ABSTRACT                    

Grain-size sensitive component separation (GSCS) methods are pivotal for paleoenvironmental reconstruction but remain underexplored in tectonically active deep-sea seamount settings like the central South China Sea (SCS).This study presents the first systematic comparison of three GSCS techniques, namely standard deviation (STD), endmember modeling (EMM), and principal component analysis (PCA). These methods are applied specifically to sediment core GT-06 recovered from the Zhongnan Seamount in the central South China Sea. By integrating grain-size unmixing with complementary proxies, including magnetic susceptibility (MS) and loss on ignition (LOI), we assess the resolving power of each method and relate the resulting components to site-specific sedimentary processes (e.g., summer monsoon, volcanic activity, seamount collapse-induced turbidity currents). Key results show that all methods identified three dominant grain-size ranges (clay, silt-sand, coarse sand) with robust inter-method correlations. Critically, EMM and PCA uniquely resolved a volcanic-derived silt component (EM2) that was undetectable using the STD method, highlighting a key limitation of STD in complex settings. Four geologically meaningful end-members were identified: (1) fine-grained terrigenous clay linked to the East Asian summer monsoon (EASM), (2) volcanic detritus reflecting Quaternary submarine eruptions, (3) siliceous biogenic debris indicative of monsoon-modulated productivity, and (4) coarse calcareous fragments associated with seamount collapse-induced turbidity currents. The results highlight the superior ability of EMM/PCA to resolve complex signals, whereas STD serves as an efficient yet limited tool for first-order screening. Together, these methods form a site-adapted framework for tectonically active deep-sea seamounts: EMM/PCA enabling fine-scale interpretation of monsoon-volcanic-turbidity interactions, and STD supports rapid large-dataset comparison. This workflow improves the reliability of paleoenvironmental reconstructions in mixed-signal settings like the Zhongnan Seamount.