Vol. 4, No. 1, 52-69, 2005

Abstract
The AVHRR data provides the only long-term synoptic coverage that we have of the globe for use in global change and dynamics studies. At present three temporal datasets (Pathfinder, GIMMS and Fasir) have been developed from the AVHRR data. A major use of such datasets is in the analysis of trends that may be occurring in vegetation. The goal of the work reported in this paper was to conduct trend analysis using the data sets and to compare the results for similarities and differences. The work was conducted in two phases. The first phase derived trends in NDVI over the period of the datasets and compared these trends and their differences. The second comparison was to analyse the temporal correlation at each pixel between the datasets so as to better understand the discrepancies between the datasets.

All three of the datasets showed an average global trend of increasing NDVI over the period 1981 - 2000, with this trend varying between 1.8 - 4.5% depending on the dataset used. The correlation analysis showed that the Fasir and GIMMS datasets are more highly correlated than either are with the Pathfinder dataset. There are significant regional variations from this average global trend, with some of these variations being recorded in a consistent way between the different datasets, and some not. Some of these variations form strong linear anomalies, with the Norwegian and Sahelian linear anomalies being the most obvious. The regional differences between the datasets are of such a magnitude that it is considered that these datasets cannot be used for regional mapping or monitoring tasks that involve estimation unless the user is aware that quite different results are likely to be derived from the different datasets.

ISSN 1729-3782