Diatom-based inference model for paleosalinity reconstructions in estuaries along the northeastern coast of Argentina

Abstract

The relationship between benthic diatom assemblages and environmental variables was analyzed in 40 sites at three estuaries from the northeastern coast of Argentina by using multivariate analyses. Surface sediment samples were collected along the estuarine gradient and analyzed for diatoms, organic matter, water content and grain size. Water salinity, pH, turbidity and temperature were measured seasonally at every sampling site. Canonical Correspondence Analysis (CCA) was applied in order to recognize which environmental variables correlated most closely with the distributions of diatom taxa. Salinity and sediment properties were identified as the most important variables influencing diatom distribution in the studied estuaries. A two-component Weighted-Averaging Partial Least Squares (WA–PLS) calibration model was developed to infer salinity based on diatom assemblages. The reliability of the model was evaluated by the error of prediction (RMSEP jack-knifed = 4.42‰) and the correlation ( r 2 jack-knifed = 0.75) between observed and diatom-inferred values. The transfer function was applied to fossil diatom assemblages recovered from a Holocene sequence from the Argentinean coast. The application of the model to the fossil record allowed quantitative inferences of salinity changes in southern South America, improving the quality of the inferences performed from diatom analysis.