Numerous researches have suggested that leaf phenological cycles delimit the plant growing season and thus have a large impact on terrestrial ecosystems process. Leaf phenological cycles contain certain amount of inter- and intra- specific variation therefore estimating such variation at larger scales is importance to understand the potential effect of global climate change on the phenological responses of plants and its consequences on ecosystem process. However, researches that directly estimated the species level leaf phenology at large scale are still limited. In this study we estimated latitudinal gradient of leaf onset behavior using simulation model and satellite observed LAI data. We employed particle filter, a non-parametric and computationally intensive parameter optimization method, to calibrate the parameters of simulation model. We then estimated the leaf emerged day (LED) and total time for leaf completely out (TLO) for 4 representative Japanese deciduous tree species. Results showed that 1) LED was delayed as latitude increase 2) TLO decrease as latitude increase. 3) LED was strongly affected by spring temperature. 4) TLO was affected by both spring temperature and temperature fluctuation during spring season but there are certain inter-specific variation. Phenological responses of trees to future climate change would differ among tree species.