Abstract: Using data from 34 Pinus densata permanent sampling plots distributed in Shangri-La in 1987, 1992, 1997, 2002, 2007 and 2012, and time series datasets created based on Landsat images, combined Google Earth Engine and Python to reconstruct time series data with 3 different filtering algorithms. The random forest algorithm is used to estimate the aboveground biomass, and the estimation results of time series data before and after reconstruction are analyzed according to the model evaluation indicators. The results show that the non-parametric model trained by time series data reconstructed by 3 different filtering methods has higher fitting accuracy and prediction accuracy than the pre-filtering time series. The overall root mean square error and relative root mean square error are both better than the pre-filter data, and the ARMIA method performs best. The application of the filtering method does eliminate a large amount of noise and uncertainty carried by the image itself to a certain of extent, which effectively improves the data quality and improves the accuracy of remote sensing estimation of Pinus densata aboveground biomass.