地理科学 ›› 2020, Vol. 40 ›› Issue (1): 149-157.doi: 10.13249/j.cnki.sgs.2020.01.018

• • 上一篇    下一篇

基于Landsat时间序列遥感数据的华北平原农田火烧迹地检测与制图

张素梅1, 杜惠琳1, 刘良2, 白翔宇1, 冯凯东1, 赵红梅3()   

  1. 1. 太原理工大学矿业工程学院,山西 太原 030024
    2. 武汉大学遥感信息工程学院,湖北 武汉 430079
    3. 中国科学院东北地理与农业生态研究所,吉林 长春 130102
  • 收稿日期:2019-02-10 修回日期:2019-05-09 出版日期:2020-01-10 发布日期:2020-03-17
  • 通讯作者: 赵红梅 E-mail:zhaohongmei@iga.ac.cn
  • 作者简介:张素梅(1982-),女,河南巩义人,讲师,主要从事资源环境遥感研究。E-mail: zsm_2004@126.com
  • 基金资助:
    国家自然科学基金项目(41771504);国家重点研发计划(2017YFC0212303)

Detecting and Mapping Burned Areas for Croplands Based on Landsat Time Series Remote Sensing Data in North China Plain

Zhang Sumei1, Du Huilin1, Liu Liang2, Bai Xiangyu1, Feng Kaidong1, Zhao Hongmei3()   

  1. 1. College of Mining and Engineering, Taiyuan University of Technology, Taiyuan 030024, Shanxi, China
    2. School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, Hubei, China
    3. Key Laboratory of Wetland Ecology and Environment, Northeast Institute of Geography andAgroecology, Chinese Academy of Sciences, Changchun 130102, Jilin, China
  • Received:2019-02-10 Revised:2019-05-09 Online:2020-01-10 Published:2020-03-17
  • Contact: Zhao Hongmei E-mail:zhaohongmei@iga.ac.cn
  • Supported by:
    National Natural Science Foundation of China(41771504);National Key R&D Plan of China(2017YFC0212303)

摘要:

以河南省漯河市、周口市和驻马店市交界处为研究区,基于2000年10月~2011年12月间的101幅图像,采用谐波模型和断点识别算法拟合Landsat时间序列实现对过火像元的检测,并将结果与目视解译结果、MODIS火烧迹地产品MCD64A1检测结果对比进行精度分析。结果表明:① 随着燃烧面积指数BAI (Burned Area Index)异常值阈值增大,焚烧火点误判误差减小,漏判误差增大,火烧迹地制图的总体精度先增大后减小;② 当BAI异常值阈值2.9×RMSE(Root Mean Square Error)时,该方法总体精度达到最高93.25%,MCD64A1产品总体精度为70.25%;③ 本文算法的漏判误差和误判误差相对平衡,而MCD64A1产品的漏判误差远大于误判误差。研究表明,相比MODIS火烧迹地产品数据,Landsat时间序列火烧迹地法可更有效地检测农田火烧迹地。

关键词: 秸秆焚烧, 火烧迹地, 谐波模型, Landsat时间序列, BAI

Abstract:

Straw burning releases large qualities of gases and aerosol particles threatening the human health and air quality. Remote sensing plays a more and more important role in monitoring burned area for croplands either at regional scales or at global scales in recent years. The algorithm was based on harmonic model and break point identification to detect annual burned area using Landsat time series and then the algorithm was tested in the border area of Luohe,Zhoukou and Zhumadian in Henan Province using all available Landsat imagery between year 2000 and year 2011. The results were then compared with visual interpretation results and MODIS product MCD64A1 respectively for accuracy analysis and assessment,which show that: 1) As Burned Area Index(BAI) threshold increases,both the commission errors decrease and the omission errors increase. The overall accuracy of mapping cropland burned areas first increases and then decreases. 2) When BAI threshold is 2.9 times root mean square error(RMSE), the overall accuracy reached at the peak value,which is 93.25%. This represents a significant improvement in comparison with MCD64A1 product (The overall accuracy of MCD64A1 product is 70.25%). 3) The algorithm in this paper had balanced omission and commission errors,while MODIS product had more omission errors than commission errors. The results sufficiently show that,the Landsat time series based method in this paper can detect cropland burned area more effectively in comparison with MODIS burned area product.

Key words: straw burning, burned area, Harmonic model, Landsat time series, BAI

中图分类号: 

  • X87/TP732