Land surface phenology (LSP), the satellite-derived seasonal recurring stages of leaf development, influences many ecosystem goods and services (e.g. food production, carbon sequestration, and biodiversity conservation) that are the foundations for human well-being. It is also widely used as a “footprint” of ecological consequences from global climate change. Current knowledge of how climate change impact LSP relies heavily on key phenological transition dates (i.e. start and end of growing season from satellite-based vegetation greenness time series) and highly aggregated climatic variables (growing- and chilling-degree days). Few efforts have been dedicated to model LSP as a continuous and dynamic process through the whole period of leaf development at daily temporal scales. It also remains poor understand how extreme weather and climate change act and interact in influencing LSP. To fill these knowledge gaps, this study implemented a Bayesian hierarchical space time (BHST) framework that incorporates both spatial and temporal dimensions to model the speed of both spring green-up and fall brown-down in 914 ecoregions of conterminous United States from year 1981 to 2014 based on EVI2 dataset from NASA’s Vegetation Index and Phenology (VIP) collection and Daymet gridded meteorological record. The model performed very well with R2 of 0.80 ± 0.17 for spring onset and R2 of 0.72 ± 0.20 for fall senescence. We also found that daily maximum temperature, daily minimum temperature, and accumulated precipitation can accelerate the rate of leaf development for both spring and fall phenology while frost- and drought-related extreme weather events significantly slow down the speed of LSP. The sensitivity of rate of LSP to climate change and extreme weather events also differ across biogeographic gradients and varies among different vegetation types (i.e. deciduous forest, evergreen forest, grassland, shrubland, cropland, and urban vegetation). Findings from this study can improve our understanding of how speed of LSP responds to complex environmental drivers at regional-to-continental scales and thus enhance our capabilities to project future LSP shifts due to future climate change and climate variations.