Jia Kun, Yao Yunjun, Wei Xiangqin, Gao Shuai, Jiang Bo, Zhao Xiang
Fractional vegetation cover is a key parameter for characterizing land surface vegetation coverage, and plays an important role in global change research, earth surface processes simulation and hydro-ecological models. Remote sensing can provide the vegetation coverage information and variation trend on different spatial scales, and become the important means in obtaining the information of regional or global fractional vegetation cover. In this paper, the commonly used remote sensing data sources including hyperspectral data, multispectral data, microwave data and LiDAR data for fractional vegetation cover estimation are analyzed, and multispectral data will be the long term main data source for fractional vegetation cover estimation because of the advantage of its easily acquisition, wide coverage and continuous observation advantages. Then, characteristics and advantages of different estiation methods are analyzed, which include the regression model, the pixel unmixing model, the machine learning method, the physical model, the spectral gradient difference and the forest canopy density mapping model. The dimidiate pixel model of the pixel unmixing model is extensively used for its simple form and certain physical significance, and the neural network method is widely used for generating products because of its extendibility and rapidity of calculation. Furthermore, the existing remote sensing data based fractional vegetation cover products are presented. However, the validation results indicate that almost all of the fractional vegetation cover products have underestimation problem and variational estimation accuracy in different regions. Finally, the fractional vegetation cover estimation research prospect is discussed and high spatial-temporal resolution global fractional vegetation cover dataset, multi-source remote sensing data fusion and assimilation method are the future development of fractional vegetation cover estimation using remote sensing data.