The potential of Hudson Valley glacial floods to drive abrupt climate change

During the last deglaciation, periodic meltwater inputs from the retreating Laurentide Ice Sheet were hypothesized to hinder North Atlantic Deep Water (NADW) formation, weaken the AMOC, and trigger cold episodes such as the Younger Dryas and the 8.2 ka event. Although geomorphic evidence documents numerous meltwater discharges broadly coincident with centennial-to-millennial cool periods, uncertainties in flood volumes and durations hinder definitively linking specific floods to specific cold events or even eliciting a clear AMOC response. Recent modeling indicates that the geographic location and duration of meltwater input strongly influence AMOC sensitivity. Motivated by this, the study tests the hypothesis that meltwater floods routed down the Hudson Valley—specifically the drainage(s) of Glacial Lake Iroquois and successors around ~13.3 ka—triggered the Intra-Allerød Cold Period (IACP). The purpose is to assess whether realistic combinations of flood volume, duration, frequency, and background runoff from the Hudson Valley could substantially weaken AMOC under Allerød-like atmospheric conditions and sea levels 75 m lower than present.

Prior work proposed meltwater routing from the Laurentide Ice Sheet as a driver of abrupt North Atlantic cooling via AMOC weakening (e.g., Broecker 1989). Geomorphic and chronological evidence indicates two large late-glacial floods via the Hudson Valley: an initial ~700 km3 outburst from Lakes Iroquois and Coveville, followed by a larger ~2500 km3 flood from Lakes Frontenac and Fort Ann, both occurring between ~13,050–13,310 cal BP, close to the IACP. Proxy records suggest suppressed AMOC during the IACP from intermediate-water temperature and radiocarbon ventilation changes in the North Atlantic and Nordic Seas. Modeling studies show AMOC response depends strongly on the location and duration of freshwater forcing; thresholds near a sustained 0.1 Sv are often needed to weaken and maintain a reduced AMOC, and freshwater inputs closer to deep convection regions (e.g., Hudson Bay, Greenland margins) produce a stronger response than inputs farther south. Some eddy-resolving models show differing persistence of AMOC reductions depending on forcing location and model resolution. The large Agassiz drainage (~9500 km3) is often invoked for the Younger Dryas, and Hudson Bay forcings are implicated for the 8.2 ka event, but sustaining multi-century AMOC reductions with short-lived outbursts alone remains challenging in models.

A suite of numerical experiments was conducted with the MITgcm in a global, eddy-permitting 1/6° (~18 km) configuration with 50 vertical levels, coupled to a dynamic–thermodynamic sea-ice model (viscous-plastic rheology). Sea level was set 75 m lower than present (closing the Bering Strait) to approximate ~13.3 ka bathymetry. Atmospheric boundary conditions used ERA-40 monthly means (1979–2002) to emulate relatively warm Bølling–Allerød conditions. Control simulations with and without a continuous background Hudson meltwater flux of 0.05 Sv were spun up for 20 years. Glacial meltwater properties were set to 0 °C and 0 PSU. Floodwaters were introduced into four ocean grid cells near the intersection of the 75 m bathymetric contour and the modern Hudson Canyon, consistent with geomorphic evidence of southward routing across the shelf. Experimental design explored: (1) a “Realistic” two-flood scenario of 1-month durations with fluxes 0.31 Sv (~700 km3) and 1.09 Sv (~2500 km3) separated by 10 years (total ~3700 km3); (2) single-event scenarios releasing the combined volume (~3700 km3, ±15% allowed in design) over 1 month (1.40 Sv) or over 1 year (0.12 Sv); (3) frequency tests releasing the same volume three times at 2-year intervals with either 1-month (1.40 Sv per event; total 11,100 km3) or 1-year (0.12 Sv per event; total 11,100 km3) durations. Each experiment was run both with and without the continuous 0.05 Sv background flux, yielding ten total flood experiments. Diagnostics included sea surface salinity (SSS) anomalies vs. control, mixed layer depth (MLD) in Labrador and Greenland Seas (MLD defined where density exceeds surface by 0.125 kg m−3 in waters deeper than 2000 m), and AMOC strength at 25°N (mean and variability over 20-year simulations). Geographic boxes were defined for Labrador Sea (52–65°N, 70°W–20°W), Greenland Sea (70–80°N, 20°W–20°E), and the outflow region (37–43°N, 74–70°W).

- In all experiments, meltwater released from the Hudson Valley produced initial near-field surface freshening (~1.5 PSU) near the paleo-Hudson mouth, but freshwater was rapidly entrained by the Gulf Stream and mixed into the interior ocean, becoming undetectable by the time it reached subpolar deep water formation regions. - No appreciable reduction in surface salinity occurred in the Labrador or Greenland Seas relative to control conditions. - Mixed layer depth in the Labrador and Greenland Seas differed by less than 4% from the Control across all scenarios. - AMOC strength at 25°N deviated by less than 1% from the Control in all experiments, with variability comparable to modern observations. With a 0.05 Sv background flux, mean AMOC (Sv; mean with standard deviation) and percent differences were: Control 29.95 (3.16); Realistic 29.90 (3.07), −0.18%; 1 yr 29.86 (3.47), −0.32%; 1 mo 29.98 (3.57), +0.11%; 3x_1yr 29.68 (3.11), −0.90%; 3x_1mo 29.82 (3.42), −0.43%. - Even the most extreme scenario (three successive 1-month floods totaling 11,100 km3) failed to produce a persistent salinity anomaly in deep convection regions or a significant AMOC response. - Experiments without the 0.05 Sv background flux yielded consistent results (no significant AMOC weakening). With the background flux, AMOC was only ~1–2% weaker than without it, and no prolonged weakening occurred when outburst floods were superimposed. - The results indicate that Hudson Valley floods around ~13.3 ka—despite timing close to the IACP—were unlikely to have been the sole cause via AMOC weakening. Flood location, long transport distance, and turbulent mixing in the Gulf Stream limit freshwater delivery to NADW formation regions.

The findings directly challenge the hypothesis that Hudson Valley outburst floods alone triggered the IACP by weakening AMOC. The experiments show that input location and downstream pathways critically determine freshwater fate: meltwater injected far south of deep convection zones is rapidly mixed and diluted by the energetic Gulf Stream and eddies during its long transit, preventing capping of the subpolar gyre and suppression of NADW formation. High-resolution, eddy-permitting dynamics likely enhance realistic mixing compared to coarse models that may overestimate persistence of freshwater anomalies. Varying flood duration (1 month vs 1 year) and frequency (single vs three events) did not alter this outcome; even cumulative volumes comparable to Lake Agassiz discharges failed to weaken AMOC from the Hudson input location. Although a continuous 0.1 Sv freshwater forcing is often required in models to weaken and sustain a reduced AMOC, glacial outburst floods are short-lived and intermittent, making sustained AMOC suppression unlikely from Hudson-sourced floods alone. The small additional weakening (~1–2%) under a constant 0.05 Sv background flux suggests that broader-scale background freshwater from other LIS outlets (e.g., Gulf of St. Lawrence, Hudson Bay, Mackenzie River) and/or atmospheric changes associated with the LIS might have preconditioned the system. The results imply that multiple forcings—combined freshwater sources, sea-ice export, sea-level rise, and atmospheric circulation changes—may be necessary to produce prolonged IACP-like cooling, rather than Hudson Valley floods in isolation.

Using an eddy-permitting global ocean–sea-ice model, the study finds that plausible Hudson Valley glacial flood scenarios near ~13.3 ka do not significantly weaken AMOC. Rapid mixing and long transport pathways prevent sufficient freshwater from reaching deep convection regions, regardless of flood volume, duration, or frequency, and even with a continuous 0.05 Sv background flux. Thus, the Hudson Valley floods were unlikely to be the sole driver of the IACP via AMOC weakening. The work underscores the dominant role of input geography and ocean dynamics in mediating freshwater–AMOC interactions. Future research should explore: (1) combined and temporally overlapping freshwater sources from multiple outlets; (2) longer-duration or sustained fluxes from ice-sheet collapse and ice discharge; (3) coupled atmosphere–ocean–sea-ice processes (including sea-ice export, atmospheric circulation, and sea-level rise effects); and (4) the contributions of other mechanisms such as volcanism and sea-ice feedbacks to abrupt climate shifts.

- The atmosphere was prescribed (uncoupled) using modern ERA-40 climatology, omitting potential feedbacks and preconditioning by contemporaneous atmospheric circulation changes due to the Laurentide Ice Sheet. - Background runoff was limited to 0.05 Sv from the Hudson Valley based on geomorphic constraints; potential concurrent background discharge from other outlets (e.g., Gulf of St. Lawrence, Hudson Bay, Mackenzie) was not included and could alter preconditioning. - Freshwater forcing focused on Hudson Valley input location; alternative routing pathways (closer to deep convection regions) were not tested in this study. - While eddy-permitting resolution improves mixing realism, model-dependent mixing and advection characteristics could influence freshwater dispersion compared with other eddy-resolving setups. - Flood volumes and durations are constrained by available geomorphic and chronologic evidence and include uncertainty (±15% considered); true event characteristics may differ.