Due to the close relationship between climate and plant phenology, changes in plant phenological patterns have been used as a surrogate of climate change. We analysed Moderate Resolution Imaging Spectroradiometer (MODIS) images to investigate the onset, offset and length of growing season, as well as spatial and inter-annual patterns of Normalized Difference Vegetation Index (NDVI) across six types of vegetation/land use in Taiwan. Regression models indicate that temperature was moderately to strongly related to NDVI for each of the six vegetation/ land-use types (coefficients of determination (R 2 ) ¼ 0.45-0.86). There was a 1-2 month lag time between changes in temperature and NDVI in the forests that are distributed in mid-to high-elevation areas, but not in low-elevation unirrigated fields, paddy fields and urban areas. The relationship between precipitation and changes in NDVI was only significant for unirrigated fields and urban areas (R 2 ¼ 0.37-0.43). Growing season ended considerably earlier at low elevations than at high elevations, possibly because of the earlier start and more severe dry period in low-elevation areas, such that the length of the growing season was longer in the forests than in the unirrigated fields, paddy fields and urban areas.