Vegetation is an important indicator of terrestrial ecosystem,and is a sensitive indicator for climate change. Although MODIS NDVI has higher accuracy,the length of time-series is limited. Therefore,it is necessary to expand and improve the time-series AVHRR NDVI data sets. According to the linear regression models between MODIS NDVI and AVHRR NDVI for each month,we produced a long time-series NDVI data sets (1981 - 2013). Using the NDVI as an indicator of vegetation activity,this study analyzed vegetation dynamic changes in northwest arid region of China for recent 33 years. The results showed that: 1)MODIS NDVI and LTDR NDVI have linear correlations with a high determination coefficient (> 0.8). 2)In recent 33 years,vegetation activity has been enhancing in whole study area,in which area of the sparse vegetation has been increasing continuously,and area of grassland vegetation has not been changed greatly. 3)The climate of this area has been becoming warmer and wetter than before. There are significant correlation among NDVI data,mean annual temperature and annual precipitation. Vegetation activity is sensitive for climate change,topography and human activities.