A model for partitioning the particulate scattering coefficient into the contributions of suspended mineral particle and organic particle was proposed based on the measured data. The independent variables, i.e. the concentrations of mineral particles and organic particles in this study, were used to determine the mass-specific scattering cross section with the concurrent total suspended particulate scattering coefficients collected during the field trip in Poyang Lake 2009. Results show that the scattering spectra of inorganic particles and organic particles can be successful derived by the proposed model, and the reconstructed total particulate scattering coefficients are in better agreement with the measured values by the ordinary least square linear regression. For the whole South Poyang Lake, mean absolute percentage errors between the measured scattering coefficients and reconstructed value were less than 25% over the main remote sensing effective wavebands such as 440, 532, 555 and 676 nm. A remarkable lower predicted error, which can be controlled within 15%, were found at all stations with higher concentration of total suspended matters, while the spectral partitioning is less efficient at stations with total suspended particle concentration less than 15 mg·L-1. Particulate scattering spectrum retrieved by RMA shows that illite and montmorillonite are the major constituents of inorganic matters which dominate the light scattering properties of Poyang Lake. It is possible that scattering spectrum partitioned by the model could infer the major effective components in waters, and could be used to predict particulate scattering properties for highly turbid waters.