欢迎访问《草业学报》官方网站,今天是

草业学报 ›› 2026, Vol. 35 ›› Issue (7): 1-14.DOI: 10.11686/cyxb2025297

• 研究论文 •    

中国半干旱牧区天然打草场空间格局遥感提取与分析

黄埔1,2(), 黄青1,2(), 辛晓平1,2   

  1. 1.北方干旱半干旱耕地高效利用全国重点实验室,北京 100081
    2.中国农业科学院农业资源与农业区划研究所,北京 100081
  • 收稿日期:2025-07-17 修回日期:2025-09-03 出版日期:2026-07-20 发布日期:2026-05-21
  • 通讯作者: 黄青
  • 作者简介:Corresponding author. E-mail: huangqing@caas.cn
    黄埔(2001-),男,四川内江人,在读硕士。E-mail: 2209913206@qq.com
  • 基金资助:
    国家重点研发计划(2021YFD1300500);国家重点研发计划(2023YFD1500102);现代农业产业技术体系建设专项资金(CARS-34)

Remote sensing extraction and analysis of the spatial pattern of natural cutting pastures in semi-arid pastoral areas of China

Pu HUANG1,2(), Qing HUANG1,2(), Xiao-ping XIN1,2   

  1. 1.National Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China,Beijing 100081,China
    2.Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081,China
  • Received:2025-07-17 Revised:2025-09-03 Online:2026-07-20 Published:2026-05-21
  • Contact: Qing HUANG

摘要:

天然打草场作为半干旱牧区草地资源的重要组成部分,对维持草地生态系统的稳定与可持续发展具有重要意义。然而,长期以来,我国天然打草场空间分布数据匮乏,现有数据难以准确反映其现状,制约了草地资源的科学管理与保护。为解决这一问题,本研究基于多源遥感数据融合技术(Sentinel-2与Landsat-8影像),引入地形、光谱指数和纹理特征,采用面向对象分割和支持向量机(support vector machine, SVM)算法,对半干旱牧区天然打草场进行了高精度遥感提取。研究结果显示,2020-2022年间半干旱牧区天然打草场总面积为794.92万hm2,其中内蒙古北部牧区占比最大(81.19%),温性草原为主要打草类型(61.87%)。固定打草场和机动打草场面积分别为477.12万和317.80万hm2,占比分别为60.02%和39.98%,空间分布呈现明显区域差异。遥感提取总体精度达84.76%,Kappa系数为0.69,与地面调查数据一致性较高,其中内蒙古北部牧区精度最高(87.16%)。本研究填补了近10年来天然打草场空间分布数据的空白,为草地资源的科学管理、生态保护与可持续利用提供了可靠的数据支持。

关键词: 半干旱牧区, 打草场, 多源数据融合, 面向对象分割, 机器学习模型, 遥感提取

Abstract:

As a vital component of grassland resources in semi-arid pastoral areas, natural cutting pastures play a significant role in maintaining the stability and sustainable development of grassland ecosystems. However, China has long faced a scarcity of spatial distribution data on natural cutting pastures, and existing data inadequately reflect their current status, hindering scientific management and conservation of grassland resources. To address this issue, this study employed a multi-source remote sensing data fusion approach (Sentinel-2 and Landsat-8 imagery), incorporating topographic features, spectral indices, and texture features. Using object-oriented segmentation and a support vector machine (SVM) algorithm, high-accuracy remote sensing extraction to identify natural cutting pastures in semi-arid pastoral areas was conducted. The results revealed that the total area of natural cutting pastures in semi-arid pastoral areas from 2020 to 2022 was 7.9492 million hectares, with the pastoral area in northern Inner Mongolia accounting for the largest proportion (81.19%), and temperate grasslands being the dominant type (61.87%). The areas of fixed cutting pastures and mobile cutting pastures were 4.7712 million and 3.1780 million hectares, accounting for 60.02% and 39.98% of the total area, respectively, showing distinct regional disparities in spatial distribution. The overall accuracy of remote sensing extraction reached 84.76%, with a Kappa coefficient of 0.69, indicating high consistency with ground survey data. The pastoral area in northern Inner Mongolia achieved the highest accuracy (87.16%). This study fills the gap in spatial distribution data of natural cutting pastures over the past decade and provides reliable data support for the scientific management, ecological conservation, and sustainable utilization of grassland resources.

Key words: semi-arid pastoral areas, cutting pastures, multi-source data fusion, object-oriented segmentation, machine learning model, remote sensing extraction