四川农业大学学报 ›› 2016, Vol. 34 ›› Issue (04): 456-463.doi: 10.16036/j.issn.1000-2650.2016.04.011

• • 上一篇    下一篇

不同遥感影像空间分辨率对川西南山地常绿阔叶林有效叶面积指数估测的影响

赵安玖1, 陈银华2, 毛加勇3   

  1. 1. 四川农业大学林学院/水土保持与荒漠化防治省级重点实验室, 成都 611130;
    2. 四川省林业调查规划院, 成都 610081;
    3. 四川省林业厅, 成都 610081
  • 收稿日期:2016-05-31 出版日期:2016-12-31 发布日期:2017-02-15
  • 作者简介:赵安玖,博士,副教授,主要从事林业GIS、林木生长与环境关系研究,E-mail:zaj9828@aliyun.com。
  • 基金资助:
    国家重点科技攻关项目(2011BAC09B05)

Effects of Different Spatial Resolution on LAIe Estimation of Evergreen Broad-leaved Forests in Southwest Sichuan

ZHAO An-jiu1, CHEN Yin-hua2, MAO Jia-yong3   

  1. 1. College of Forestry, Sichuan Key Laboratory of Soil & Water Conservation and Desertification Combating, Sichuan Agricultural University, Chengdu 611130, China;
    2. Sichuan Forestry Inventory and Planning Institute, Chengdu 610081, China;
    3. Forestry Department of Sichuan Province, Chengdu 610081, China
  • Received:2016-05-31 Online:2016-12-31 Published:2017-02-15

摘要: [目的] 分析遥感影像不同空间分辨率对LAIe估测结果的影响。[方法] 基于地面调查的83个20 m×20 m样地和Landsat-8、SPOT-5、Pleiades-1遥感数据,以川西南山地常绿阔叶林为研究对象,运用偏最小二乘回归分析法,估测了2 m、10 m、30 m 3种尺度(粒度)上区域森林有效叶面积指数(LAIe)。[结果] 3种分辨率的遥感数据提取的植被指数NDVI、SAVI对LAIe估测最为重要(Landsat-8:NDVI、SAVI的VIP=1.662;SPOT-5:NDVI、SAVI的VIP=1.573;Pleiades-1:NDVI、SAVI的VIP=1.423)。3种传感器的NDVI、SAVI的相关系数大于0.8,均达极显著水平。对LAIe回归估测检验显示,Landsat-8的决定系数R2=0.793,精度P=79.8%;SPOT-5的决定系数R2=0.853,P=84.4%;Pleiades-1的R2高达0.898,估测精度最高,达89.5%。[结论] 不同空间分辨率的影像对LAIe估测有显著影响,使用高空间分辨率数据能显著提高LAIe估测精度。

关键词: 有效叶面积指数估测, 偏最小二乘回归法, 空间分辨率

Abstract: [Objective] The aim of the study was to analyze the effects of different spatial resolutions on the LAIe estimation. [Method] Effective leaf area index(LAIe) of montane evergreen broad-leaved forests in southwest Sichuan was inventoried and assessed based on 83 field plots of 20 m×20 m and spectral vegetation index(Normalised Difference Vegetation Index, NDVI; Soil Adjusted Vegetation Index, SAVI; Ratio Vegetation Index, RVI) were extracted from different spatial resolution image data of Landsat-8(30 m), SPOT-5(10 m), Pleiades-1(2 m). Based on remotely sensed data, inventoried LAIe and auxiliary variables(eg. relief data), regress models were established by partial least multiplicative regression method. [Results] NDVI and SAVI extracted from three kinds of resolution remote sensing data were most important to LAIe estimation. The VIP value of NDVI and SAVI extracted from Landsat-8 was 1.662. The VIP value of NDVI, and SAVI extracted from SPOT-5 was 1.573. Similar to Pleiades-1, the VIP value was 1.423. According to three kinds of sensors, all correlation coefficients between NDVI, SAVI and LAIe were more than 0.8, all of which were significant. The coefficient of determination R2 was 0.793 and precision accuracy was 79.8% for the regress model using Landsat-8 data. Likewise, the coefficient of determination R2 was 0.853 and precision accuracy was 84.4% for the regress model using SPOT-5 data. Moreover, R2 and P of regress model using Pleiades-1 data was highest compared to other regress models, which reached to 0.898 and 89.5%, respectively. [Conclusion] Different image spatial resolution have significant effects on LAIe estimation and the performance of LAIe estimation can be improved significantly using high spatial resolution optical data, such as Pleiades-1 remote sensing data.

Key words: LAIe estimation, partial least multiplicative regression(PLSR), spatial resolution

中图分类号: 

  • P966