To understand response of terrestrial ecosystem to climate changes, it is important to study variations in vegetation greenness (VG) under the changing climate. We firstly extend the Advanced Very High Resolution Radiameter (AVHRR)-based Normalized Difference Vegetation Index (NDVI) from 2006 to 2011 by calibrating the Moderate Resolution Imaging Spectroradiometer-based NDVI. Using hierarchical clustering and empirical orthogonal function (EOF) analysis, this study picked out main models of variations of autumn VG (September to November) which was indicate by NDVI from 1982 to 2011. Then, by using Singular Value Decomposition (SVD), we integrated temporal variations and spatial variations of NDVI and temperature to reveal climate background of variations in VG. We found that the MODIS-NDVI captures the main variations of VG as the AVHRR-NDVI; therefore, it is reliable to extend the AVHRR-NDVI with the MODIS-NDVI. Both of cluster analysis and EOF analysis show the variations in autumn VG had two main models. One (model I)was featured by an increasing trend of VG from 1982 to 2011. This model spreads across the North America land, northern part and (west, east) two ends of Eurasia. The mean rate of increase in NDVI was ~0.014 per decade. The other model (model II) was featured by turn of NDVI variations. This model mostly spread in the center area of Eurasia. The turn point of NDVI variations was about 1994. From 1982 to 1994, the NDVI increased at the rate of 0.027 per decade; while, from 1995 to 2011, the NDVI decreased at the rate of 0.017 per decade. EOF analysis as well as illustrates that above two models could explain 21.1% and 8.1% of total variations of NDVI, respectively. The SVD analysis illustrated that variations in NDVI matched well with variations in temperature. In the Model I area, the temperature changes were featured by an increasing trend. The time coefficients of NDVI and temperature had a strong correlation (R=0.82). In the model II area, the temperature changes were featured by a turn point. Before the turn point, temperature increased; after the turn point,temperature decreased. The time coefficients of NDVI and temperature also had a strong correlation (R=0.92).These results suggest that the persisting greening in the Model I area occurred in the climate warming background and conversion from greening to browning in the Model II area co-occurred with the climate changes from warming to cooling. These findings demonstrate that the main models of variations in VG might result from temperature variations. In the continental scale, the spatial heterogeneity of temperature variations resulted in spatial heterogeneity of VG variations.