Extraction of climatic signals from fossil organic compounds in marine sediments up to 11.7Ma old (IODP-U1318)

Abstract

This study focuses on the extraction of climate signals and processes using a combined approach which includes the analysis of a high number of lipid molecules in marine sediments, and the chemometric analysis of the acquired data. Neutral and acidic fractions of marine sediments from site IODP-U1318 (south-west of the UK, Porcupine Seabight) were quantified by GC-MS. The alkenone unsaturation index, U(k')37, was estimated from the composition of C37 alkenones and it was then used for the estimation of sea surface temperatures (SST) for reference. Principal component analysis (PCA), explained 77.45% of the total data variance, and differentiated neutral fraction GC-MS total ion current (TIC) profiles according to SST values of the different sediment sections. GC-MS TIC chromatograms were correlated to sea surface temperatures (SST) by partial least squares regression (PLSR). The compounds more robustly in line with SST values at each sediment section explained 93% of the SST variance and they were identified using the variable importance in projection (VIP) scores method. The proposed approach enables an objective identification of organic compounds sensitive to SST variability throughout complete chromatographic profiles. As a result of this multivariate unbiased approach, lipid composition of sediments was differentiated between compounds of marine (long chain n-alkanes, long chain n-alkan-1-ols) and terrestrial (short chain n-alkan1-ols, alkenols, cholesterol, squalene) origin, whose concentrations were directly and inversely correlated to SST, respectively.

Comparison of the variable importance in projection (VIP) and of the selectivity ratio (SR) methods for variable selection and interpretation

“…There were three variables with absolute values of r near to one ( p < 0.005), which were rejected by SR and not by VIP (variables 4, 15 and 73 of concentration X2 matrix, in Table S5). Two of these variables, identified as nonadecane and tetracosane, were relevant for the interpretation of y variance (SST changes) . All other variables selected by VIP and not by SR methods presented an absolute r value between 0.62 and 0.72 ( p < 0.005).…”

“…Previous to their analysis, the chromatographic profiles of the 10 samples were baseline corrected using asymmetric least squares (AsLS) method , peak aligned using correlation optimised warping (COW) algorithm and mean‐centred. Further details of experimental methodology and of data pre‐treatment are given in Farrés et al . Annual mean SST values were reconstructed using a global core‐top calibration method and the alkenone unsaturation index U k' 37 and mean‐centred prior to PLSR analysis.…”

“…Additionally, a second X matrix of predictor variables, X2 (refer to Table S2), was obtained using 77 fossil compounds recognised in the original chromatogram and identified by comparison of their corresponding measured spectra to library data and synthetic standards . Their chromatographic areas were then integrated to estimate the relative concentration of each of the identified organic compounds.…”

“…VIP and SR are two of the most frequently used methods in chemometrics for variable selection. VIP scores selection method has been extensively used in different fields and thus for a variety of data types . Recently, Galindo‐Prieto and coauthors proposed a new VIP approach for orthogonal projections to PLS latent structures to enhance model interpretability.…”

“…In the first data set (water quality data set), physiochemical water quality parameters were related to sensorial data of the same water samples . In the second data set (climate data set), GC‐MS chemical (organic compound) profiles from fossil sea sediment samples were correlated with sea surface temperature (SST) changes . And, in the third data set (transcriptomic data set), genes from Daphnia magna female samples exposed to sublethal doses of the selective serotonin reuptake inhibitors (SSRIs: fluoxetine and fluvoxamine) were correlated to their corresponding total offspring production .…”