| 28 Nov 2025
New planktonic foraminifera-derived transfer function for the South Atlantic Ocean: Palaeoceanographic implications for the Brazil- Malvinas Confluence
Abstract. Planktonic foraminiferal assemblages are extensively used for reconstructing sea surface temperature through the application of transfer functions. Nonetheless, it has been observed that several parameters present throughout the water column also influence compositional changes within these assemblages. Selection of driving factors and evaluation of transfer function performances are method-specific processes that require the combination of prior ecological knowledge and objective variable selection approaches. In this study, we compiled a 171 core-top samples dataset of planktonic foraminifera and productivity-related variables to quantify the relationship between the assemblages and modern productivity conditions in the South Atlantic Ocean. Multivariate statistical analyses revealed that planktonic foraminiferal species were related to austral summer nitrate, explaining an independent and significant proportion of variance in the species data. We evaluated different prediction models, and estimated their performances considering spatial autocorrelation. The calibration model Weighted Averaging with tolerance downweighting and inverse deshrinking (WATOL_inv) with h-block cross validation showed a regression coefficient of r2cv = 0.938, with a root-mean-square error of prediction RMSEP = 1.578 սmol l-1. The resulting transfer function was applied then to sediment core GeoB2806-4 (~37° S – 53° W; 3500 m) in order to reconstruct variations of summer nitrate concentration during the Holocene. Our reconstructed summer nitrate shows a general decreasing trend from early to mid-Holocene associated with increased biological uptake, and a later increase of it towards the late Holocene. We suggest that changes in summer surface nitrate concentration are linked to the latitudinal shifts of the Brazil-Malvinas Confluence. Understanding the displacement of the Confluence, and the associated shifts in the upper layers’ nutrient availability, is crucial to evaluate the implications of these changes on the local to regional ecosystem dynamics and trophic structure, particularly when considering future climate projections.
Albarracin et al present a transfer function model that relates planktonic foraminifera abundance to seawater nitrate concentration. The model is based on a set of core top sediment species assemblages from the South Atlantic Ocean. Before building the transfer function model the authors assess which nutrient and productivity related variables could serve as independent predictors of the species assemblage composition and they evaluate several different models. The final model is used to reconstruct the nitrate concentration during the Holocene using fossil assemblages from a core in the Brazil-Malvinas Confluence.
Whereas there is ample evidence that on a global and basin-wide scale planktonic foraminifera species composition is best predicted by temperature (Morey et al. 2005; Rillo et al. 2021), there are also indications that on smaller spatial scales and in isolated basins the drivers of species assembly differ (Azibeiro et al. 2023). It is hence valuable to assess the influence of other variables than temperature on species assemblage composition and the study by Albarracin is therefore of potential interest for the readership of Climate of the Past. However, there are several issues including one fundamental methodological flaw in the study that prevent me from recommending the study for publication in its present form.
My main concern is that the authors do not demonstrate that the nitrate concentration is more important than temperature (or other physical environmental variables) for planktonic foraminifera species assembly. The authors restrict their analysis to nutrient and productivity related variables, but a priori ruling out the influence of temperature is no evidence that it plays no role and that an (independent) reconstruction of nitrate concentration is possible. The study by Lessa et al in itself is not sufficient proof as it is based on plankton tows and the implications for sedimentary assemblages, which are vertically and temporally integrated, is not yet clear. The authors may be right, but they need to redo their analysis with a wider suite of environmental variables (including temperature) to prove this. Only if this analysis reveals that nitrate is an important predictor of assemblage composition that can be disentangled from the temperature effect can they proceed with the reconstruction. In this regard it is surprising to see that the authors compare their reconstructions of nitrate concentration with reconstructions of seawater temperature based on exactly the same species assemblages, but there is no indication that is possible to derive independent reconstructions of both variables.
My second concern is with the selection of the samples included in the calibration dataset. In the method section it is written that they use over 300 samples, but in the results fewer than 200 are mentioned. It is unclear why some samples were excluded (and what the effect is). On a more fundamental level, what is the rationale of working with a subset of the available data (Siccha and Kucera 2017)? Please explain as this likely influences the transfer function model and the reconstruction.
My third major concern is about what nitrate concentrations actually tell about the environment. The authors seemingly use nutrient concentrations and productivity interchangeably (e.g. L 545 “4.4 Holocene paleoproductivity reconstruction in the BMC (WSA)”), but they are different. The nutrient concentration reflects what is left over after utilisation by primary producers and it is hence not related to primary production in a straightforward way. The fact that nitrate concentration and chlorophyll-a concentration (probably a better indicator of productivity) plot perpendicular to each other in the RDA plots underscores this. So if it is possible to reconstruct nitrate concentrations in the South Atlantic, then a reinterpretation of the results is still needed.
Minor comments
L1: the title is vague as to what this transfer function is actually for.
L109: “In order to…” I don’t understand the reasoning here.
L111: please describe the selection process (see also above).
L124: “The temporal ranges…” please describe better. I can see how temperature is affected by global warming, but the effect on the other variables is less clear. For most variables there is also observational and climatology data available for earlier times, so why not use those if the effect of global change is a concern?
L126: “Therefore, we analyzed seasonal averages…” I don’t understand the “therefore” here. How does this reduce the influence of global warming?
L131: “Sediment core GeoB2806-4 (37°50’S - 53°08.6’W, 3500 m depth; García Chapori et al., 2015) was used for testing the transfer function developed here.” Testing implies that the true nitrate concentrations during the Holocene were known. Please reword as this is not a test, but an application.
L153: “morphotypes” should be subspecies (Morard et al. 2019).
L154: why were rare species excluded from the analysis?
L155: how does log transformation standardise the variance?
L203: which dissimilarity metric was used. And why five analogues?
Transfer function performance: I don’t understand why the authors assess the transfer function not immediately using h-block crossvalidation and neither why they don’t evaluate all models in this way. Perhaps a different model than the WATOL_inv one performs better with h-block CV than shown in Table 4? Also, is the evaluation of the performance based on the R2 and RMSE sufficient? I realise that this is the usual approach, but the authors show two other metrics (related to bias), but don’t use them.
Azibeiro, Lucia A., Michal Kučera, Lukas Jonkers, Angela Cloke-Hayes, and Francisco J. Sierro. 2023. “Nutrients and Hydrography Explain the Composition of Recent Mediterranean Planktonic Foraminiferal Assemblages.” Marine Micropaleontology 179 (March): 102201.
Morard, Raphaël, Angelina Füllberg, Geert-Jan A. Brummer, et al. 2019. “Genetic and Morphological Divergence in the Warm-Water Planktonic Foraminifera Genus Globigerinoides.” PloS One 14 (12): e0225246.
Morey, Ann E., Alan C. Mix, and Nicklas G. Pisias. 2005. “Planktonic Foraminiferal Assemblages Preserved in Surface Sediments Correspond to Multiple Environment Variables.” Quaternary Science Reviews 24 (7-9): 925–950.
Rillo, Marina C., Skipton Woolley, and Helmut Hillebrand. 2021. “Drivers of Global Pre‐industrial Patterns of Species Turnover in Planktonic Foraminifera.” Ecography, ahead of print, December 13. https://doi.org/10.1111/ecog.05892.
Siccha, Michael, and Michal Kucera. 2017. “ForCenS, a Curated Database of Planktonic Foraminifera Census Counts in Marine Surface Sediment Samples.” Scientific Data 4 (August): 170109.