A preliminary evaluation of the suitability of a currently operational satellite data archive, namely the NOAA global vegetation index (GVI) product, as a basis for providing information on land cover characteristics for global climate modelling has been undertaken by comparing it to several existing data bases of soil and vegetation cover. It is found that temporal composites of GVI values often continue to change beyond the accepted compositing period of three or four weeks. In extreme cases the addition of another week to the compositing period changes over 90% of the pixels. Selection of GVI class boundaries is found to be rather hard and is probably a strong function of the classification technique employed. The resulting GVI classes are generally explicable by reference to other archival material although this study identifies some specific cases of locational divergence from 'consensus' classification. In these locations monitoring with the GVI may prove useful. More importantly GVI values for very similar ecotypes vary widely from region to region thus rendering global classification very dangerous despite the graphic appeal of coloured maps. In its present form the NOAA global vegetation index is not a useful data source for climate modellers.