JOURNAL OF NATURAL RESOURCES ›› 2017, Vol. 32 ›› Issue (12): 2125-2135.doi: 10.11849/zrzyxb.20161086

• Research Method • Previous Articles     Next Articles

Anisotropic Reflectance Effect on the Spectral Mixture Analysis for Vegetation Coverage Estimation

DUAN Li-min, TONG Xin, LIU Ting-xi   

  1. Water Conservancy and Civil Engineering College, Inner Mongolia Agricultural University, Inner Mongolia Water Resource Protection and Utilization Key Laboratory, Hohhot 010018, China
  • Received:2016-10-11 Revised:2017-04-19 Online:2017-12-20 Published:2017-12-20
  • Supported by:
    National Natural Science Foundation of China, No.51369016 and 51620105003; Excellent Young Scientist Foundation of Inner Mongolia Agricultural University of China, No.2014XYQ-11; Natural Science Foundation of Inner Mongolia, No.2015MS0566; Ministry of Education Innovative Research Team, No.IRT_17R60; Innovation Team in Priority Areas Accredited by the Ministry of Science and Technology, No.2015RA4013

Abstract: Spectral mixture analysis (SMA) models are highly effective methods used to deal with sub-pixel vegetation coverage estimation, among which linear spectral mixture analysis (LSMA) is the most commonly used one. However, the precision of vegetation coverage estimation retrieved from LSMA is mainly affected by the multiple scattering and end-member spectral variability. Also, anisotropic reflectance effect (ARE), one of the distinctive and inherent properties of surfaces, is very likely to be ignored. This research conducted in situ spectral experiments by using checkerboard-style mixture design which incorporated three types of surfaces. After discussing and analyzing the traditional multiple scattering and the end-member spectral variability, the effect of anisotropic reflectance on the spectral mixture analysis for vegetation coverage estimation was further evaluated. The results indicated that the impact of ARE cannot be neglected. The Carex duriuscula coverage estimation was more accurate after considering of ARE, when minimizing the effect of the traditional multiple scattering and end-member spectral variability. The root mean square error (RMSE) decreased nearly 52%. These results not only emphasized the importance of integrating ARE into vegetation coverage estimation but also indicated that ARE can be regarded as another significant source of variability within the same end-member class. This study broadens the scope of end-member spectral variability, and may put forward a new thinking and direction for vegetation coverage estimation based on SMA.

Key words: anisotropic reflectance effect, checkerboard-style mixture, spectral mixture analysis, vegetation coverage estimation

CLC Number: 

  • P237