannas were widely distributed in the Otindag Sandy Land and the eastern and northern Horqin Sandy Land, arranged in stripes and distributed alternately with mobile dunes. Oak/maple savannas were mainly found in northern Horqin Sandy Land and grew on low hills. The largest temperate savannas were shrub savannas, which were primarily distributed in the western Otindag Sandy Land and the central and western Horqin Sandy Land. This study developed a detailed map of temperate savannas in Horqin Sandy Land and Otindag Sandy Land in eastern China, providing their precise area and geographical distribution. This map can support the conservation and utilization of temperate savannas, scientific research, and decisionmaking related to combating desertification and ecosystem restoration in the sandy lands of eastern China. This method provides a new vision for precisely mapping savannas on a global scale and analyzing savanna dynamics.