Gateway-driven weakening of ocean gyres leads to Southern Ocean cooling

Declining atmospheric CO2 concentrations are considered the primary driver for the Cenozoic Greenhouse-Icehouse transition, ~34 million years ago. A role for tectonically opening Southern Ocean gateways, initiating the onset of a thermally isolating Antarctic Circumpolar Current, has been disputed as ocean models have not reproduced expected heat transport to the Antarctic coast. Here we use high-resolution ocean simulations with detailed paleobathymetry to demonstrate that tectonics did play a fundamental role in reorganising Southern Ocean circulation patterns and heat transport, consistent with available proxy data. When at least one gateway (Tasmanian or Drake) is shallow (300 m), gyres transport warm waters towards Antarctica. When the second gateway subsides below 300 m, these gyres weaken and cause a dramatic cooling (average of 2–4 °C, up to 5 °C) of Antarctic surface waters whilst the ACC remains weak. Our results demonstrate that tectonic changes are crucial for Southern Ocean climate change and should be carefully considered in constraining long-term climate sensitivity to CO2.


Geological proxy data from drill sites
We compare the results of our simulations, of different depths of both gateways, to geological constraints in the Southern Ocean (Supplementary Figure 1) covering the geological time from Eocene to Early Oligocene (~41 to 30 Ma).
Geological sites which are located close to each other (50-300 km apart) and show similar microfossil species assemblages have been summarized into one data point in  Table 1) are derived using two paleothermometer methods, tetraether index of 86 carbon atoms TEX86 (1) and alkenone unsaturation index U k '37 (2,3) .
Previously, it has been suggested that biomarker-based TEX86 proxies for Eocene SSTs may be biased towards warm temperatures in both polar regions (4,5) , as seasonal biased export production of the biomarkers may have been dependent on the seasonality in grazing and fecal pellet formation (6). However, this mechanism for a warm bias is not well-supported by the available sediment trap studies. Although biomarker-based SST proxies tend to record warmer SSTs than other proxies, the SST trends per site are robust and are independent of the applied calibration (1) .
We calculate our modelled SST differences from our simulations' absolute SST outputs between two gateway configurations. We compare these to the proxy SST changes within the respective geological time periods. The time period 41 to 36 Ma contains the Middle Eocene Climatic Optimum (MECO, ~40 Ma), a short-lived global warming event with global temperatures increasing by ~5 °C 7) . For studies presenting data including this anomaly (8) , we removed this peak from the SST range presented in Supplementary Table 1 (noted with *). Best fit between model and proxy data ( Figure 3) is as followed: prior 41 Ma both gateways remain at ≤300 m; from 41 to 36 Ma one gateway deepens to >600 m depth; from 36 to 33.6 Ma, the second gateway transitions from 300 m to 600 m depth; and from 33.6 Ma onwards the second gateway deepens further (>600 m). Although, we do not compare modelled SST with proxy SST for the time period 49 to 41 Ma, we add the available proxy data for this particular time period to this overview for completeness.  (20) , nutrient levels (21) , salinity (22) , and upwelling, typically with presence of Southern Ocean sea ice (23,24) . The Paleogene

Supplementary
Southern Ocean harbours strictly endemic dinocysts from Antarctic coastal regions, as well as a high abundance of bipolar dinocyst species that occur in both north and south polar regions (25)(26)(27)(28) . Although the records of dinocyst assemblages show robust qualitative signals of surface currents connecting ocean basins, the water masses' vigour and vertical extension cannot be quantified from these microfossil assemblages.
In addition, dinocysts are transported spatially by ocean currents before settling down to the seafloor, and drill site. The travelled distances depend on several factors, including current speed, sinking speed of the particle and seafloor depth. Present-day modelling with particle tracing python module OceanParcels reveals that dinocysts (with an average sinking speed of 6 m/day) can be transported up to 1500 km in regions such as the strong Weddell Gyre (29) . No similar quantitative study exists for Eocene models. However, we follow Nooteboom et al.
(2019) (30) 's study assuming meridional dinocyst transport of several hundred kilometres in the subpolar gyres (<1500 km transport due to the weaker gyre strengths compared to presentday; about 60-65 Sv, Weddell Gyre (29) ; and 15-30 Sv, Ross Gyre (31,32) ) and several hundred kilometres of zonal transport in regions dominated by circumpolar flow pattern. Endemic species are found in drill sites located along the gyres' western boundary currents, likely transported from the Antarctic coast to the mid-latitudes ( Figure 3). Cosmopolitan species are likely transported zonally by a circum-polar flow; however, we assume less transport distance due to the weaker-than-today ACC transport (see main text, Figure 2). In order to quantify the exact transport distances, additional particle tracing experiments for the Eocene are required in the future.  (36,37) , as weathering of young volcanic West Antarctic rocks result in more radiogenic εNd values in the nearby Pacific Ocean seawater (-3 to -5 (36,38) ), whereas the Atlantic Ocean's seawater shows less radiogenic values (about -9 (36,39) ), due to weathering of old East Antarctic cratonic material. εNd measurements on fossil fish teeth or ferromanganese crusts in sediments are used to track changes in bottom water masses, and potential interoceanic connections through the gateways (36,40) .   (36,40) as the proxy for bottom current flow direction (in comparison to Pacific endmember (purple square): -3 to -5 (36,38) , Atlantic endmember (blue square): -9 (36,39) ). Details of the sites' paleolocations, recorded geological time periods, as well as all data used in this study are collated in Supplementary Table 3. We extended the northern boundary from 25°S to 0°S for the model with a shallow TG (300 m) and a deep DP (1000 m) in order to test the hypothesis of a "proto-ACC"-type current flowing north of Australia after the opening of a deep Drake Passage (45) . Such a flow has been detected in previous model simulations (45) . stream function and sea surface temperatures (legend as in Figure 2, main text).

Present-day bathymetry
This ocean simulation uses the present-day bathymetry ETOPO-1 (41) with the same paleo atmospheric forcing as the other simulations presented in this study (Supplementary Figure   8). A prominent circumpolar current dominates the Southern Ocean flowing through wide and deep Southern Ocean gateways. However, the net transport through the gateway reaches only about 22% of today's ACC transport and the ACC frontal systems are further south (modelled zonal wind stress has the maximum at 55°S) compared with today's ACC conditions. In addition, our model does not produce significantly strong gyres in the Weddell and Ross Sea, which reach about 60-65 Sv (Weddell Gyre (29) ) and 15-30 Sv (Ross Gyre (31,32) ) strength today. Furthermore, the SSTs offshore of Antarctica reach values of ~10 °C which is much warmer than today's conditions, but very close to the SST observation in the model run with both gateways below 600 m (Figure 2; Supplementary Figure 3).
These results lead us to suggest that it is the increasing equator-to-pole SST gradients that are responsible for the ACC to strengthen during post-EOT evolution until reaching its presentday strength of 170 Sv (42) .
bathymetry, b. stream function and sea surface temperatures (legend as in Figure 2, main text).

Closed gateway (at 0 m)
We simulate the oceanographic consequences with one gateway entirely closed (TG at 0 m depth) and a deep second gateway (DP at 1000 m; Supplementary Figure 9a). With the early opening (from 0 m to 300 m), the subpolar gyres remain relatively stable in size; however, the subpolar Pacific gyre decreases slightly in strength, from ~50 Sv to ~40 Sv. The observed SST changes are small, not exceeding 1 °C differences, during the initial, shallow opening of the second gateway (Supplementary Figure 9c). It requires a deepening to depths > 300 m to achieve significant surface water cooling along the Antarctic coast.  Figure 2), c. shows the resulting sea surface temperature change with TG deepening from 300 m to 1500 m.

Smooth bathymetry
The 0.25° model run using smooth bathymetry with a 300 m deep TG yields a subpolar Atlantic gyre similar in strength and size to that in the "rough" counterpart. In addition, a very strong gyre centre in the southeast Pacific is observed (reaching 90 Sv; Supplementary Figure 13a).
This occurs in a region of closed "f/H contours" and is likely eddy-driven (e.g. 46,47) . In the subpolar Atlantic the SSTs offshore Antarctica are about 13-14 °C, which is 3-4 °C colder than the SSTs in the "rough" simulation ( Figure 2a). Furthermore, the Antarctic coast along the Southwest Pacific reaches slightly colder temperatures with the smooth bathymetry.
However, no significant differences in the Australian-Antarctic SST distributions can be observed between the "rough" and "smooth" run. The "smooth" net transport through the 300 m TG and 1000 m DP are 15.6 Sv and 5.1 Sv, respectively, which is stronger in eastward direction compared to the "rough" equivalent (~3.2 Sv and -6.2 Sv; Figure 2a).
When TG deepens to 1500 m, the gyres weaken (although, the subpolar Pacific one remains >60 Sv). The "smooth" gyres are about twice as strong as their "rough" counterpart.
The "smooth" SSTs offshore Antarctica show a cooling trend, however, not as much as in the 1500 m (DP at 1000 m; legend as in Figure 2), c. shows the resulting sea surface temperature change with TG deepening from 300 m to 1500 m.
It can be concluded that both improved ocean model resolution and the presence of detailed seafloor roughness are crucial for the strong SST cooling and changes in the ocean circulation dynamics to reconstruct in the Southern Ocean.