Super typhoon Haiyan makes landfall
Super typhoon Haiyan
Subscribe here to receive special images

VEGETATION - 2000

Lake Maggiore - Italy, 3-6 April 2000


INTERCOMPARISON OF DEKADAL VEGETATION INDEX FROM NOAA/AVHRR AND SPOT4/VEGETATION OVER THE IGAD REGION


SCOT, 8-10 rue Hermes, 31526 Ramonville Cedex , France
Email : bicheron@scot.cnes.fr

Paper (pdf file, 878 k)

Numerous operational systems (MARS, World Fire Web) are supplied with data acquired by the AVHRR instrument. An important archive has been thus constituted but with a critical data quality. Since 2 years, the VEGETATION instrument delivers data with a quality largely greater, but do no offer AVHRR temporal depth. In this framework, the AVHRR/VEGETATION "intercomparison" has all its interest, since the determination of transfer functions between both instruments products will allow to constitute a virtual VEGETATION archive. This study will thus allow on one hand to improve the quality of the existing archive and on the other hand to guarantee a transition and a substitution of AVHRR data with VEGETATION data in operational projects. We will focus more precisely on the comparison of temporal dekadal NDVI series derived from both instruments.

The study area locates in Oriental Africa (IGAD region) over a region of 1500 x 1500 km² (19°N, 25°E ; 5°N, 39°E). This area covers a part of Eritrea, Sudan, and the eastern part of Ethiopia, offering a broad vegetation gradient from North to South. The land cover is characterised by 27 classes following the USGS land cover classification realised with LANDSAT data. Over our study area, 10 vegetation classes are covered (grazing, cropland and grazing, grazing and cropland, Shrub Savannah, Other types of savannah, Forest, Dry vegetation), more two non vegetation classes (Urban, Water).

The first step of the study consist of processing the AVHRR and VEGETATION Dataset to make them comparable. This data processing includes geometrical and atmospherical standard corrections, and the synthesis of NDVI according to the MVC method. According to a regular sampling by step of 1°, we extract the NDVI temporal signature from AVHRR and VEGETATION Dataset for 175 pixels belonging to several classes. After an analysis of the different signatures, we propose different filtering techniques to discard erroneous values taking into account or not the temporal evolution of the pixel. We present then the processing of simple regression coefficients between AVHRR and VEGETATION NDVI.