Subducted sediments play an important role in arc magmatism and crust-mantle recycling. Models of continental growth, continental composition, convergent margin magmatism and mantle heterogeneity all require a better understanding of the mass and chemical fluxes associated with subducting sediments. We have evaluated subducting sediments on a global basis in order to better define their chemical systematics and to determine both regional and global average compositions. We then use these compositions to assess the importance of sediments to arc volcanism and crust-mantle recycling, and to re-evaluate the chemical composition of the continental crust. The large variations in the chemical composition of marine sediments are for the most part linked to the main lithological constituents. The alkali elements (K, Rb and Cs) and high field strength elements (Ti, Nb, Hf, Zr) are closely linked to the detrital phase in marine sediments; Th is largely detrital but may be enriched in the hydrogenous Fe-Mn component of sediments; REE patterns are largely continental, but abundances are closely linked to fish debris phosphate; U is mostly detrital, but also dependent on the supply and burial rate of organic matter; Ba is linked to both biogenic barite and hydrothermal components; Sr is linked to carbonate phases. Thus, the important geochemical tracers follow the lithology of the sediments. Sediment lithologies are controlled in turn by a small number of factors: proximity of detrital sources (volcanic and continental); biological productivity and preservation of carbonate and opal; and sedimentation rate. Because of the link with lithology and the wealth of lithological data routinely collected for ODP and DSDP drill cores, bulk geochemical averages can be calculated to better than 30% for most elements from fewer than ten chemical analyses for a typical drill core (100-1000 m). Combining the geochemical systematics with convergence rate and other parameters permits calculation of regional compositional fluxes for subducting sediment. These regional fluxes can be compared to the compositions of arc volcanics to asses the importance of sediment subduction to arc volcanism. For the 70% of the trenches worldwide where estimates can be made, the regional fluxes also provide the basis for a global subducting sediment (GLOSS) composition and flux. GLOSS is dominated by terrigenous material (76 wt% terrigenous, 7 wt% calcium carbonate, 10 wt% opal, 7 wt% mineral-bound H2O+), and therefore similar to upper continental crust (UCC) in composition. Exceptions include enrichment in Ba, Mn and the middle and heavy REE, and depletions in detrital elements diluted by biogenic material (alkalis, Th, Zr, Hf). Sr and Pb are identical in GLOSS and UCC as a result of a balance between dilution and enrichment by marine phases. GLOSS and the systematics of marine sediments provide an independent approach to the composition of the upper continental crust for detrital elements. Significant discrepancies of up to a factor of two exist between the marine sediment data and current upper crustal estimates for Cs, Nb, Ta and Ti. Suggested revisions to UCC include Cs (7.3 ppm), Nb (13.7 ppm), Ta (0.96 ppm) and TiO2 (0.76 wt%). These revisions affect recent bulk continental crust estimates for La/Nb and U/Nb, and lead to an even greater contrast between the continents and mantle for these important trace element ratios. GLOSS and the regional sediment data also provide new insights into the mantle sources of oceanic basalts. The classical geochemical distinction between 'pelagic' and 'terrigenous' sediment sources is not valid and needs to be replaced by a more comprehensive understanding of the compositional variations in complete sedimentary columns. In addition, isotopic arguments based on surface sediments alone can lead to erroneous conclusions. Specifically, the Nd/Hf ratio of GLOSS relaxes considerably the severe constraints on the amount of sediment recycling into the mantle based on earlier estimates from surface sediment compositions.
