Factors controlling the silicon isotope distribution in waters and surface sediments of the Peruvian coastal upwelling

We present the first systematic study of the silicon isotope composition in the water column (δ30SiSi(OH)4) and in diatoms (δ30Sidiatom) from the underlying surface sediments in a coastal upwelling region. The surface waters upwelling on the shelf off Peru are mainly fed by southward flowing subsurface waters along the coast, which show a mean δ30SiSi(OH)4 of +1.5‰. The concentration of dissolved silicic acid (Si(OH)4) increases towards the south in these waters and with increasing water depth, suggesting lateral mixing with water masses from the south and intense remineralisation of particulate biogenic silica (bSiO2) in the water column and in the surface sediments. Surface waters in the realm of the most intense upwelling between 5°S and 15°S have only marginally elevated δ30SiSi(OH)4 values (δ30SiSi(OH)4=+1.7‰) with respect to the source Si isotope composition, whereas further north and south, where upwelling is less pronounced, surface waters are more strongly fractionated (δ30SiSi(OH)4 up to +2.8‰) due to the stronger utilisation of the smaller amounts of available Si(OH)4. The degree of Si(OH)4 utilisation in the surface waters along the shelf estimated from the Si(OH)4 concentration data ranges from 51% to 93%. The δ30Sidiatom values of hand-picked diatoms in the underlying surface sediments vary from +0.6‰ to +2.0‰, which is within the range of the expected fractionation between surface waters and diatoms. The fractionation signal in the surface waters produced during formation of the diatoms is reflected by the δ30Sidiatom values in the underlying sediments, with the lowest δ30Sidiatom values in the main upwelling region. The silicon isotope compositions of bSiO2 (δ30SibSiO2) from the same surface sediment samples are generally much lower than the δ30Sidiatom signatures indicating a significant contamination of the bSiO2 with biogenic siliceous material other than diatoms, such as sponge spicules. This shift towards lighter δ30SibSiO2 values by up to −1.3‰ compared to δ30Sidiatom signatures for the same surface sediment samples potentially biases the interpretation of δ30Si paleorecords from sediments with low bSiO2 concentrations, and thus the reconstruction of past Si(OH)4 utilisation in surface waters.