Two types of algorithms, empirical and physical methods, have been widely used to estimate FAPAR from satellite remote sensing data (Knyazikhin et al., 1998a&b; Gobron, 2006; Baret et al., 2007&2013). The empirical methods use the statistical relationships between FAPAR and vegetation indices to estimate FAPAR. The functional relationships between FAPAR and vegetation indices have been widely investigated using in-situ observations and canopy radiative transfer models (Daughtry et al., 1992; Goward & Huemmrich, 1992; Myneni & Williams, 1994; Chen, 1996; Knyazikhin et al., 1998b; Myneni et al., 2002; Gitelson et al., 2014). However, these statistical relationships are influenced by the vegetation type, soil background and imaging geometrics (Myneni & Williams, 1994; Chen, 1996; Xiao et al., 2015; Peng et al., 2012; Zhang et al., 2014; Liu et al., 2017). FAPAR estimates using physical methods, on the other hand, are based on the inversion of canopy radiative transfer models. Most currently available global FAPAR products are generated by physical methods (Knyazikhin et al., 1998a&b; Myneni et al., 2002; Gobron et al., 2006; Plummer et al., 2006; Baret et al., 2007).