Palaeoenvironmental Reconstructions Through Compositional Data Analysis

Abstract

The modern analogue technique has been accordingly revised with compositional data analysis framework. The method adopts the Aitchison distance, obtained from isometric log-ratio coordinates of relative abundances, as a natural measure of similarity among assemblages. The number of analogues from which obtain the estimates was determined through leave-oneout verification of modern assemblages. Mean distances and local outlier factor are considered to evaluate the quality of palaeoestimates. The method has been tested on Atlantic Ocean and Mediterranean planktonic foraminiferal assemblages to reconstruct past sea surface temperatures (SST). In comparison with previous planktonic foraminiferal based reconstructions, the Portugal offshore SST record for the last 210 ka shows a higher coherence with other paleoclimate proxies (i.e. the stable isotope and alkenone records). In comparison with raw data analysis, CoDa-MAT yields lower estimates of the western Mediterranean last glacial period SST