The influence of canopy green vegetation fraction on spectral measurements over native tallgrass prairie

TitleThe influence of canopy green vegetation fraction on spectral measurements over native tallgrass prairie
Publication TypeJournal Article
Year of Publication2002
AuthorsRundquist, BC
JournalRemote Sensing of the Environment
Pagination129 -135
Accession NumberKNZ00843
KeywordsGreen vegetation fraction; Field spectroscopy; Tallgrass prairie

Spectral vegetation indices (SVIs) calculated from remotely sensed data are routinely used to monitor spatial and temporal changes in vegetation biophysical characteristics. The most commonly used SVI, the Normalized Difference Vegetation Index (NDVI), has been criticized because of its sensitivity to atmospheric conditions and substrate reflectivity, as well as its insensitivity to increases in vegetation biomass past particular thresholds. Yet, the use of NDVI remains widespread and is attractive because of the ease with which it is calculated. This article examines the utility of NDVI for monitoring the biophysical characteristic of green vegetation fraction (GVF) in comparison to other SVIs suggested as improvements. Statistical relationships between spectral response, presented in the form of SVIs, and GVF of a native tallgrass prairie canopy are explored. Broadband spectra were gathered from close-range during the 1999 growing season at the Konza Prairie Biological Station (KPBS), located in the Flint Hills region of Kansas, USA. Through simple regression analyses, spectra were related to GVF estimates derived from digital color photographs. SVIs evaluated are the NDVI, the Soil Adjusted Vegetation Index (SAVI), and the square of scaled NDVI (N*2). Results show that NDVI and N*2 were statistically related to GVF (R2 for NDVI=.77, N*2=.78) throughout the growing season. The least-squares line defining the relationship between N*2 and GVF approximated a 1:1 line. For June sample dates, all three SVIs were significant statistical predictors of GVF (R2 for NDVI=.89, N*2=.91, SAVI=.89). Regression coefficients for late-season sample dates were weaker, yet still significant in statistical terms (R2 for NDVI=.70, N*2=.70). While encouraging, these results suggest that further analyses are required to determine the usefulness of SVIs calculated from broadband devices for estimation of GVF when leaf litter dominates the scene.