Atlantic Circulation and Ice Sheet Influences on Upper South Atlantic Temperatures During the Last Deglaciation

Abstract

Atlantic Meridional Overturning Circulation (AMOC) disruption during the last deglaciation is hypothesized to have caused large subsurface ocean temperature anomalies, but records from key regions are not available to test this hypothesis, and other possible drivers of warming have not been fully considered. Here, we present the first reliable evidence for subsurface warming in the South Atlantic during Heinrich Stadial 1, confirming the link between large-scale heat redistribution and AMOC. Warming extends across the Bølling-Allerød despite predicted cooling at this time, thus spanning intervals of both weak and strong AMOC indicating another forcing mechanism that may have been previously overlooked. Transient model simulations and quasi-conservative water mass tracers suggest that reduced northward upper ocean heat transport was responsible for the early deglacial (Heinrich Stadial 1) accumulation of heat at our shallower (~1,100 m) site. In contrast, the results suggest that warming at our deeper site (~1,900 m) site was dominated by southward advection of North Atlantic middepth heat anomalies. During the Bølling-Allerød, the demise of ice sheets resulted in oceanographic changes in the North Atlantic that reduced convective heat loss to the atmosphere, causing subsurface warming that overwhelmed the cooling expected from an AMOC reinvigoration. The data and simulations suggest that rising atmospheric CO 2 did not contribute significantly to deglacial subsurface warming at our sites.

Less Remineralized Carbon in the Intermediate‐Depth South Atlantic During Heinrich Stadial 1

“…Liu et al () show that simulated weakening of the AMOC from an LGM state reduces convective heat exchange in the North Atlantic as well as northward heat transport, warming subsurface waters in the South Atlantic down to ~2,000‐m water depth. At intermediate depths, simulated temperatures increased by 1.5 to 2 °C (Liu e al., 2009), somewhat smaller than the reconstructed HS1 signal for 90GGC based on the Mg/Li ratio of benthic foraminifera (2–3 °C; Umling et al, ). These results suggest that 75% to 100% of the δ 18 O signal at the core site can be attributed to subsurface warming.…”

“…From 18 to 16 kyr BP, [CO 3 2− ] abruptly increased by ~22 μmol/kg, achieving a maximum value of 130 ± 2 μmol/kg during HS1. Over this same interval, benthic δ 13 C increased by 0.4‰ ± 0.03‰, while benthic δ 18 O decreased by 0.30‰ ± 0.05‰ (Curry & Oppo, ; Umling et al, ). The HS1 [CO 3 2− ] maximum was followed by a gradual ~10‐μmol/kg decline into the late Bølling‐Allerød (~13 kyr BP).…”

“…(f) Cibicidoides wuellerstorfi B/Ca time series (right y ‐axis) with converted [CO 3 2− ] values (left y ‐axis) from Lacerra et al (). Stable isotope data for 90GGC are from Umling et al (), while radiocarbon data are from Lund et al (). Radiocarbon age control points for 90GGC and 78GGC are shown as black triangles in (c) and (f), respectively.…”

“…The results of Umling et al () suggest that air‐sea gas exchange signature of water masses bathing the site may account for the HS1 δ 13 C signal. In their study, Cd/Ca (a proxy for ambient phosphate concentrations) in core 90GGC remains relatively constant throughout HS1 instead of decreasing as expected due to less remineralization (Schmittner & Lund, ) or deepening of the thermocline (Hain et al, ; Umling et al, ). Two or more mechanisms working in concert may have produced a net increase in δ 13 C and no net change in Cd/Ca.…”

“…In order to reconstruct ΣCO 2 variability during the deglaciation, we analyzed benthic foraminifera from an intermediate‐depth Brazil Margin core (KNR159‐5‐90GGC; 1,105‐m water depth). The core site is located within the core of AAIW today (Figure ) and is complemented by detailed time series of benthic δ 18 O and δ 13 C (Curry & Oppo, ; Umling et al, ). We compare our results to the output of a coupled ocean circulation‐biogeochemistry model capable of resolving tracer field (δ 13 C, [CO 3 2− ], and ALK) responses to variable AMOC (Schmittner et al, ).…”