Tracking westerly wind directions over Europe since the middle Holocene

The Mediterranean Basin, home to about 400 million people and a recognized climate hot spot, exhibits large hydroclimate variability and has experienced exceptionally low rainfall over recent decades. Drought has been linked to enhanced sea-level pressure (SLP) contrast between the Azores High and the Icelandic Low, i.e., the positive NAO phase, which migrates the westerlies northward and diverts moisture from the Mediterranean. The spatial position of this SLP dipole is not stable; its displacement determines westerly track angle and the phase of the East Atlantic (EA) pattern, leading to non-stationary NAO-correlated rainfall regimes across Europe on decadal to multidecadal timescales. To better understand westerly variability on multiple timescales, reliable proxies are needed. Prior work from Iberian terrestrial and marine sediments and Moroccan tree rings with Scottish speleothems indicates prolonged positive NAO during the Medieval Climate Anomaly, but broader Northern Hemisphere proxy assemblages and models have highlighted contention in paleo-NAO reconstructions due to complex Atlantic SLP interactions and scarcity of direct SLP-sensitive proxies. Because NAO-correlated rainfall regimes reflect the westerlies’ position, comparing a network of westerly-sensitive records can constrain past westerly positions. This study presents a multi-annually resolved precipitation record from northern Italy (Bàsura cave) and integrates it with regional archives to reconstruct large-scale hydroclimate changes and variability in the SLP dipole and westerlies over the last 6,500 years.

Previous studies have linked Mediterranean hydroclimate extremes to NAO phases and SLP dipole variability, with evidence suggesting prolonged positive NAO during the Medieval Climate Anomaly from Iberian marine/lake records and combined Moroccan tree-ring and Scottish speleothem records. However, model-tested reconstructions have shown that paleo-NAO indices can be contentious due to complex Atlantic atmospheric dynamics. Only a limited number of proxies directly reflect Atlantic SLP patterns, complicating efforts to track the westerlies’ position over mainland Europe. The EA pattern, representing pressure anomalies over the eastern North Atlantic, modulates the location of the SLP dipole and influences NAO-correlated precipitation regions. Studies have documented non-stationary relationships between NAO and European precipitation, with contributions from EA and other modes, and highlighted decadal to multidecadal variability in the NAO’s spatial footprint during the instrumental era. This work builds on stalagmite, lacustrine, marine sediment, and pollen records from across Europe and the North Atlantic to resolve the spatial-temporal behavior of westerlies and the NAO/EA interaction over the Holocene.

Study site and sampling: Bàsura cave (44°08′N, 8°12′E; Toirano, Liguria, northern Italy) lies in a Mediterranean climate where >70% of ~1276±310 mm annual precipitation falls from September to February. Two stalagmites (BA14-1: 190 mm, aragonite; BA18-4: 90 mm, calcite) were collected 500–800 m from the cave entrance (RH 97–100%). Cave air T ~15.6 °C. U-Th chronology: 96 subsamples (78 for BA14-1; 18 for BA18-4; 10–100 mg) were dated by MC-ICP-MS (Thermo-Finnigan Neptune) at NTU HISPEC using a 229Th–233,236U triple spike and isotope dilution. Ages are relative to 1950 CE with 2σ uncertainties; initial 230Th corrections were 0–2 years. A Monte Carlo StalAge model produced age-depth models, yielding coverage from 6437±12 to 5±18 yr BP. Stable isotope analyses: 997 powdered subsamples (732 for BA14-1; 265 for BA18-4; 10–50 μg) were micromilled at 0.05–0.10 mm resolution along growth axes. δ18O was measured using Thermo Finnigan MAT 253 instruments (Nanjing Normal Univ., Fujian Normal Univ., Hokkaido Univ.) and Micromass IsoPrime for Hendy tests; reproducibility ±0.08–0.12‰ (1σ), referenced to VPDB via NBS-19. Hendy tests on 13 layers, plus coeval Sr/Ca data, indicate near-equilibrium calcite precipitation. Composite Δ18O construction: BA14-1 (aragonite) δ18O values were 1.2‰ higher than BA18-4 during overlap (702–752 yr BP); BA14-1 data were offset by −1.2‰ and merged with BA18-4. The overlap was averaged and resampled at 4-year intervals to match resolution. An additional ±0.2‰ combination error, combined with instrumental error, gave ±0.23‰ maximum uncertainty. The composite was normalized to a mean of −5.8‰, producing the Δ18O record used as a westerly-sensitive proxy (negative/positive Δ18O indicates strong/weak westerlies and high/low rainfall over northern Italy). Trace element analyses: Sr/Ca was measured (10–50 μg subsamples; 0.1–0.5 mm spacing) on a Finnigan Element II ICP-SF-MS with matrix-matched standards every 4–5 samples (2σ reproducibility ±0.5%). Different mineralogies (calcite vs aragonite) produce offsets; each series was z-standardized and merged using the overlapping interval (728–875 yr BP) assuming coherent hydroclimate response. Instrumental and reanalysis data: Relationships between Genoa precipitation (DJFM), 850-mb zonal winds around Gibraltar/Iceland, and SLP were assessed using NCAR/NCEP Reanalysis v3. EOF analyses of European-Atlantic precipitation (60°W–40°E, 20–80°N) identified EA and NAO patterns and their correlation with Genoa precipitation. Modeling: A transient simulation (TraCE-21ka, CCSM3) provided 700-mb height pressure variability over the eastern Atlantic for centennial–millennial comparison with Δ18O. CESM2 preindustrial control (approx. 2°×2°, 300-year window with 5-year running mean) was used to correlate SST fields with 700-mb height pressure variations at 52°30′N, 27°30′W to assess coherence between East Atlantic SLP anomalies and North Atlantic Current SSTs. Correlation and uncertainty methods: For series without age uncertainty, probable error (P.E.) bounds were used: P.E.=0.6745×(1−r²)/√N. For series with age uncertainties, a Monte Carlo approach (Fohlmeister et al., 2012) generated artificial series preserving variance, autocorrelation, resolution, and age uncertainties to obtain distributions of best correlations and significance. Multi-proxy comparisons were performed at centennial to multicentennial scales by resampling proxies into 15-year windows, averaging into consecutive 150-year bins, and correlating with the transformed Bàsura record.

• The Bàsura cave composite Δ18O provides a multi-annual to decadal-resolution record of westerly strength and precipitation over northern Italy spanning the last ~6,500 years.
• Δ18O variations align with regional hydroclimate reconstructions: significant correlations with northern Italy lacustrine flood records (r = −0.42, n = 31, p < 0.05) and central Italy speleothem wet/dry indices (r = 0.67, n = 31, p < 0.05). Coherent regional dry (5–4 kyr BP) and wet (4–3 kyr BP) phases match circum-Mediterranean records and coincide with known societal impacts (e.g., 5.2 kyr BP, 4.8–4.1 kyr BP events).
• Modern Genoa precipitation correlates with 850-mb zonal winds and is anti-correlated with SLP over NW Europe, consistent with EA modulation; SLP in NW Europe is anti-correlated with 850-mb zonal wind near Toirano (r = −0.6 ± 0.04, n = 97, p < 0.01, DJF 1920–2016), emphasizing westerly control on local hydroclimate.
• On multi-decadal to multi-centennial scales over the past millennium, Bàsura-inferred precipitation correlates positively with reconstructed NAO (r = 0.69, n = 292, 99% significance). Comparisons with western Greenland lacustrine temperature show positive correlations at 2.2–1.2 kyr BP (r = 0.37, n = 64, 95%) and 0.8–0.3 kyr BP (r = 0.69, n = 41, 95%), but negative correlation at 4.2–2.2 kyr BP (r = −0.48, n = 84, 99%), evidencing shifts in NAO-correlated regions since the middle Holocene.
• Instrumental-era running correlations reveal non-stationary NAO behavior, with stronger NAO–precipitation links in 1870–1900 and 1950–1980 CE than 1905–1935 CE, consistent with EA-phase-driven migrations of the SLP dipole and westerly tracks.
• Centennial–millennial similarity between Δ18O and simulated 700-mb height pressure over the eastern Atlantic (TraCE-21ka) and coherence with reconstructed EA indices support EA modulation of NAO dipole location and European precipitation domains.
• Spatial syntheses show contrasting hydroclimate under positive NAO phases: at 5.4–3.5 kyr BP, cool/dry conditions below ~45°N and wet/warm in northern Europe; at 2.2–1.2 kyr BP, multi-centennial wet/warm conditions in southern, central, and northern Europe with ambiguous signals in the eastern Mediterranean and northern Africa. Storminess in northern Europe coincides with southern European droughts at 5.8–5.5, 4.5–3.95, 3.3–2.4 kyr BP, but with southern wetness at 1.9–1.05 and 0.6–0.25 kyr BP.
• Running correlations with Norway, Germany, and Austria proxies indicate domain shifts around ~3 kyr BP; Austria remains aligned with Bàsura throughout, while Norway and Germany shift, consistent with a southward migration of the SLP dipole after ~2.2 kyr BP.
• The Δ18O similarity with Kaite cave (Spain) δ18O implies co-occurring westward Icelandic Low and eastward Azores High shifts, i.e., counterclockwise SLP dipole rotation affecting NAO-correlated regions and westerly orientation over Europe.
• North Atlantic ocean conditions likely modulate SLP dipole migrations: over the past 2.2 kyr, increases in high-latitude SSTs (N Norway, E Greenland) and SSS around the Subpolar Gyre, along with increased Atlantic water inflow to the Nordic Seas, suggest a reduced meridional SST gradient and a southward-shifted SLP dipole (positive EA). CESM2 preindustrial simulations show strong coherence between East Atlantic SLP anomalies and SST changes along the North Atlantic Current.

The study demonstrates that European hydroclimate and the position/orientation of the westerlies have varied on decadal to millennial timescales due to migrations of the North Atlantic SLP dipole. The Bàsura Δ18O record, tightly linked to regional precipitation and westerly dynamics, reveals that NAO-correlated regions over Europe have not been stationary since the middle Holocene. Periods like 4.2–2.2 kyr BP show an inversion in the relationship with NAO compared to later times, indicating geographic shifts of NAO-positive and NAO-negative precipitation domains. The EA pattern emerges as a key modulator that, together with NAO, controls the centennial–millennial positioning and rotation of the SLP dipole and hence the westerly tracks (e.g., SW–NE tilt prior to 2.2 kyr BP vs a more zonal tilt afterward). Spatial syntheses across >20 westerly-sensitive archives corroborate that under differing combinations of NAO and EA phases, the westerly-affected regions and storm tracks migrate, altering hydroclimate across Europe. Comparisons with model simulations (TraCE-21ka, CESM2) and instrumental reanalysis strengthen the inference that ocean–atmosphere coupling, particularly Atlantic SST/SSS gradients and AMV-like variability, influences EA phase and SLP dipole position. Thus, the findings address the central question of how and why westerly wind directions and their European hydroclimate impacts have evolved since the middle Holocene.

This work provides a multi-annually resolved 6.5-kyr record from Bàsura cave that tracks European westerly variability and reveals non-stationary NAO-correlated regions driven by migrations and rotations of the North Atlantic SLP dipole. The synthesis of >20 westerly-sensitive proxies across Europe, together with reanalysis and climate model outputs, shows that combined NAO and EA modes govern the spatial domain of westerly influence and storminess on multi-decadal to millennial scales. Evidence suggests a southward migration of the SLP dipole after ~2.2 kyr BP, likely linked to changes in Atlantic meridional SST gradients and increased advection of warm, saline Atlantic waters to higher latitudes. These insights underscore a strong oceanic control on atmospheric circulation pathways over Europe. Future research should refine chronological alignment across proxies, expand westerly-sensitive archives geographically and temporally, improve isotope-enabled modeling of NAO–EA interactions, and assess how ongoing changes in AMOC strength, greenhouse gas forcing, and aerosols may further modulate the SLP dipole and European westerlies.

• Reliability of early reanalysis: Correlation maps before ~1920 CE are less reliable due to sparse observations in NCAR/NCEP Reanalysis v3.
• Age and proxy uncertainties: Although U-Th dating is high precision with small initial 230Th corrections, age uncertainties and composite construction (including mineralogical offsets and δ18O combination error ±0.23‰) introduce uncertainty. Monte Carlo approaches were needed to assess correlation significance.
• Spatial and temporal heterogeneity: Shifts in SLP dipole and westerly tracks are likely time-transgressive, making exact transition boundaries (e.g., around ~2.2–3 kyr BP) difficult to define.
• Proxy interpretational complexity: Paleo-NAO reconstructions can be contentious due to complex SLP interactions; some regions (eastern Mediterranean, northern Africa) show ambiguous hydroclimate signals during certain intervals.
• Coverage limitations: Despite >20 records, proxy network density varies across regions, potentially biasing spatial interpretations.