收藏设为首页 广告服务联系我们在线留言

干旱区地理 ›› 2022, Vol. 45 ›› Issue (2): 478-487.doi: 10.12118/j.issn.1000–6060.2021.177

• 地球信息科学 • 上一篇    下一篇

极干旱区稀疏荒漠植被地上生物量遥感估算

叶静芸1(),吴波1,2(),贾晓红1,2,费兵强1,2,高君亮3,成龙1,2,庞营军1,2,姚斌1,2,孔德庸4   

  1. 1.中国林业科学研究院荒漠化研究所,北京 100091
    2.国家林业和草原局荒漠生态系统与全球变化重点实验室,北京 100091
    3.中国林业科学研究院沙漠林业实验中心,内蒙古 磴口 015200
    4.韶关学院,广东 韶关 512005
  • 收稿日期:2021-04-19 修回日期:2021-11-26 出版日期:2022-03-25 发布日期:2022-04-02
  • 通讯作者: 吴波
  • 作者简介:叶静芸(1988-),女,博士研究生,主要从事植被定量遥感研究. E-mail: 517304329@qq.com
  • 基金资助:
    国家自然科学基金项目(41471151);国家重点研发计划项目(2016YFC0500806);国家重点研发计划项目(2016YFC0500801);科技基础资源调查专项项目资助(2017FY100206)

Estimation of aboveground biomass of sparse desert vegetation based on remote sensing techniques in hyper-arid area

YE Jingyun1(),WU Bo1,2(),JIA Xiaohong1,2,FEI Bingqiang1,2,GAO Junliang3,CHENG Long1,2,PANG Yingjun1,2,YAO Bin1,2,KONG Deyong4   

  1. 1. Institute of Desertification Studies, Chinese Academy of Forestry, Beijing 100091, China
    2. Key Laboratory of Desert Ecosystem and Global Change, State Administration of Forestry and Grassland, Beijing 100091, China
    3. Experimental Center of Desert Forestry, Chinese Academy of Forestry, Dengkou 015200, Inner Mongolia, China
    4. Shaoguan University, Shaoguan 512005, Guangdong, China
  • Received:2021-04-19 Revised:2021-11-26 Online:2022-03-25 Published:2022-04-02
  • Contact: Bo WU

摘要:

以库姆塔格沙漠南部阿尔金山北麓山前戈壁区为研究区,借助无人机影像数据准确提取植被覆盖区;采用高空间分辨率WorldView-3数据的估算结果对中空间分辨率Landsat-OLI数据的估算结果进行修正;选取红光反射率波段(RED)、近红外反射率波段(NIR)、比值植被指数(RVI)、归一化植被指数(NDVI)和修正土壤调节植被指数(MSAVI)反射率波段作为遥感特征参数,采用逐步线性回归(SLR)、套索回归(LASSO)和岭回归(RR)模型分别对稀疏荒漠植被地上生物量进行估算。结果表明:(1) WorldView-3数据可准确刻画极干旱区稀疏荒漠植被的时空分布特征,其估算精度高于Landsat-OLI数据。(2) 采用非线性的多元回归模型SLR、LASSO和RR构建遥感特征参数与地面样方数据之间的回归模型可以有效提高模型的稳定性。RR模型的回归效果优于SLR模型和LASSO模型。(3) 提取稀疏荒漠植被信息时,采用WorldView-3数据的植被生物量估算结果对Landsat-OLI数据的植被地上生物量估算结果进行修正,可以有效提高Landsat-OLI数据的提取精度。

关键词: 遥感特征参数, 稀疏荒漠植被, 地上生物量, 回归模型

Abstract:

The aboveground biomass (AGB) is an essential indicator of desert ecosystem health evaluation and desertification monitoring in arid areas. The estimation accuracy of the general remote sensing model is reduced because of the influence of small and sparse desert vegetation, low AGB per unit area, and background soil information. Improving biomass estimation accuracy of sparse desert vegetation can provide essential parameters for terrestrial ecosystem carbon storage estimation and carbon cycle research. It can also provide technical support for desertification monitoring. Herein, the Gobi area in front of the Altun Mountains in the south of Kumtag Desert, northwest China with an extremely arid climate is considered as the research area. The vegetation coverage area was accurately extracted using UAV image data. The estimation results based on the high-spatial resolution WorldView-3 remote sensing data were used to correct the estimation results based on the medium spatial resolution Landsat-OLI remote sensing data. Then, stepwise linear regression (SLR), LASSO regression, and ridge regression (RR) models were used to estimate the AGB of sparse desert vegetation. The results showed that: (1) High-spatial resolution satellite remote sensing data can accurately describe the spatial and temporal distribution characteristics of vegetation in arid areas. The RMSErel mean value of WorldView-3 remote sensing data was 44.68%. (2) The RR model had a better regression effect than SLR and LASSO models. The RMSErel of the WorldView-3 RR model was 44.68%, better than the SLR model (58.33%) and LASSO model (45.26%). Additionally, the RMSErel of the Landsat-OLI RR model was 114.34%, better than the SLR model (134.94%) and LASSO model (115.81%). (3) Using the WorldView-3 remote sensing data as an intermediate medium between ground and Landsat-OLI remote sensing data, the AGB estimation results of desert vegetation based on Landsat-OLI remote sensing data were corrected, and the RMSErel decreased by 31.18%. The spatial resolution of remote sensing data is the main factor affecting the estimation accuracy of the sparse desert vegetation AGB regression model. The main reason for the decrease in the estimation accuracy of AGB is the decrease in the sensitivity of Landsat-OLI remote sensing data to sparse desert vegetation information. The nonlinear multiple regression methods, SLR, LASSO, and RR, are used to build the regression model between remote sensing characteristic parameters and ground sample data, effectively improving model stability. When extracting sparse desert vegetation information, using high-spatial resolution remote sensing data as a medium between ground and medium spatial resolution remote sensing data can effectively improve the extraction accuracy of sparse desert vegetation information from medium spatial resolution remote sensing data.

Key words: remote sensing eigenvalue, sparse desert vegetation, aboveground biomass, regression model