THE APPLICATION OF GRAIN-SIZE END MEMBER ALGORITHM TO PALEOENVIRONMENTAL RECONSTRUCTION ON INNER SHELF OF EAST CHINA SEA

In order to assess the difference and applicability of different end-member algorithms in paleoenvironmental reconstruction for the inner shelf of the East China Sea (ECS), six common methods, including nonnegative matrix factorization, eigenvector rotation, hierarchical Bayesian algorithm, grain-size class vs. standard deviation, fitting with theoretical function, and factor analysis, are used to extract the grain-size end members of sediments in core DC1 from the ECS inner shelf. Based on comparison and evaluation of different end members extracted from the above six approaches, their deviations and availabilities are discussed. Two sedimentologically meaningful end members (coarse-grained and fine-grained end members) are deduced by all the six methods. Particle size of mode for the end members from five of the six modeling, except that from factor analysis, are consistent with each other, and the content variations of those end members exhibit fairly uniform downcore pattern along the core DC1. The coarse-grained end members for the five modeling represent transgressive sand deposit, and the fine-grained end members represent fluvial fine-silty deposit. However, the coarse-grained end member from factor analysis indicates storm deposits, and the fine one indicates re-suspended deposits induced by current-wave. Although there are some differences in the particle size distribution and content variation in different end members, the six grain-size end-member approaches have good availabilities for the past environment reconstruction on the ECS inner shelf. The variations in end members are effective to indicate the phase change in hydrodynamic environment caused by the sea-level fluctuation since the last glacial maximum.