Testing the applicability of a benthic foraminiferal-based transfer function for the reconstruction of paleowater depth changes in Rhodes (Greece) during the early Pleistocene

Abstract

We present paleo-water depth reconstructions for the Pefka E section deposited on the island of Rhodes (Greece) during the early Pleistocene. For these reconstructions, a transfer function (TF) using modern benthic foraminifera surface samples from the Adriatic and Western Mediterranean Seas has been developed. The TF model gives an overall predictive accuracy of ~50 m over a water depth range of ~1200 m. Two separate TF models for shallower and deeper water depth ranges indicate a good predictive accuracy of 9 m for shallower water depths (0–200 m) but far less accuracy of 130 m for deeper water depths (200–1200 m) due to uneven sampling along the water depth gradient. To test the robustness of the TF, we randomly selected modern samples to develop random TFs, showing that the model is robust for water depths between 20 and 850 m while greater water depths are underestimated. We applied the TF to the Pefka E fossil data set. The goodness-of-fit statistics showed that most fossil samples have a poor to extremely poor fit to water depth. We interpret this as a consequence of a lack of modern analogues for the fossil samples and removed all samples with extremely poor fit. To test the robustness and significance of the reconstructions, we compared them to reconstructions from an alternative TF model based on the modern analogue technique and applied the randomization TF test. We found our estimates to be robust and significant at the 95% confidence level, but we also observed that our estimates are strongly overprinted by orbital, precession-driven changes in paleo-productivity and corrected our estimates by filtering out the precession-related component. We compared our corrected record to reconstructions based on a modified plankton/benthos (P/B) ratio, excluding infaunal species, and to stable oxygen isotope data from the same section, as well as to paleo-water depth estimates for the Lindos Bay Formation of other sediment sections of Rhodes. These comparisons indicate that our orbital-corrected reconstructions are reasonable and reflect major tectonic movements of Rhodes during the early Pleistocene.

Vertical and lateral distribution of Foraminifera and Ostracoda in the East Frisian Wadden Sea – developing a transfer function for relative sea-level change

“…Kemp & Telford, 2015). Bootstrapping cross-validation (1000 cycles) was used to evaluate the TF performance based on the coefficient of determination (R² boot ), providing an estimation of the grade of the linear relationship between observed and estimated elevations in the training set, and the root mean squared error of prediction (RMSEP), enabling the evaluation of the general predictive capability of the TF (Milker et al, 2017). The most appropriate of the three modelled components was chosen based on the lowest RMSEP and highest R² Boot .…”

“…We expect that a higher number of training samples will result in a better TF performance as most studies base their TF on a minimum of 40 samples, resulting in smaller error ranges (e.g. Kemp et al, 2012;Müller-Navarra et al, 2016, 2017Milker et al, 2017). In a next step we will, therefore, increase the total number of samples in order to fill the elevation gap created by the weak sand flat samples.…”

The Effect of Size Fraction in Analyses of Benthic Foraminiferal Assemblages: A Case Study Comparing Assemblages From the >125 and >150 μm Size Fractions

“…This section has a length of approximately 16 m, and was deposited during the early Pleistocene. It mainly comprises homogeneous and bioturbated marls with intercalations of laminated marls (Milker et al, 2017). Prior to sampling, the weathered surface was removed, and sediment samples of 2 cm thickness each were taken every 10 cm.…”

“…The general trends are fully comparable between the >125 and >150 µm size fractions, and qualitative interpretations are not influenced by the chosen size fraction in which benthic Foraminifera have been counted. Studies for which such qualitative or low-precision quantitative interpretations are sufficient, such as past environmental reconstruction studies (e.g., Herguera and Berger, 1991;Erbacher et al, 1999;Gooday, 2003;Kouwenhoven and van der Zwaan, 2006;Milker et al, 2017), where errors are generally rather large, will not have to deal with any negative impacts on their performance. Rather, they can fully compare published studies and use them as a basis for their own, regardless of whether those studies employed the >125 µm or the >150 µm size fraction for their benthic foraminiferal assemblage counts.…”

“…Studies of benthic Foraminifera are numerous, because benthic foraminiferal assemblages are widely used to characterize recent and to reconstruct past environmental conditions, employing various tools such as diversity analyses (e.g., Debenay, 1991;Almogi-Labin et al, 1996;Armynot du Châtelet et al, 2004;Badawi et al, 2005;Mojtahid et al, 2009), multivariate analyses (e.g., Frontalini et al, 2009;Mojtahid et al, 2009;Cosentino et al, 2013;Minhat et al, 2016), and transfer function approaches (e.g., Horton et al, 1999;Gehrels, 2000;Leorri et al, 2010;Kemp et al, 2015;Milker et al, 2017).…”