Vegetation indices are radiometric measures of photosynthetically active radiation absorbed by chlorophyll in the green leaves of vegetation canopies and therefore good proxies of the functioning of terrestrial ecosystems. A series of articles, in the early 1980s, showed how to use normalized difference vegetation index (NDVI) data to map land cover and to monitor vegetation changes at continental and global scales. This research opened a whole can of investigations that demonstrate the important role of long-term records of vegetation indices from Earth Observing instruments in monitoring and further our understanding on the varying effects of climate on vegetation. We describe 34+ years of an improved non-stationary 8-km NDVI produced from Advanced Very High Resolution Radiometer (AVHRR) instruments that extends from 1981 to the present. The main difficulty in processing AVHRR NDVI data is to properly deal with limitations of the AVHRR instruments such as calibration loss, orbital drift, volcanic eruptions, etc. This long AVHRR record is comprised of data from two different sensors: the AVHRR/2 instrument that spans July 1981 to November 2000 and the AVHRR/3 instrument that continues these measurements from November 2000 to the present. Complicating among-instrument AVHRR inter-calibration of channels one and two is the dual gain introduced in late 2000 on the AVHRR/3 instruments for both these channels. We have processed NDVI data derived from the Sea-Viewing Wide Field-of-view Sensor (SeaWiFS) from 1997 to 2010 to overcome among-instrument AVHRR calibration difficulties. We use Bayesian methods with high quality well-calibrated SeaWiFS NDVI data for carefully deriving AVHRR NDVI calibration parameters. Evaluation of the uncertainties of our resulting NDVI values gives an error of ± 0.005 NDVI units for our 1981 to present data set that is independent of time within our AVHRR NDVI continuum and has resulted in a non-stationary climate data set.