Leaf area index （LAI） is one of the key parameters that describe the plant physical structure. However

the LAI product is consistently discontinuous at spatial and temporal scales due to the contamination of atmospheric factors

which limits its application. In this paper

multi-year historical LAI datasets were used as a priori knowledge to establish the LAI background library

based on which

the improved Savitzky-Golay （SG） algorithm was designed to reconstruct the high quality LAI prof iles. The results indicated that by comparison with a traditional SG algorithm

the new algorithm performed better in aspects of both robustness and efficiency.