欢迎访问《生态学杂志》官方网站,今天是 分享到:

生态学杂志

• 研究报告 • 上一篇    下一篇

秋茄类胡萝卜素含量高光谱反演

高灯州1,3,章文龙1,3,陈美田1,张新中1,曾从盛2,3**   

  1. 1福建师范大学地理科学学院, 福州 350007; 2湿润亚热带生态-地理过程教育部重点实验室, 福州 350007; 3福建师范大学亚热带湿地研究中心, 福州 350007)
  • 出版日期:2014-11-10 发布日期:2014-11-10

Retrieval of carotenoid contents of Kandelia candel based on hyper-spectral remote sensing data. 

GAO Deng-zhou1,3, ZHANG Wen-long1,3, CHEN Mei-tian1, ZHANG Xin-zhong1, ZENG Cong-sheng2,3**   

  1. (1School of Geographical Sciences, Fujian Normal University, Fuzhou 350007, China; 2Key Laboratory of Humid Subtropical Ecogeographical Process of Ministry of Education, Fuzhou 350007, China; 3Research Centre of Wetlands in Subtropical Region, Fuzhou 350007, China)
  • Online:2014-11-10 Published:2014-11-10

摘要:

类胡萝卜素(Car)作为植物主要色素,对诊断植被生理状态有重要作用。于2013年4月和7月采集闽江口秋茄(Kandelia candel)叶片,室内测定其叶片正面和背面反射光谱,同时测定其Car含量\[单位面积(μg·cm-2)和单位质量(mg·g-1)\]。选取常见Car含量估算的光谱参数,同时分析确定最佳比值植被指数(SR),基于回归分析,建立秋茄叶片Car含量估算与验证模型。结果表明,叶片光谱反射率表现为叶片背面大于正面(350~2350 nm);基于叶面背面光谱计算的SR与叶片Car含量(μg·cm-2)的相关系数优于其他组合,相关系数较高的区域分布在520~540 nm与1000~1100 nm波段组合,700~720 nm与800~1100 nm波段组合;基于背面光谱计算的大部分光谱参数与Car含量(μg·cm-2)的相关系数要高于基于正面光谱计算的。因此,以叶片背面光谱作为Car含量估算的光谱数据,以单位面积Car含量为估算量纲建立反演模型。本研究表明,光谱指数LCI、DD、NDVI(770,713)、NDVI(773,562)、SR(723,770)SR(1000,700)均可实现Car含量的反演,估算与检验模型的R2均>0.65,RMSE均<1.52;并且新构建的SR(1000,700)估算精度最好,模型和检验R2分别为0.77和0.87,模型和检验RMSE分别为1.08和1.11。这些预味着基于高光谱遥感对闽江河口湿地秋茄Car含量进行估算是可行的。
 

关键词: 聚集度系数, Ross-Li模型, MODIS BRDF

Abstract: Carotenoid (Car), the main pigment of green leaves, plays an important role in diagnosing the physiological state of vegetation. The leaves of Kandelia candel were sampled from the Minjiang River estuary in April and July, 2013. In the laboratory, spectral reflectance of leaves (both front and back), and the Car contents (two dimensions: μg·cm-2 and mg·g-1) were determined. The common parameters and the best simple ratio spectral index (SR) were used to build the estimation models. The results showed that the spectral reflectance of leaf back was higher than that of leaf front in 350-2350 nm. The relationship between SR calculated from spectral reflectance of leaf back and Car content (μg·cm-2) was better than other combinations. SR calculated by bands in two zones, including the 520-540 nm and 1000-1100 nm, 700-720 nm and 800-1100 nm had higher correlation coefficient compared with that in other zones. Additionally, most of the parameters calculated by leaf back reflectance had higher correlation coefficients with Car content (μg·cm-2) than that of leaf front. Consequently, parameters calculated by leaf back reflectance and the Car content in per unit area were selected to establish the estimation and validation models. The model parameters further indicated that the models built by LCI, DD, NDVI(770,713), NDVI(773,562), SR(723,770) and SR(1000,700) could achieve the estimation of Car contents (R2>0.65, RMSE<1.52). The best estimation parameter was SR(1000,700):R2 values of its estimation and validation models were 0.77 and 0.87, and RMSE values were 1.08 and 1.11, respectively. The results impied that hyperspectral remote sensing data could be used to estimate Car contents of K. candel.

Key words: Ross-Li model, clumping index, MODIS BRDF