Abstract: It is crucial to understand vegetation phenology changes and their relationship with climate change at biome-level when projecting regional ecosystem carbon exchange and climate-biosphere interactions. To further understand the relationship between vegetation growth and climatic factors, in this study, we investigated the variation in vegetation phenology and its linkage with climate change on the Chinese Loess Plateau through analyzing the Land Long Term Data Record (LTDR) NOAA/AVHRR Normalized Difference Vegetation Index (NDVI) and concurrent temperature and precipitation during 1982-2011. Firstly, the maximum value composite (MVC) method was used to composite the 10 d LTDR NDVI dataset in order to reduce effects of atmospheric and cloud noise. The Harmonic Analysis of Time Series (HANTS) method of HANTS software was used to filter points which were still affected by cloud noise after the MVC was used composite and reconstruct the NDVI time series datasets. Secondly, the 30-year average seasonal NDVI curves for the whole study area and each vegetation type were calculated. Pixels with yearly mean values below 0.1 were excluded from the analysis to ensure the inclusion of sparsely vegetated areas in the analysis. The relative change ratio of NDVI was then calculated from the 30-year average NDVI seasonal curves. We then used the maximum and minimum values for relative change ratio of NDVI as the threshold for the onset dates of vegetation green-up (the beginning of growing season, BGS) and dormancy (the end of growing season, EGS). Finally, linear least square regression was employed to estimate the trends of phenology. Partial correlation analysis was performed between the EGS/ BGS and mean monthly temperature and total monthly precipitation. The results showed that vegetation phenology in the study area generally commenced on Julian day 96-150 for natural vegetation and 72-112 for artificial vegetation. The vegetation dormancy usually began on Julian day 283-305 for natural vegetation and 291-323 for artificially planted vegetation. Over the study period, the growing season was increased by 39 days across the Chinese Loess Plateau. In spring, sixty six percent of the study area showed an advance in the vegetation green-up while only 39% of the study area experienced an apparent advance. These areas were mainly covered by grass and shrub. In autumn, areas subject to a significant delayed vegetation dormancy occupied 62% of the study region, being located in Gansu, Northern Shaanxi, Inner Mongolia and Northern Shanxi. The BGS and EGS varied with vegetation types. The highest and lowest advances in the advances in the BGS occurred in open shrub land (1.31 d/a) and evergreen needle-leaf forest (0.19 d/a), respectively. The EGS was delayed to a highest degree in Orchard (1.18 d/a) and to a lowest degree in paddy land (0.17 d/a). Across the whole Loess Plateau, changing temperature was the dominating factor driving the vegetation phenology. A warming winter (February) and pre-autumn (September-November) could trigger an earlier onset of spring green-up and a warming in late spring and early summer (May-June) could result in a delayed onset of autumn dormancy. Results also suggested that summer and autumn precipitation played an important role in autumn vegetation dormancy. At biome level, the climate warming may be responsible for the earlier onset spring green-up for open forest, deciduous broadleaf shrub (DBS), meadow, steppe and herbosa. Decreased precipitation may be the major reason for delayed onset green-up in paddy land. In mixed forestry land, a warming winter (December and January) could lead to a delayed spring green-up. The delay of EGS in DBS and herbosa could also be partly explained by the climate warming in spring and early summer. The precipitation in summer and autumn may be responsible for the delay of EGS for meadow, sparse grassland and dry land. The correlations between deciduous broadleaf forest, evergreen needle-leaf forest and climate were not statistically significant either for BGS or EGS, indicating that these two vegetation types may be not sensitive to climate change. This study provided a useful reference for evaluation and protection of ecological environment and establishment of climate models.