湖北农业科学 ›› 2024, Vol. 63 ›› Issue (1): 199-205.doi: 10.14088/j.cnki.issn0439-8114.2024.01.036

• 信息工程 • 上一篇    下一篇

农牧交错地带撂荒地遥感识别研究——以青海省海东市乐都区为例

叶鹏帅1, 杨海镇1, 马涛2, 胡碧霞2, 包喜文3, 赵之重2   

  1. 1.青海民族大学政治与公共管理学院,西宁 810007;
    2.青海大学农牧学院,西宁 810016;
    3.东北农业大学公共管理学院与法学院,哈尔滨 150030
  • 收稿日期:2023-03-17 出版日期:2024-01-25 发布日期:2024-02-05
  • 通讯作者: 赵之重(1965-),男,教授,博士,主要从事耕地资源监测与评价研究,(电子信箱)1989990003@qhu.edu.cn。
  • 作者简介:叶鹏帅(1998-),男,河南驻马店人,在读硕士研究生,研究方向为耕地资源监测与评价,(电话)15539686518(电子信箱)1796220951@qq.com。
  • 基金资助:
    青海省重点研发与转化计划项目(2022-QY-225)

Research on remote sensing identification of abandoned farmland in agricultural and animal husbandry interzone:Taking Ledu District, Haidong City, Qinghai Province as an example

YE Peng-shuai1, YANG Hai-zhen1, MA Tao2, HU Bi-xia2, BAO Xi-wen3, ZHAO Zhi-zhong2   

  1. 1. College of Politics and Public Administration, Qinghai Minzu University, Xining 810007, China;
    2. College of Agriculture and Animal Husbandry, Qinghai University, Xining 810016, China;
    3. School of Public Administration and Law, Northeast Agricultural University, Harbin 150030, China
  • Received:2023-03-17 Online:2024-01-25 Published:2024-02-05

摘要: 为了实现对耕地及时、准确的识别,借助遥感技术对农牧交错地带撂荒地进行识别提取,摸清撂荒地的空间分布特征。基于谷歌地球引擎(Google earth engine,GEE)平台,调用研究区Sentienel-1和Sentienel-2遥感影像并进行预处理,采用随机森林算法开展研究区土地利用分类研究,并通过GEE平台获取研究区2017—2022年NDVI月最大值合成数据,结合撂荒地样本和非撂荒地样本NDVINDVI差值和NDVINDVI差值,设定分割阈值来提取研究区撂荒地。研究区2017—2022年总体分类精度OA均≥0.85,Kappa系数均≥0.80,整体分类效果良好,可以进行后续的耕地提取;从水平尺度看,研究区撂荒地集中分布在南北山地,其次分布在沿湟水河两岸;从垂直尺度看,随着海拔上升,撂荒率呈正态分布,撂荒地集中分布在2 000~2 500 m,撂荒率随着坡度的增加而增加,这与坡度的增加会导致耕地质量下降和农业机械的难以利用有很大关系。相较于传统土地利用遥感分类研究,借助GEE平台开展的撂荒地识别研究能够快速获悉区域尺度下的撂荒地分布情况,为提取该地区撂荒地和土地利用保护提供参考。

关键词: 耕地, 撂荒地, 空间分布特征, GEE, NDVI, 撂荒率, 青海省海东市乐都区

Abstract: In order to achieve timely and accurate identification of farmland, remote sensing technology was used to identify and extract abandoned farmland in the agricultural pastoral transitional zone, and to understand the spatial distribution characteristics of abandoned farmland.Based on the Google Earth Engine (GEE) platform, the study area’s Sentienel-1 and Sentienel-2 remote sensing images were called and preprocessed. The random forest algorithm was used to conduct land use classification research in the study area,and obtain the monthly maximum NDVI composite data of the study area from 2017 to 2022 through the GEE platform. Combined with the NDVI summer and spring differences and NDVI summer and autumn differences of abandoned and non abandoned farmland samples, segmentation thresholds to extract abandoned farmland in the study area were set. The results showed that the overall classification accuracy OA of the study area from 2017 to 2022 was ≥0.85, and the Kappa coefficient was ≥0.80. The overall classification effect was good, and it could be used for subsequent farmland extraction;from a horizontal scale, the abandoned farmland in the study area was mainly distributed in the north-south mountainous areas, followed by along the banks of the Huangshui River;from a vertical scale perspective, as the altitude increased, the abandonment rate followed a normal distribution, with abandoned farmland concentrated between 2 000 and 2 500 meters. The abandonment rate increased with the increase of slope, which was closely related to the decline in farmland quality and the difficulty in utilizing agricultural machinery caused by the increase of slope.Compared to traditional land use remote sensing classification research, abandoned farmland identification research conducted using the GEE platform could quickly obtain the distribution of abandoned farmland at the regional scale, providing reference for extracting abandoned farmland and land use protection in the region.

Key words: farmland, abandoned farmland, spatial distribution characteristics, GEE, NDVI, abandonment rate, Ledu District,Haidong City,Qinghai Province

中图分类号: