Abstract Time series of the NOAA Global Vegetation Index data set for the period 1985–1991 were analyzed. The Advanced Very High Resolution Radiometer (AVHRR) cloud-free reflectance data were recalibrated to remove the trends due to the sensor sensitivity change and then normalized to nadir at 45° solar zenith angle using anisotropic top-of-the-atmosphere reflectance models. The results of such a normalization procedure show a substantial reduction in week-to-week variability of the spatial means. The present results demonstrate that relatively stable time series can be obtained by applying empirically derived regional anisotropic models. Thus, the improved data set is potentially more useful for monitoring biosphere-hydrology interactions and for use in numerical climate models.