Abstract: Remote sensing technology provides long time series of NDVI datasets in studying vegetation changes and has good advantages. At present, relatively in-depth change of vegetation NDVI are rarely studied by using long time series data over 30 years and in larger scale. The paper built the regression of different sourced remote sensing data, GIMMS and MODIS with overlapping year, and then reconstructed a long time series NDVI dataset. Combined the reconstructed NDVI, GLC2000(Global Land Cover 2000)product and temperature and precipitation data of 128 meteorological stations in the study area over 30 years, this paper analyzed spatial and temporal vegetation NDVI changes of different vegetation types in northwest China in recent 30 years(from 1984 to 2013)by using regression analysis and correlation analysis methods. The results are as follows: (1)Reconstructed long time series NDVI dataset has good timing characteristics and can be effectively used in monitoring dynamic changes of vegetation in northwest China in temporal and spatial scale.(2)Within the whole study period, vegetation change in northwest China showed an upward trend. Specifically, vegetation change showed a slight upward trend from 1984 to 1993, a slow decline trend from 1993 to 2002 and an upward trend again from 2002 to 2013.(3)Except for Qilian Mountains in Gansu, southeastern Qinghai, and Qinling in Shaanxi where NDVI change was driven mainly by temperature, vegetation NDVI and temperature in northwest China were weakly correlated and vegetation NDVI in arid and semi-arid regions was negatively correlated with temperature. In northwest, vegetation NDVI and precipitation showed a strong positive correlation, except for some areas with more precipitation, irrigated agricultural or meadows such as south Gansu, west Qilian Mountains and central Shaanxi. Precipitation was the main natural factor that affected the natural vegetation changes.(4)In spatial scale, vegetation NDVI change showed obvious regional differences that NDVI increased obviously in Kunlun Mountains, northern Tarim Basin, Qilian Mountains, western and middle eastern of Qinghai Province, northern Shaanxi and Gansu province, and decreased obviously in arid areas such as Altai Mountains and Ili Kazak Autonomous Prefecture.( 5)NDVI changes of different vegetation types showed various spatial and temporal variations and the correlation of different vegetation types between temperature and precipitation were also various. For temperature, from strong to week, the correlation ranked as: farmland > shrub > meadow > swamp > woodland; for precipitation, the correlation ranked as: farmland > grassland > shrub> woodland > swamp.(6)Based on remote sensing and GIS technology, the article discussed the dynamics changes of vegetation and its relationship between temperature and precipitation via NDVI, which provided relative scientific basis for further study of the response mechanism of vegetation and climate.

Key words: vegetation changes, NDVI, temperature, precipitation, correlation, piecewise linear fitting