Title: Mediterranean-wide Green Vegetation Abundance for Land Degradation Assessment Derived from AVHRR NDVI and Surface Temperature 1989 to 2005
Publisher: OPOCE
Publication Year: 2008
JRC N°: JRC47272
ISBN: 978-92-79-09777-5
ISSN: 1018-5593
Other Identifiers: EUR 23500 EN
URI: http://publications.jrc.ec.europa.eu/repository/handle/JRC47272
DOI: 10.2788/9597
Type: EUR - Scientific and Technical Research Reports
Abstract: NOAA AVHRR data stemming from the MEDOKADS archive and ranging from 1989 to 2005 was processed and decomposed into their fractions of the vegetated, non-vegetated and the so called ¿cold¿ endmember. Decomposition occurred via Linear Unmixing within a triangle spanned up by NDVI (y-axis) and surface temperature (x-axis), separately for each of the 612 10-day composites. Endmembers were derived statistically using percentiles and the inverse relationship between NDVI and Ts. The cold endmember was fixed at -20 degrees Celsius, the vegetated endmember at NDVI = 0.7, the latter was then empirically corrected for illumination effects. Linear Unmixing occurred for the whole Mediterranean area, separately for a western and eastern window. Outcomes are the vegetation abundance, soil abundance and ¿cold¿ abundance, indicating the individual coverage of a pixel by each of these. The vegetation abundance was re-scaled to the so-called Grenn Vegetation Fraction (GVF), re-distributing the ¿cold¿ abundance on vegetation and soil abundance proportionally. Unmixing led to a higher stability of GVF data in comparison to NDVI data with regard to atmospheric effects. The data was post-processed for missing values and outliers and it was filtered. The GVF shows close parallelism on several test sites in comparison to a re-scaled NDVI within the endmember limits. The positive effect of the cold abundance, which is amongst other accounting for negative effects from poor atmospheric conditions and which was used to improve the GVF, could be clearly shown. Comparison with high and low resolution SPOT data shows a linear relationship and higher values for GVF. Squared GVF values were found to be closely correlated with independently derived high and low resolution vegetation cover (fCover), confirming this relationship known from literature. Coefficients of determination (R2), slope and offset of linear relations between squared GVF on one side and the two validation data sets on the other side were 0.69, 0.91, 0.07 and 0.58, 1.27, 0.06, respectively. In addition to the ¿per se¿ value of the derived abundances, validation results indicate that squared GVF may be used as approximation for vegetation cover.
JRC Directorate:Sustainable Resources

Files in This Item:
File Description SizeFormat 
sandtreportgvflatest-vercw-kb.pdf1.69 MBAdobe PDFView/Open

Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.