https://dx.doi.org/10.1016/j.rse.2005.07.011">
 

Title

A comparison of methods for estimating fractional green vegetation cover within a desert-to-upland transition zone in central New Mexico, USA

Abstract

We compared a set of methods for estimating fractional green vegetation cover (fc) over a ∼4000 km2 region of central New Mexico, USA. The models used were trained and tested independently using high-resolution, true-color orthoimagery with 0.3 m spatial resolution. Simple NDVI-based methods performed well for estimating fc regionally but overestimated fc in sparsely vegetated areas with bright soils, and areas with abundant non-photosynthetic vegetation (e.g. dry shrubs). Three-, four-, and five-endmember spectral mixture models (SMA3, SMA4, and SMA5) were also compared. Constrained versions of these models all produced similar accuracy regionally, but constrained and unconstrained versions of the SMA5 model best captured fc for the rarer landscapes (bright soils, riparian vegetation) found throughout the region. This indicates that heterogeneous landscapes can be stratified into relatively homogeneous strata, and three or four endmembers may be adequate to characterize the spectral variability within each stratum. Including NDVI along with the six reflective bands of ETM+ data, provided enough data dimensionality to support the five-endmember SMA model. This permitted a more complete representation of the range of spectral landscape types that are germane for separating out green vegetation in this semi-arid region. We also note that green woody vegetation and green grass cover should be spectrally represented by two different endmembers in SMA because these two vegetation types are spectrally different, particularly in the near-infrared (NIR) wavelength.

Publication Date

10-15-2005

Journal Title

Remote Sensing of Environment

Publisher

Elsevier

Digital Object Identifier (DOI)

https://dx.doi.org/10.1016/j.rse.2005.07.011

Document Type

Article

Rights

Copyright © 2005 Elsevier Inc. All rights reserved.