|Extending the SPOT-VEGETATION NDVI time series (1998-2006) back in time with NOAA-AVHRR data (1985-1998) for southern Africa|Swinnen, E.; Veroustraete, F. (2008). Extending the SPOT-VEGETATION NDVI time series (1998-2006) back in time with NOAA-AVHRR data (1985-1998) for southern Africa. IEEE Trans. Giosci. Remote Sens. 46(2): 558-572. dx.doi.org/10.1109/TGRS.2007.909948
In: IEEE transactions on geoscience and remote sensing. Institute of Electrical and Electronics Engineers: New York, N.Y.. ISSN 0196-2892, more
image processing; spectral response function (SRF); time series
A new consistent long-term normalized difference vegetation index (NDVI) time series at a 1-km2 resolution for Southern Africa that is based on the data from Satellite Pour l'Observation de la Terre VEGETATION (VGT) (1998-2006) and the National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (AVHRR) (1985-1998) has been produced for vegetation-dynamics monitoring purposes. This paper presents the evaluation of the newly processed AVHRR data set, as well as the integration of this data set with the VGT archive. First, the AVHRR processing chain and the resulting AVHRR data set have been investigated with respect to calibration accuracy, cloud masking, and atmospheric and geometric correction. Second, different calibration approaches, spectral response (SR) functions, spatial resolutions, overpass times, and geometries of observation for the VGT and AVHRR data sets have been compared for a common observation period. The application of published correction functions accounting for the SIR differences for both sensors considerably improved the consistency between both data sets. An r2 of 0.85 is obtained between paired samples of the NDVI from the VGT and the newly processed AVHRR archive. After the application of the correction functions, the slope of the regression line between the two NDVI data sets was much closer to the 1: 1 line. The performance of the correction functions differed among vegetation types. The largest reduction in the root-mean-square error between the NDVI of both sensors is obtained from areas with higher biomass. Large parts of the remaining variability are suggested to be attributed to the bidirectional reflectance distribution function effects, as demonstrated by the intersensor NDVI time-series variability versus the intrasensor NDVI time-series variability.