The utility of mangrove foraminifera, diatoms, and stable carbon isotope and C/N geochemistry in relative sea-level reconstruction in the Pearl River Delta, China

Abstract

We investigated the utility of foraminifera, diatoms, and δ¹³C and C/N geochemistry as proxies for generating late Holocene relative sea-level (RSL) records from Deep Bay on the eastern side of the Pearl River Delta. We described the contemporary distribution of these proxies from 66 samples across subtidal, intertidal (mudflat and mangrove), and upland environments from three transects. Using Partitioning Around Medoids (PAM) cluster analysis, we identified ten distinct, vertically-zoned groups of foraminifera and two groups of diatoms. Canonical Correspondence Analysis (CCA) revealed that tidal elevation was the primary environmental variable controlling foraminiferal assemblages, while salinity, which covaried with elevation, primarily influenced diatom assemblages. We also identified 6 vertically-zoned, environmental zones on the basis of δ¹³C and C/N geochemistry using linear discriminant analysis (LDA). Bayesian transfer functions (BTFs) were developed from modern foraminifera and diatom training sets incorporating priors from geochemically-defined environmental zones and BTF predictions derived from the other microfossil group. These BTFs were applied to a sediment core collected from contemporary mangroves, which began accumulating around 1960 CE. Comparison of our RSL reconstructions with local tide gauge records demonstrated that incorporating informative priors improved the accuracy and precision of RSL reconstructions. Among the approaches tested, microfossil BTF-derived priors yielded greater improvement, producing lower prediction uncertainties compared to those based on geochemical priors. The foraminifera BTF captured all tide gauge observations within its 1σ uncertainty bounds regardless of whether informed by priors, while using diatom BTF priors led to a reduction in Mean Squared Error (MSE) by 121 % (MSE: 0.014 m²), and average 1σ uncertainty by 36 %. The incorporation of foraminifera BTF priors in the diatom BTF did not result in substantial improvements in accuracy but reduced its average 1σ uncertainty by 43 %. Our results demonstrate that foraminifera, diatoms, and δ¹³C and C/N geochemistry together offer robust tools for reconstructing late Holocene RSL in the Pearl River Delta, although their utility in isolation is limited by poor preservation of foraminiferal tests and the influence of planktonic diatom taxa in sediment archives.