Abstract: Based on NDVI data from 2010 to 2021 in Sichuan Province, three methods were used for fitting reconstruction, including Gaussian function fitting (SG), double Logistic function fitting (DL) and Savitzky-Golay (SG) methods. The applicability evaluation of the reconstruction effect was conducted from spatiotemporal pattern, fidelity, and vegetation type. The results show that all three reconstruction methods could effectively repair abnormally low values, but the SG filtering method was more prone to overestimating the peak value. The high fidelity regions of reconstruction were mainly distributed in the northwest of Sichuan Plateau, the northeastern part of the Sichuan Basin and low altitude regions of the Panxi region and the southern Sichuan Plateau. Among the three land types (forest land, grassland, and farmland), the reconstruction effect of grassland was the best, while the forest land was relatively poor. Among the three reconstruction methods, the DL fitting method had the best effect, while the SG filtering method had poor reconstruction effect.