Palaeo-productivity record from Norwegian Sea enables North Atlantic Oscillation (NAO) reconstruction for the last 8000 years

The North Atlantic Oscillation (NAO) is the dominant driver of atmospheric variability in the North Atlantic region, influencing westerlies and thereby surface ocean circulation, mixing, and current speeds. Defined as the sea-level pressure difference between the Icelandic Low and the Azores High, a positive (negative) NAO phase yields warmer (colder), wetter (drier) conditions in north-western Europe, with strongest expression in winter (December–March). While numerous palaeo-reconstructions of the NAO exist, only few are based on open-ocean sediment records due to challenges in achieving high temporal resolution and precise age control. Overcoming these challenges could provide crucial insights into ocean–atmosphere linkages and the broad-scale, non-stationary character of the NAO. Prior work suggests the NAO influences phytoplankton dynamics via air–sea fluxes and circulation in the eastern North Atlantic. Biogenic CaCO3 in sediments can reflect palaeoceanographic changes, but its relation to sea-surface conditions is complex because sources (coccolithophores, planktonic foraminifera) and dilution by terrigenous input and dissolution play roles. Productivity of CaCO3-secreting algae is controlled by light, nutrients, and temperature, with seasonal blooms in spring and autumn, and also exhibits decadal to millennial climate-driven variability. This study addresses: (i) whether ITRAX-derived Ca/Fe variability can serve as a proxy for changes in sea-surface primary productivity in the south-eastern Norwegian Sea, and (ii) how this proxy reflects known oceanic and atmospheric variability, particularly the NAO. Two well-dated, high-resolution sediment cores from 850 and 960 m water depth on the south-eastern Norwegian margin spanning the last 8000 years overlap with late 20th-century observational time series, enabling calibration of proxy to observations.

Several studies have numerically and empirically linked NAO variability to phytoplankton productivity changes in the eastern North Atlantic through altered mixing and circulation. Sedimentary biogenic CaCO3 has long been used to infer palaeoceanographic changes, though its relationship to surface conditions depends on varying contributions from coccolithophores and foraminifera, dilution by terrigenous material, and potential dissolution. In high-latitude sediments, coccoliths commonly contribute a large fraction (often 40–60% or more) of CaCO3. Coccolithophore distributions, notably Coccolithus pelagicus and Emiliania huxleyi in the Nordic seas, are influenced by nutrients and irradiance, with peak concentrations in summer and seasonal degradation of preservation. Prior regional work indicates CaCO3 flux at these latitudes is dominated by coccoliths (especially C. pelagicus due to larger size), suggesting potential for using Ca-related signals as productivity proxies in suitable sedimentary contexts.

Two sediment cores from the south-eastern Norwegian Sea were analyzed: GS13 (CALYPSO piston core; 63.64°N, 05.51°E; 960 m water depth; 1777.7 cm marine deposits recovered) and P1-003 (63.76°N, 05.25°E; 850 m; spliced MultiCore plus SelanticCore totaling 670.3 cm). The region is influenced by the Norwegian Atlantic Slope Current and Front Current; at the core depths (700–1200 m) sediments are pelitic muds reflecting low-energy conditions. Core stratigraphies intersect Storegga Slide debris dated to ~7650 ± 250 14C yr BP; marine mud overlies slide debris in GS13. Grain size is predominantly silt and clay (<63 µm ~98.5%) with very fine sand composing most of the >63 µm fraction; grain-size variability is small and shows no clear relation to Ca/Fe. Geochemical analyses: ITRAX XRF core scanning (Mo tube) on GS13 and P1-003SC at 500 µm resolution (10 s count time; 100 s for P1-003 MC) produced semi-quantitative elemental counts. To mitigate closed-sum constraints and lithogenic dilution, Ca counts were normalized to Fe (similar results with Ti or K); Ca/kcps yielded comparable patterns. Water content increased modestly (50–60%) down-core; immediate scanning reduced water film effects; Cl comparisons supported minimal bias. Subsamples from GS13 were taken at 5 cm intervals (continuous in top 2 m), treated with H2O2, and wet-sieved at 63, 125, 150, and 1000 µm. Foraminiferal assemblages (150–1000 µm) averaged 64 tests g−1 (67% Neogloboquadrina incompta), implying foraminiferal CaCO3 contributes <0.1% per gram of dry sediment; thus CaCO3 is dominated by coccoliths, mainly Coccolithus pelagicus. In P1-003, absolute concentrations of CaCO3, CaO, and Fe2O3 were measured by wavelength-dispersive XRF; TIC was measured to estimate %CaCO3 (CaCO3 = TIC×8.33). Chronology: Radiocarbon dating of planktonic foraminifera (Neogloboquadrina incompta, Globigerina bulloides) calibrated with Marine13 and local ΔR; identification of three Icelandic tephras (Hekla 1947, Katla 1918, Askja 1875) via shard counts in 63–125 µm; 210Pb/137Cs dating from continuous 2-cm samples in top 50 cm. An initial Bayesian age model (rbacon v2.3.6) was constructed, then refined by correlating GS13 Ca/Fe to the well-dated nearby core P1-003 via 33 tie-points to reduce uncertainty post-1850 AD. Final age-model ensemble comprises ~19,000 equally probable realizations; 2σ uncertainty: ~2–13 years (last 150 years), ~2–24 years (last 200 years), ~17–110 years between 0–1800 AD, and ~58–190 years older than 0 AD/BC. Statistical analysis: Time series were linearly resampled (annual) using R; cross-correlations were computed across all 19,000 age models allowing ±15-year lags (up to ±50 for long series), yielding ~600,000 combinations. Significance assessed via Pearson correlation with p<0.05, adjusted for autocorrelation using AR(1) and Bartlett correction; results summarized as median R, best-fit R, lag distributions, and % of significant models. Observational datasets: CPR survey standard area B4 (upstream of GS13) total phytoplankton (sum of dinoflagellate and diatom counts) used as a proxy for primary productivity; during overlap (1999–2016), coccolithophore counts covary strongly with total phytoplankton (r=0.81, summer), supporting the use of total counts as an overall productivity metric. OWSM dissolved O2 (1–50 m) and temperature (1948–2008) provided additional productivity-related constraints; seasonal means computed for DJFM, MAMJ, JJAS, SOND, discarding seasons with >1 missing month. Proxy calibration considered core-top median age ~1993 AD, providing ~40-year overlap with instrumental series. NAO index: seasonal NAO (DJFM focus) from early instrumental pressure observations (1824–1992; 1950–1992 high-quality subset) used for validation. For palaeo-NAO reconstruction, the Ca/Fe series was detrended by removing low-frequency (>500-year) components; bandpass filters (10–500 years; and high-pass <100 years) were applied to define NAO_Ca/Fe for comparisons to published reconstructions (NAOgrowth, NAOmed, NAOPC1) and a functional palaeoclimate network ‘degree of belief’ NAO phase record.

• Age control and sediment context: Chronological uncertainties are ~2–13 years for 1992–1850 AD and 17–190 years from 1850 AD to 8000 a BP. Sediments are pelitic muds with minimal grain-size variability and little evidence for current-speed-driven terrigenous flux changes at the site. Foraminifera contribute <0.1% of sediment mass; coccoliths dominate CaCO3, with Coccolithus pelagicus the principal contributor. Scanning electron microscopy showed no dissolution features; high sedimentation rates further reduce post-depositional dissolution concerns.
• Ca/Fe as CaCO3 and productivity proxy: Ca/Fe tracks bulk CaCO3 variability; normalization choices (Fe vs. kcps, Ti, K) yield consistent patterns, indicating limited terrigenous influence. Ca/Fe shows distinct peaks around 5000, 3000, 2000, and 0 a BP.
• Calibration to productivity observations: Cross-correlation with CPR B4 summer (JJAS) total phytoplankton yields significant correlations in ~84% of age models with lags of 0–1 year in ~70%; median R=0.54, best fits >0.80. Correlation with OWSM dissolved O2 (spring–early summer, MAMJJA) shows median R=0.40 (best age models), consistent with spring bloom timing.
• NAO relationship (instrumental era): For 1950–1992, Ca/Fe anti-correlates robustly with winter NAO (DJFM): median R=−0.48 (best R=−0.73), with modal lag 0–1 year. Co-variability decreases for spring, summer, autumn. Extending to 1824–1992 weakens the fit (median R=−0.20; best R=−0.44), likely due to increased chronological/proxy uncertainties and NAO non-stationarity.
• Other climate modes: Significant co-variability detected with the Atlantic Multidecadal Oscillation (AMO S2; 1572–1992: median R=0.40; best 0.56) and Subpolar Gyre index (1950–1992: median R=−0.60; best −0.83), suggesting additional climate influence on the proxy.
• Multi-centennial to millennial NAO reconstruction: After removing low-frequency (>500-year) components, the NAO_Ca/Fe index captures annual to multi-centennial variability over 8000 years. High-frequency components (<100 years) correspond well with published NAO reconstructions over the last millennium: NAOgrowth (1700–1992: median R=0.55; best 0.70; 1000–1992: median 0.28; best 0.55), NAOmed (1700–1989: median R=−0.27; best −0.49), and NAOPC1 with a notable lag (~39 years). NAO_Ca/Fe also aligns with a functional palaeoclimate network ‘degree of belief’ NAO-phase record; a sign-test shows ~50% agreement over the last 1000 years, increasing to >77% when a 25-year lag is applied to the network record.
• Holocene perspective: The NAO_Ca/Fe amplitude varies through time, with generally lower amplitudes around the Holocene Thermal Maximum and Medieval Warm Period. The reconstruction does not show a distinct prevailing positive NAO phase during the Medieval Warm Period. Overall, the Ca/Fe-based record provides a sub-decadal to multi-decadal perspective on NAO variability over the last 8000 years.

The study demonstrates that Ca/Fe from high-sedimentation-rate, open-ocean sediments in the south-eastern Norwegian Sea primarily reflects coccolith-derived CaCO3 deposition and can serve as a proxy for regional primary productivity. Mechanistically, positive winter NAO phases intensify westerlies and Atlantic inflow, deepen the mixed layer, delay spring restratification, shorten the growing season, and reduce bloom magnitude, leading to decreased primary production—consistent with the observed anti-correlation between Ca/Fe (productivity) and winter NAO. Calibration against instrumental observations (CPR phytoplankton and OWSM dissolved O2) supports the proxy’s sensitivity to productivity changes on seasonal-to-decadal scales. Cross-correlation with the instrumental NAO confirms a robust winter anti-correlation during the high-quality period (1950–1992), while reduced correlations back to 1824 reflect cumulative age-model and data uncertainties and NAO non-stationarity. Not all Ca/Fe variance is attributable to the NAO; additional influences from modes such as the AMO and Subpolar Gyre, as well as sedimentological and chronological uncertainties, contribute. Despite these complexities, the detrended and bandpass-filtered NAO_Ca/Fe reproduces key features of existing NAO reconstructions over the last millennium and provides a continuous, open-ocean, sub-decadal perspective extending back 8000 years. The absence of a sustained positive NAO signal during the Medieval Warm Period aligns with some reconstructions, and amplitude modulation across the Holocene suggests evolving background states affecting NAO expression. The record’s broad oceanic ‘catchment’ may mitigate some biases associated with the NAO’s spatial non-stationarity, offering complementary insights to terrestrial or marginal-marine proxies.

This work establishes ITRAX-derived Ca/Fe from Norwegian Sea slope sediments as a robust proxy for coccolith-dominated primary productivity and, via an observed anti-correlation, for winter NAO variability. The study provides a sub-decadally resolved, open-ocean NAO reconstruction spanning the last 8000 years, calibrated against instrumental phytoplankton counts, dissolved oxygen observations, and the instrumental NAO, and cross-validated against multiple independent NAO reconstructions and a functional palaeoclimate network record. The NAO_Ca/Fe record captures high-frequency variability and reveals temporally varying amplitudes without a persistent positive NAO during the Medieval Warm Period. Future work should further interrogate Holocene NAO dynamics, their links to major climatic episodes (e.g., HTM, RWP, LIA) and human history, and disentangle the influences of other climate modes (e.g., AMO, SPG) and low-frequency oceanographic changes on primary productivity and the proxy signal.

• Chronological uncertainty increases back in time (to ~58–190 years for the oldest sections), impacting cross-period comparisons and weakening correlations with early instrumental NAO records (1824–1992).
• The XRF Ca/Fe metric is semi-quantitative and may be affected by residual grain-size, water-content, or core-surface effects, though tests suggest minimal bias; subtle current-speed variations cannot be entirely excluded.
• Calibration relies on a relatively short overlap with instrumental observations; the modern period may be influenced by anthropogenic changes in the marine environment.
• The NAO is non-stationary; shifting centers of action and evolving background states can modulate proxy–NAO relationships, and not all variance in Ca/Fe is driven by NAO. Other climate modes (AMO, SPG) and sedimentological processes likely contribute to variability.
• Dissolution appears negligible, but potential pre- or post-depositional effects cannot be completely ruled out despite SEM evidence and high sedimentation rates.