草地学报 ›› 2025, Vol. 33 ›› Issue (4): 1258-1266.DOI: 10.11733/j.issn.1007-0435.2025.04.026

• 研究论文 • 上一篇    下一篇

基于无人机遥感的荒漠草地地上生物量反演研究

李文雄, 靳瑰丽, 刘文昊, 李嘉欣, 王生菊, 陈梦甜, 李超, 杜玟霖, 萨木尕尔·达吾列提开勒得, 叶里达那·赛力克艾力   

  1. 1. 新疆农业大学草业学院/新疆草地资源与生态重点实验室/西部干旱荒漠区草地资源与生态教育部重点实验室, 新疆 乌鲁木齐 830052
  • 收稿日期:2024-06-18 修回日期:2024-09-12 出版日期:2025-04-15 发布日期:2025-04-28
  • 通讯作者: 靳瑰丽,E-mail:jguili@163.com
  • 作者简介:李文雄(1999-),男,土家族,重庆人,硕士研究生,主要从事草地资源与生态研究,E-mail:939314017@qq.com
  • 基金资助:
    国家自然科学基金项目(31960360)资助

Research on Aboveground Biomass Inversion of Desert Grassland Based on UAV Remote Sensing

LI Wen-xiong, JIN Gui-li, LIU Wen-hao, LI Jia-xin, WANG Sheng-ju, CHEN Meng-tian, LI Chao, DU Wen-lin, Davuletti Khaled Samugal, Serik Eli Yeridana   

  1. 1. College of Grassland Sciences of Xinjiang Agricultural University/Xinjiang Key Laboratory of Grassland Resources and Ecology,/Key Laboratory of Grassland Resources and Ecology for Western Arid Desert Rengion, Ministry of Education Urumqi, Urumqi, Xinjiang 830052, China
  • Received:2024-06-18 Revised:2024-09-12 Online:2025-04-15 Published:2025-04-28

摘要: 荒漠草地地上生物量(Aboveground biomass,AGB)是评估植被状况与荒漠化进程的重要指标。为迅速、精确且高效地评估荒漠草地地上生物量,本研究以新疆伊犁绢蒿(Seriphidium transiliense)荒漠草地为研究区,在植被生长旺季采集草地AGB数据,同步获取无人机(Unmanned aerial vehicle,UAV)数据;选取10种植被指数为特征变量,利用3种机器学习算法构建AGB反演模型,并引入遗传算法(Genetic algorithm,GA)优化模型参数,进而筛选出最佳的AGB反演模型。结果表明:3种算法均表现出较高的预测性能,其中XGBoost模型优势显著,尤其是在融合了4种典型植被指数并采用遗传算法优化后,其预测精度达到最高(R2=0.94,RMSE=3.44),其中RVI贡献最大,占比35%。因此,基于4种典型植被指数并结合GA优化的XGBoost模型被认定为最适用于研究区域草地AGB遥感反演的模型。此研究结果可为监测草地生物量遥感反演方法的选择和精度的提高提供一定参考。

关键词: 荒漠草地, 地上生物量, 无人机, 极限梯度提升, 随机森林, 轻量级梯度提升

Abstract: Aboveground biomass (AGB) is an important index to evaluate vegetation status and desertification process in desert grassland. In order to evaluate the aboveground biomass (AGB) of desert grassland rapidly, accurately and efficiently, the desert grassland of Seriphidium transiliense in Xinjiang was taken as the research area in this study. The AGB data of grassland were collected in the vegetation growth season, and the unmanned aerial vehicle (UAV) data were obtained simultaneously. Ten vegetation indices were selected as the characteristic variables, and three machine learning algorithms were used to construct the AGB inversion model. The genetic algorithm (GA) was introduced to optimize the model parameters, and then the best AGB inversion model was selected. The results showed that the three algorithms all had high prediction performance, among which the XGBoost model had significant advantages. Especially after integrating four typical vegetation indices and using genetic algorithm (GA) optimization, the prediction accuracy reached the highest (R2=0.94, RMSE=3.44), of which RVI contributed the most, accounting for 35 %.Therefore, the XGBoost model based on four typical vegetation indices combined with GA optimization was identified as the most suitable model for grassland AGB remote sensing inversion in the study area. The results of this study could provide a reference for the selection of remote sensing inversion methods for monitoring grassland biomass and the improvement of accuracy.

Key words: Desert grassland, Aboveground biomass, Unmanned aerial vehicle, eXtreme gradient boosting, Random forest, Light gradient boosting machine

中图分类号: