This dataset is associated with a research article entitled "Satellite_Observed Changes in Forest Cover over Northern China from 1996_2020". As an important part of the land surface, forest is an important factor affecting global carbon, water cycle and climate change; Fractional Forest Cover (FFC) represents the proportion of forest canopy cover area to the whole pixel observed from the vertical direction. It is an important parameter for monitoring forest resources and is related to forest structure or attributes, such as forest area or forest distribution density. It is a variable representing forest cover on a continuous scale. Based on the remote sensing images of Gaofen-2 and Landsat-8, a model for estimating the FFC in the three-north regions of China is constructed based on the ensemble machine learning method. From 1996 to 2020, the annual FFC products with a resolution of 30 meters covering the three northern regions of China with long time series have been generated. The data values contained in the FFC range from 0 to 100, the "Nodata" value is set to -99, and the data file is provided in Geo-Tiff format.