HR: 0800h
AN: B51B-0194    [Abstracts]
TI: Multiscale Geostatistical Analysis of AVHRR, SPOT-VGT, and MODIS NDVI products
AU: Tarnavsky, E
EM: elena@rohan.sdsu.edu
AF: King's College London, Strant, London, WC2R 2LS United Kingdom
AU: * Brown, M E
EM: molly.brown@gsfc.nasa.gov
AF: NASA Goddard Space Flight Center, Code 614.4, Greenbelt, MD 20771 United States
AU: Garrigues, S
EM: garrigue@pop600.gsfc.nasa.gov
AF: NASA Goddard Space Flight Center, Code 614.4, Greenbelt, MD 20771 United States
AB: Regional and global monitoring of terrestrial ecosystems is routinely achieved with remote sensing measurements processed into spectral vegetation indices, from which the normalized difference vegetation index (NDVI) has been among the most widely used. Time series NDVI records are available from 1981 to present from the NOAA AVHRR sensor, and for recent years from the SPOT-Vegetation (SPOT-VGT) and MODIS/Terra imaging instruments. AVHRR, SPOT-VGT, and MODIS NDVI products are similar in terms of general procedures for deriving vegetation index measurements but have different sensor characteristics and image processing chains. In this paper, our goal was to investigate whether these similarities and dissimilarities affect the variance of images at the different spatial scales (250 m, 500 m, 1 km, and 8 km). Previously, such cross-scale comparisons have been conducted using spatially aggregated data from a single sensor and image. The proposed methodology for this study involved geostatistical analysis of multi-sensor data (AVHRR, SPOT-VGT, and MODIS NDVI image subsets) for eight Earth Observing System (EOS) Land Validation Core Sites spread across different ecosystem types. Our first objective was to investigate the inter-relationship between empirical semivariograms derived from each image type for each validation site. The second objective was to investigate whether similar ecosystem types are characterized by common variogram curve parameters (curve shape, sill and range values). We found that in general the response of the multi-sensor dataset is consistent with geostatistical theory, with the exception of sporadic heterogeneity observed in the SPOT-VGT data relative to coarser resolution AVHRR NDVI and nearly the same resolution MODIS 1-km NDVI product. Overall, the cross-scale geostatistical analysis of spatial variance proved a promising method for comparison of actual multi-sensor datasets.
DE: 0400 BIOGEOSCIENCES
DE: 0480 Remote sensing
DE: 1545 Spatial variations: all harmonics and anomalies
DE: 3252 Spatial analysis (0500)
SC: Biogeosciences [B]
MN: Fall Meeting 2005