Monitoring Rangeland Plant Community Composition Using Spectral Mixture Analysis of Hyperspectral Remote Sensing Data

摘要

This paper investigates the abundance mapping of rangeland plant communities using hyperspectral remote sensing data. Spectral Mixture Analysis (SMA) was used to estimate the cover fraction of five rangeland components: green grass, yellow grass, litter, shrubs and soil. Two types of endmembers were assessed using canopy reflectance modeling and tested over real data. The first type is the leaf endmember based on the laboratory reflectance measurements of different samples of leaves. The second is the canopy endmember based on reflectance simulation using the canopy radiative transfer model SAIL. These two endmember types were first assessed in SMA using a number of homogenous canopy simulations with different Leaf Area Index (LAI). Subsequently, the leaf and the canopy endmembers were evaluated using ground spectra, and cover fractions were compared to actual data. Finally, both endmember types were applied in SMA to CHRIS/PROBA data to estimate the rangeland component cover fractions. Performances of leaf and canopy endmembers were evaluated based on the field knowledge of the area of interest. Results showed overall that the cover fraction estimates using the canopy endmembers tend to better agree with actual data.