In-phase millennial-scale glacier changes in the tropics and North Atlantic regions during the Holocene

The study investigates how past climate variability influenced centennial- to millennial-scale fluctuations of small mountain glaciers during the Holocene, to clarify the roles of natural versus anthropogenic forcings in current and future glacier retreat. While orbital forcing led to decreasing summer insolation in the Northern Hemisphere and increasing in the Southern Hemisphere, observed Holocene glacier histories are not uniformly consistent with this trend. In the North Atlantic sector (Greenland, Scandinavia, Iceland, European Alps), some glaciers achieved larger extents in the early Holocene than during the Little Ice Age, suggesting additional mechanisms beyond insolation. In the tropical Andes, glacier behavior depends on both temperature and precipitation, and records show irregular retreat despite increasing austral summer insolation, with a notable lack of mid-Holocene moraines. The Atlantic Meridional Overturning Circulation (AMOC) exerts a strong impact on climate and has shown substantial Holocene variability, potentially affecting temperatures and precipitation patterns across the North Atlantic and tropics. The research question is whether millennial-scale AMOC changes can explain in-phase Holocene glacier variations in the tropical Andes and North Atlantic regions (TANAR), superimposed on external forcings (orbital, greenhouse gases, ice-sheet extent, volcanic activity).

Prior work indicates hemispheric contrasts in Holocene glacier evolution driven by insolation, with Northern Hemisphere mid-latitude glaciers generally advancing toward a Little Ice Age maximum and many Southern Hemisphere glaciers retreating. However, North Atlantic sector glacier chronologies indicate early Holocene maxima exceeding LIA extents, inconsistent with monotonic insolation forcing. In the tropical Andes, combined sensitivity to temperature and precipitation, and scarcity/obliteration of mid-Holocene moraines, point to additional drivers. AMOC variability during the Holocene, documented from multiple proxy reconstructions, is implicated in regional surface temperature and precipitation changes around the North Atlantic and in the tropics. ENSO variability, a major source of tropical climate variability, has shown Holocene changes in variance/frequency, but reconstructed ENSO states do not align with the timing of major glacier advances and retreats in the Andes, suggesting ENSO alone cannot explain the long-term glacier changes. Transient Holocene simulations (TraCE, LOVECLIM) including external forcings often yield monotonic or inconsistent regional responses, under-representing AMOC’s role.

Glacier chronology development: The authors produced new in situ cosmogenic nuclide exposure ages (10Be and 14C) from moraines and deglaciated bedrock at five small mountain glacier sites: Charquini North and South (Cordillera Real, Bolivia, 16°30′S, 68°10′W), Saint-Sorlin Glacier (French Alps, 45°N, 6°17′E), a glacier adjacent to Clavering Ice Cap (NE Greenland, 74°37′N, 21°33′W), and composite moraines in the Enchantment Lakes region (Cascade Range, Washington State, USA). These new data were combined with published Holocene moraine chronologies from small glaciers worldwide (Ice-D Alpine database), emphasizing robust, well-dated sites (n=66) and excluding ambiguous or poorly constrained records. Sampling and laboratory analyses: 10Be extraction for most moraine samples occurred at CALM (Meudon) and LN2C (CEREGE) with AMS measurements at ASTER (CEREGE), normalized to 07KNSTD and STD11 standards. Charquini South moraine samples were processed and measured at ETH Zürich (TANDY AMS), normalized to S2010N calibrated against ICN 01-5-15. Enchantment Lakes 10Be samples were prepared at Lamont-Doherty and measured at LLNL CAMS. In situ 14C and 10Be from bedrock at Charquini North were extracted at Tulane University, with 14C measured at WHOI NOSAMS and stable isotope ratios at UC-Davis. Exposure ages were calculated with CREP using LSD scaling, assuming no denudation, and region-appropriate production rates (e.g., tropical Andes local 10Be rate; Arctic-NE mean 4.05 ± 0.23 at/g/yr for Greenland and Alps). Moraine ages are weighted means after chi-square outlier rejection. Exposure-burial modeling: A numerical model simulated coupled 10Be–14C production, decay, and erosion in bedrock columns to test 100,000 randomly generated Holocene exposure/burial histories (100-year time steps; persistence parameter P from 0.6 to 0.99). Scenarios were constrained to include burial prior to 10.10 ± 0.26 ka and around 1.20 ± 0.12 ka (from moraine ages). Erosion rates from 0–50 mm/ka were tested. Scenarios matching measured surface 14C and 10Be concentrations within 3σ (including measurement and production-rate uncertainties) for two bedrock samples (P12, P13) were retained as plausible histories. Climate simulations and forcing analysis: Two transient Holocene simulations were analyzed: TraCE (CCSM3) and LOVECLIM, including external forcings (insolation, GHGs, ice sheets; LOVECLIM also includes volcanic forcing; TraCE includes a crude freshwater flux estimate). Model outputs were compared against glacier chronologies to evaluate whether external forcings alone reproduce observed glacier patterns. Freshwater hosing experiments: Five AOGCMs (HadCM3, IPSL-CM5A-LR, MPI-ESM, EC-Earth, BCM2) were used to perform Greenland coastal freshwater hosing (0.1 Sv for 40 years) relative to historical control simulations, yielding an average AMOC weakening at 26°N of 2.6 ± 1.7 Sv. Temperature and precipitation fingerprints associated with AMOC weakening were compiled, including impacts over TANAR and ITCZ shifts. AMOC reconstructions and fingerprints: Independent Holocene AMOC proxy reconstructions (e.g., Ayache 2018; Thornalley 2013; Caesar et al. 2018, 2021; and a recent Labrador Sea convection record) were compiled. For the instrumental era, North Atlantic subpolar gyre SST indices were regressed against detrended global SSTs (HadISST) to derive an observational AMOC fingerprint (interhemispheric Atlantic SST seesaw). Observed decadal mass balance of Zongo Glacier (Bolivia) was correlated with global SSTs to establish teleconnections with the North Atlantic. Semi-empirical AMOC-correction model: A linear model was developed to superimpose reconstructed AMOC impacts onto transient simulation fields: Y = X_mod + β X_AMOC, where X_mod is the simulated temperature or precipitation, and X_AMOC is the AMOC variance scaled by model-derived fingerprints from hosing experiments. The amplitude of Holocene AMOC variations was calibrated (standard deviation ≈ 1.2 Sv) to best fit the 30–90°N Holocene temperature reconstruction (Kaufman et al. 2020) by minimizing RMSE and maximizing correlation, using the ensemble mean fingerprint from the five hosing models (AMOC decrease ≈ 2.6 Sv). The calibrated AMOC history was used to generate AMOC-adjusted temperature and precipitation time series in TANAR, which were then compared with glacier chronologies and hydroclimate proxies. Site-specific chrono highlights: Charquini South: LIA moraine mean 10Be age 420 ± 15 a (n=7); nearest downslope early Holocene moraine 10.0 ± 0.21 ka (n=4). Charquini North: moraines at 1.20 ± 0.12 ka (n=3) and 10.10 ± 0.26 ka (n=4); bedrock 10Be–14C modeling indicates mid-Holocene ice-free conditions (~9.5–5 ka) and glacier cover from ~4 ka onward. Saint-Sorlin (Alps): early Holocene moraines at 9.79 ± 0.49 ka and 11.40 ± 0.44 ka (plus 11.20 ± 0.58 ka lateral). Clavering, Greenland: moraines at 10.15 ± 0.37 ka (grouped), 1.17 ± 0.09 ka, and a near-terminus ~0.28 ± 0.25 ka; Enchantment Lakes (Cascades): late Holocene advances during Neoglacial (4–2 ka) and LIA. Comparative glacier database: A curated compilation (n=66) of small glacier Holocene chronologies with strict inclusion criteria (clear separation from Younger Dryas, dated moraines between LIA and late glacial, multiple CRE ages or integrated lake records, exclusion of highly dispersed datasets) was used to define TANAR versus other regional patterns based on normalized glacier length indices for early Holocene (11.6–9 ka), mid-Holocene (8–4 ka), and late Holocene (3 ka–LIA).

- In-phase TANAR glacier pattern: The tropical Andes and North Atlantic regions share a coherent millennial-scale Holocene glacier evolution distinct from many other regions: (1) early Holocene maximum extents (~11.6–9 ka), (2) pronounced mid-Holocene retreat/minimal extents (~8–4 ka), (3) late Holocene re-advance culminating during or prior to the LIA. - New Andean chronology: At Charquini (Bolivia), early Holocene moraines at ~10.0–10.1 ka and late Holocene advances at ~1.2 ka and LIA (~420 ± 15 years) bracket a mid-Holocene interval with glaciers smaller than late Holocene extents, corroborated by combined in situ 10Be–14C bedrock modeling (ice-free ~9.5–5 ka; mostly ice-covered from ~4 ka) and independent lake sediment and plant remains evidence for reduced ice between ~7 and 5.2 ka. - North Atlantic sector agreement: Early Holocene maxima and mid-Holocene minima occur in Alps (e.g., Saint-Sorlin ~11.4–9.8 ka), Greenland (Clavering ~10.15 ka), Scandinavia, and Iceland, with late Holocene re-advances, aligning with the TANAR pattern. - Regional contrasts: Outside TANAR, glacier histories are less uniform: Cascades broadly similar; Alaska shows late Holocene maximum; Himalayas exhibit complex responses linked to westerlies/monsoon balance; Southern Hemisphere mid-latitudes generally show progressive Holocene retreat after early Holocene maxima. - AMOC as driver: Independent AMOC reconstructions reveal early and late Holocene minima with a mid-Holocene maximum (~7 ka). This timing is consistent with TANAR glacier fluctuations: strong AMOC → warmer North Atlantic and drier tropical Andes → glacier retreat (mid-Holocene); weak AMOC → cooler North Atlantic and southward ITCZ shift intensifying the South American summer monsoon → increased tropical Andes precipitation and cooler conditions contributing to glacier advances (early and late Holocene). - Quantified impacts: Semi-empirical AMOC-corrected simulations indicate that between mid-Holocene (6–5 ka) and the last millennium (1–0 ka), a weaker AMOC could cause roughly 0.5°C cooling north of 30°N and ~1 mm/day (~40%) precipitation increase in the tropical Andes. In boreal summer, AMOC accounts for up to one-third of late Holocene Northern Hemisphere cooling, with the remainder primarily due to insolation. - Calibration and magnitude: Calibrated Holocene AMOC variability is ~4–5 Sv, with a 3–4 Sv difference between 6 ka and preindustrial, consistent with PMIP3 models. Freshwater hosing experiments robustly reproduce AMOC fingerprints (AMOC reduction at 26°N of 2.6 ± 1.7 Sv), yielding TANAR-consistent cooling and ITCZ shifts. - ENSO mismatch: Reconstructed early Holocene and LIA El Niño-like conditions and mid-Holocene La Niña-like conditions do not match the observed glacier advances/retreats, suggesting ENSO is not the principal millennial-scale driver of Holocene Andean glacier change. - Model-data inconsistency: Transient simulations (TraCE, LOVECLIM) underrepresent reconstructed AMOC evolution (simulate post-8 ka strengthening without mid-Holocene maximum), underscoring missing processes or forcings in models.

The observed in-phase Holocene glacier fluctuations across the tropical Andes and North Atlantic regions address the central hypothesis that a common driver, beyond external radiative forcings, modulated millennial-scale glacier behavior. AMOC variations provide a physically consistent mechanism linking cooler North Atlantic conditions and enhanced tropical Andes precipitation under weak AMOC states, and warmer/drier conditions under strong AMOC, producing the TANAR signature of early/late Holocene glacier advances and mid-Holocene retreat. Teleconnections via interhemispheric Atlantic SST gradients and ITCZ migration explain coherent precipitation and temperature responses affecting glacier mass balance. Semi-empirical AMOC-corrected simulations reconcile glacier chronologies with hydroclimate reconstructions better than full-forcing transient runs alone, implying that AMOC changes likely exerted a primary control in the tropical Andes and a secondary but significant influence in mid- to high-latitude North Atlantic regions, superimposed on insolation-driven trends. The discrepancy between proxy-based AMOC reconstructions and transient model AMOC highlights gaps in current modeling of Holocene ocean circulation and its climatic impacts. These findings elevate the importance of constraining past AMOC dynamics to improve attribution of Holocene glacier changes and to reduce uncertainties in future projections where AMOC weakening may modulate, but not offset, greenhouse gas–driven warming and hydrological changes.

This study compiles new and published cosmogenic nuclide chronologies to show that tropical Andes and North Atlantic region glaciers evolved in-phase during the Holocene, with early Holocene maxima, mid-Holocene minima, and late Holocene re-advances. By integrating proxy-based AMOC reconstructions, freshwater hosing fingerprints, and a semi-empirical AMOC-correction applied to transient simulations, the authors demonstrate that millennial-scale AMOC variability is consistent with the timing and structure of TANAR glacier fluctuations, via temperature changes in the North Atlantic and ITCZ-driven precipitation shifts over the tropical Andes. The work underscores AMOC’s role as an important driver superimposed on orbital forcing, and the need to better represent AMOC dynamics in climate models. Future research should: (1) expand well-dated small-glacier chronologies in under-sampled regions; (2) develop higher-resolution transient simulations including realistic meltwater fluxes and improved ocean dynamics; (3) refine AMOC reconstructions and calibrations across the Holocene; (4) improve representation of regional precipitation responses over complex terrain; and (5) investigate additional processes (e.g., Mediterranean outflow changes, SH westerly shifts) that may modulate AMOC. Such advances are crucial for constraining future climate and cryosphere responses under projected AMOC weakening and ongoing greenhouse gas forcing.

- Modeling gap: Transient Holocene simulations (TraCE, LOVECLIM) do not reproduce reconstructed AMOC evolution (lack mid-Holocene maximum), limiting direct model-data consistency. - Semi-empirical approach: The linear AMOC-correction model is simplified and does not capture full coupled dynamics or nonlinearities, especially for precipitation over complex orography (e.g., Alps). - Proxy calibration: AMOC reconstructions are standardized and require calibration (assumed SD ≈ 1.2 Sv) against temperature reconstructions, adding uncertainty to AMOC amplitude estimates. - Regional data coverage: Glacier chronologies remain sparse or ambiguous in some regions (e.g., East Africa, parts of Asia), and some mid-Holocene moraines may have been obliterated by later advances, potentially biasing interpretations. - Cosmogenic assumptions: Exposure-age calculations assume no denudation for moraines and negligible subglacial production; exposure-burial models assume specific erosion regimes and shared histories within sampling areas. - ENSO and other forcings: The study focuses on AMOC; other internal modes or regional forcings (e.g., PDO, land-surface feedbacks) may contribute but are not fully quantified. - Precipitation signal resolution: Hosing-derived precipitation fingerprints may not resolve mesoscale dynamics and orographic effects critical for glacier mass balance.