HR: 1330h
AN: B22A-0797    [PDF]
TI: The Effect of Spatial and Spectral Resolution in Determining NDVI
AU: * Boelman, N T
EM: nboelman@ldeo.columbia.edu
AF: Lamont-Doherty Earth Observatory, Columbia University, 206a Oceanography 61 Route 9W, Palisades, NY 10964 United States
AB: We explore the impact that varying spatial and spectral resolutions of several sensors (a field portable spectroradiometer, Landsat, MODIS and AVHRR) has in determining the average Normalized Difference Vegetation Index (NDVI) at Imnavait Creek, a small arctic tundra watershed located on the north slope of Alaska. We found that at the field-of-views (FOVs) of less than 20 m$^{2}$ that were sampled, the average NDVI value for this watershed is 0.65, compared to 0.77 at FOVs equal to and greater than 20 m$^{2}$. In addition, we found that at FOVs less than 20 m$^{2}$, the average NDVI value calculated according to each of Landsat, MODIS and AVHRR band definitions (controlled by spectral resolution) was similar. However, at FOVs equal to and greater than 20 m$^{2}$, the average NDVI value calculated according to AVHRR's broad-band definitions was significantly and consistently higher than that from both Landsat and MODIS's narrow-band NDVI values. We speculate that these differences in NDVI exist because high leaf-area-index vegetation communities associated with watertracks are commonly spaced between 10 and 20 m apart in arctic tundra landscapes and are often only included when spectral sampling is conducted at FOVs greater than tens of square meters. These results suggest that both spatial resolution alone and its interaction with spectral resolution have to be considered when interpreting commonly used global-scale NDVI datasets. This is because traditionally, the fundamental relationships established between NDVI and ecosystem parameters, such as CO$_{2}$ fluxes, aboveground biomass and net primary productivity, have been established at scales less than 20 m$^{2}$. Other ecosystems, such as landscapes with isolated tree islands in boreal forest-tundra ecotones, may exhibit similar scaling patterns that need to be considered when interpreting global-scale NDVI datasets.
DE: 0400 Biogeosciences
DE: 1640 Remote sensing
DE: 1851 Plant ecology
SC: Biogeosciences [B]
MN: 2003 Fall Meeting