In this study, we use the Australian Community Climate and Earth System Simulator (ACCESS) Earth System Model (ESM) version 1.5, which participated in the Coupled Model Intercomparison Project Phase 6 (CMIP6) and includes an interactive carbon cycle (Ziehn et al., 2020). The resolution of the atmospheric model component (UK Met Office Unified Model, UM) is 1.875° longitude × 1.25° latitude × 38 levels (Martin et al., 2010; The HadGEM2 Development Team, 2011). It is coupled to the land surface model, which is the Community Atmosphere Biosphere Land Exchange (CABLE) model (Kowalczyk et al., 2013). The ocean component uses the Geophysical Fluid Dynamics Laboratory Modular Ocean Model 5.1 (MOM5.1; Griffies, 2012) and sea ice component uses the Los Alamos Community Ice Code model CICE4.1 (Hunke and Lipscomb, 2010). The ocean and sea-ice components use a common horizontal grid of 360 × 300 cells at a nominal 1° resolution, with meridional refinement of grid spacing down to 0.33° at the equator and down to ∼ 0.4° in the Southern Ocean and a tripolar grid with land-based poles at 65° N over Eurasia and North America (Ziehn et al, 2020). More details of the model description and performance are provided in Ziehn et al. (2020). Previous studies have performed freshwater hosing experiments to alter the AMOC strength in pre-industrial, interglacial boundary conditions, historical and future simulations using this model (Du et al., 2025; Pontes and Menviel, 2024; Saini et al., 2025a), with acceptable AMOC sensitivity to freshwater perturbations within the range of CMIP6 models (Weijer et al., 2020).

The model was first integrated under 49 ka boundary conditions starting from the PI simulation (Table 1; Saini et al., 2025b). The model is forced by 49 ka orbital parameters (Berger, 1978), greenhouse gas concentrations (Köhler et al., 2017); and ice-sheet extent and topography corresponding to 52.5 ka (Gowan et al., 2021), with closest match to ∼ 48–52 ka sea level estimates (Shakun et al., 2015). Vegetation was modified to reflect the presence of continental ice sheets, including the conversion of forest to bare soil between the Cordilleran and Laurentide ice sheets. Additionally, C₃ (photosynthetic growth form) crops in the PI simulation were replaced by surrounding vegetation types. The land-sea mask and river runoff are also adjusted to reflect the modified topography (see details in Saini et al., 2025b).

The 49 ka simulation was run with step-wise changes in boundary conditions for a total of 1555 model years, with the last 760 years implemented with full boundary conditions (49ka-full in Saini et al., 2025b). The last 100 years in 49ka-full show relatively stable surface air temperature and sea-surface temperatures (Saini et al., 2025b, Fig. A1). We therefore consider this as a quasi-equilibrium state for the 49 ka climate, and define this period as the 49ka_control experiment in this study. This 49ka_control displays a stronger AMOC strength than PI (31 Sv compared to 21 Sv for PI), which is considered to represent interstadial conditions at 49 ka. The stronger AMOC at 49 ka is due to changes in the topography and expansion of the ice sheets in the model, which enhances the North Atlantic gyres and cyclonic circulation in the Labrador Sea (Saini et al., 2025b). Simulated globally averaged surface air temperatures are 2.7 °C lower than PI, and global mean SST is 1.2 °C lower. Drier conditions are simulated over the NH high latitudes and most land areas of the SH relative to PI (Saini et al., 2025b, Fig. 2c).

For this study, North Atlantic meltwater hosing experiments (Table 2) are performed from the quasi-equilibrium state conditions (49ka_control). Freshwater fluxes are added into the North Atlantic (50–70° N, 70–0° W) to alter the AMOC strength in the model, similar to previous studies (Du et al., 2025; Pontes and Menviel, 2024; Saini et al., 2025a).

Previous modelling studies suggest that D–O stadials are associated with a ∼ 50 % AMOC slowdown, while HS are more likely to be close to a complete shutdown of the AMOC (e.g. Malmierca-Vallet et al., 2023; Menviel et al., 2020), even though the exact changes in the AMOC transport during D–O and HS are still poorly constrained from proxy records (Henry et al., 2016; Lynch-Stieglitz, 2017).

To simulate D–O stadials, we add 0.2 and 0.3 Sv of freshwater in the northern North Atlantic under constant 49ka boundary conditions (49ka_0.2Sv and 49ka_0.3Sv experiments). This leads to ∼ 32 % and ∼ 50 % AMOC reductions relative to the 49ka_control, respectively (see detailed changes in Table 2). We also perform an AMOC shutdown experiment with a constant 0.4 Sv freshwater forcing to simulate HS (49ka_shutdown; Table 2). See Fig. S1 in the Supplement for changes in the mixed layer depth in each simulation.

The AMOC evolution for each simulation is shown in Fig. 1. In 49ka_shutdown, the AMOC takes ∼ 100 model years to shut down (below 5 Sv is considered shutdown; Saini et al., 2025a). The simulation is then continued for another 400 model years for the model to fully respond to the shutdown, especially at southern high latitudes. The last 50 years of the simulation is used for analysis (dark blue in Fig. 1).

To explore pseudo D–O stadial conditions, we consider both the 49ka_0.2Sv, and 49ka_0.3Sv experiments. For 49ka_0.2Sv, we analyse years 150–200 (dark red in Fig. 1). To further weaken the AMOC, 49ka_0.2Sv is continued first forced with 0.25 Sv meltwater input, then with 0.3 Sv (see Fig. 1). The analysis period for 49ka_0.3Sv is the 50-year interval from year 450 to 500 (dark brown in Fig. 1), which corresponds to the same integration time as the 49ka_shutdown experiment to allow for better comparisons. The 49ka_0.2Sv and 49ka_0.3Sv experiments are hereafter referred to as D–O stadial or slowdown experiment, and the 49ka_shutdown experiments are also referred to as HS or shutdown experiment. AMOC weakening refers to all three experiments.

The NH and SH HC strength is calculated as the maximum and minimum value of the atmospheric mass streamfunction at 500 hPa in DJF and JJA, respectively. The width of the HC is calculated from the latitudes of zero mass streamfunction at mid-tropospheric level (500 hPa) as edges of the ascending and descending branches (Frierson et al., 2007; Seo et al., 2014). The mass streamfunction is interpolated to 0.5° latitude resolution for the calculations.

The ITCZ position in this study is defined as the latitude of zero zonal-mean meridional mass streamfunction at 850 hPa, corresponding to the cross-equatorial ascending branch of the northern (DJF)/southern (JJA) Hadley Cells. This definition provides a more direct dynamical linkage to HC variability than the conventional precipitation-based metrics (e.g. Braconnot et al., 2007), while yielding consistent ITCZ positions across both approaches (Bian and Räisänen, 2024; Bischoff and Schneider, 2014).

SH westerlies strength is calculated as the 850 hPa zonal mean maximum wind speed between 40 to 60° S, whilst the position of the SH westerly winds refers to the latitude of the maximum wind speed interpolated to 0.5° latitude resolution, consistent with previous studies (e.g. Du et al., 2024; Saini et al., 2025b). Changes in the SH westerlies are calculated across different ocean basins, the longitude range for the Atlantic Ocean covers between 70 and 0° W, the Pacific Ocean 130° E–70° W, and the Indian Ocean from 0 to 130° E.

The STR position and intensity over Australia is calculated as the maximum mean sea level pressure (MSLP) value zonally averaged over 140–150° E between 10 and 45° S. The zonal mean MSLP values are then interpolated to 0.5° latitude resolution, followed by a fitted cubic spline to detect the latitude of maximum pressure. The latitude is the STR location, and the pressure value is the STR intensity (Grose et al., 2015).

Finally, monsoon domains are defined as regions in which the wet monsoon season precipitation rate exceeds the dry season rate by at least 2.5 mm d⁻¹, and is responsible for at least 55 % of the annual precipitation (Wang et al., 2011; Yeung et al., 2021).

First, we provide an overview of annual temperature and precipitation changes in our D–O and HS simulations relative to the 49ka_control climate and compare the linearity between the climatic responses to each hosing experiment by normalising the temperature and precipitation response to a “per Sv” AMOC decrease. Then, we evaluate the changes in seasonal large-scale atmospheric circulations by investigating the changes in Hadley Cell, MSLP, and SH westerly winds to AMOC weakening. Finally, we investigate the influence of changes in these climate features on Australasian climates.

The meridional oceanic heat transport to the North Atlantic is reduced by ∼ 77 % (∼ 1 PW at 30° N) in the HS experiment (see Fig. S2). This change is larger than previous model simulations of AMOC shutdowns under LGM conditions that found an average ∼ 40 % reduction in Atlantic meridional heat transport (∼ 0.8 PW at 30° N) (Menviel et al., 2008, 2020; Stouffer et al., 2006). The larger reduction in meridional oceanic heat transport can be explained by the large AMOC reduction simulated here (29 Sv reduction compared to an averaged of 15 Sv in previous LGM experiments; Kageyama et al., 2013). In the D–O stadial simulations with ∼ 50 % and 32 % AMOC reductions (Fig. 2e, f), the simulated reductions in ocean heat transport at 30° N are limited to 25 % and 16 %, respectively (Fig. S2), which exert weak constraints on the atmospheric response.

In the HS simulation, significantly reduced heat transport leads to a widespread significant NH mean cooling of 3.8 °C, with maximum annual mean surface air temperature (SAT) decrease of up to 26.3 °C in the Labrador Sea (Fig. 2a) due to the large decrease in the AMOC strength, while the SH displays a significant warming (mean 0.43 °C) from 0 to 55° S. Simulated SSTs show up to 15.4 °C cooling in the North Atlantic, and up to 6.5 °C warming in the South Atlantic (Fig. S3a). Over Antarctica, significantly colder conditions are simulated in the Ross and Weddell Seas (Fig. 2a), which could be related to changes in sea ice extent or negative low pressure anomalies over the region (Fig. 6b). The SAT patterns in the D–O stadial simulations show generally similar responses to the interhemispheric response in HS, with NH cooling and some SH warming anomalies relative to 49ka_control (Fig. 2b, c). The small reductions in the northward oceanic heat transport in D–O stadial simulations are only sufficient to induce some significant warming beyond the Atlantic Ocean, reaching into the Indian and eastern Pacific Oceans (Fig. 2b, c).

To compensate the significantly reduced ocean heat transport in the HS experiment, the ITCZ is pushed southwards towards the warmer hemisphere. The precipitation is significantly reduced by a mean of 2.3 mm d⁻¹ to the north of the equator in the Atlantic Ocean and greatly enhanced to the south in the HS experiment (Fig. 2d). The pattern extends to every ocean basin, indicating a global-wise southward shift of the annual mean ITCZ position in HS. This further leads to enhanced precipitation intensity in the South Pacific Convergence Zone (SPCZ) region (10–30° S, 155° E–140° W) (Fig. 2d). In the D–O stadial simulations, the southward displacement pattern of the ITCZ is mostly evident over the Atlantic Ocean, extending into the eastern equatorial Pacific. The global precipitation patterns show similar responses between the slowdown simulations (Fig. 2e, f); however, greatly different from the HS simulation, particularly in the tropical and subtropical regions. The SPCZ precipitation is slightly reduced in the D–O stadial experiments (Fig. 2e, f), which is the opposite of the shutdown experiment. Widespread drying is simulated over much of the NH associated with the cooling in all experiments, which is consistent with previous modelling results with a weaker AMOC (e.g. Jackson et al., 2015).

To further quantify the linearity of changes, we normalise the temperature and precipitation response to a “per Sv” AMOC decrease relative to 49ka_control (Fig. 3). The D–O stadial simulations with 10 and 14 Sv decrease in AMOC strength show the same degree of annual mean temperature and precipitation changes per Sv change (details in Table S1), which indicate linear relationships. This can be seen from the regions of significant changes in the 49ka_0.2Sv experiment (Fig. 2c, f) being enhanced in magnitude and expanded in areal extent in the 49ka_0.3Sv simulation (Fig. 2b, e), and similar normalised spatial patterns between the two experiments (Fig. 3b, c, e, f). Nevertheless, crossing the tipping point of a full shutdown of the AMOC under glacial conditions leads to around 1.3 times the global annual average temperature cooling per Sv AMOC decrease compared to D–O stadial experiments (see Table S1). More specifically, NH temperature displays up to 1.6 times more cooling in 49ka_shutdown, while the SH (from 0 to 55° S) experiences up to 6.8 times warmer temperatures per Sv than in the slowdown experiments (see Fig. 3a–c for normalised spatial patterns). This large difference in SH temperature changes between the D–O and HS simulations are attributed to a simulated SH mean cooling in D–O stadial experiments while a SH warming is simulated in the HS (Table S1; Fig. 3a–c). Globally averaged precipitation changes under AMOC shutdown are also ∼ 1.3 times larger per Sv of AMOC reduction than in the AMOC slowdown simulations, with the NH becoming ∼ 2 times drier and the SH ∼ 3 times wetter per Sv compared to the D–O slowdown simulations (Table S1; Fig. 3d–f).

In this section, we investigate the drivers of the diverse hydroclimate response in the simulations by examining the changes in the HC, and the associated changes in the MSLP, 850 hPa wind circulations, including the SH mid-latitude westerly winds, and the STR using the metrics described in Sect. 2.3.

In this section, we investigate the climatic changes over Australia and the surrounding region in response to AMOC slowdown and shutdown at 49 ka and propagated by large-scale changes in atmospheric circulation discussed above.

The northern half of the Australian continent, as well as parts of New Guinea, are simulated to be colder while the southern part of Australia is warmer in the shutdown simulation in both DJF and JJA seasons (Fig. 8a, d). This is consistent with previous AMOC weakening experiments performed with the same model under interglacial background climates (Du et al., 2025; Saini et al., 2025a). However, the amplitude of this temperature gradient is smaller in the slowdown experiments (Fig. 8b–c, e–f). The 49ka_0.2Sv experiment displays a similar pattern as 49ka_shutdown in DJF, even though the changes are insignificant (Fig. 8c). In JJA, an overall cooling pattern is simulated across Australia in the slowdown experiments (Fig. 8e, f), while significant warming anomalies are simulated over southern Australia due to AMOC shutdown (Fig. 8d).

Over New Zealand, significant warming is triggered by AMOC shutdown (Fig. 8a, d), while small temperature changes are suggested in the slowdown experiments, with significant cooling to the south and warming to the east of New Zealand (Fig. 8b–c, e–f).

The current hydroclimate of northern Australia and nearby Maritime Continent and New Guinea are influenced by the Indo-Australian summer monsoon in the austral summer (DJF), while southern Australian and New Zealand hydroclimate is more affected by a northward shift of the SH westerly winds in austral winter (JJA) season. Moreover, the STR, as the surface descending branch of the Hadley Cell, undergoes meridional displacement with the seasons, greatly influencing the timing of moisture delivery over Australia (Grose et al., 2015; Pepler et al., 2018).

In the shutdown experiment, significantly increased precipitation in DJF is simulated across the Australian continent south of 10 ° S, with an up to 6 mm d⁻¹ increase in northern Australia due to AMOC shutdown (Fig. 9a). The increase is likely driven by stronger north-westerly winds associated with an enhanced and expanded Indo-Australian summer monsoon (see Fig. 9a, where blue contours represent 49ka_control monsoon domain, and red contours represent the monsoon domains in the shutdown and slowdown experiments). The STR shifts southwards by 1° to around 41° S, allowing more tropical moisture to reach central Australia while drier conditions prevail around the high-pressure regions (Fig. 9a). The STR shift is associated with significantly lower summer (DJF) temperatures of more than 2 °C over northern Australia and New Guinea due to increased precipitation and cloud, whereas most of the SH mid-latitudes are warmer (Fig. 8a). In the slowdown simulations, parts of northern Australia and New Guinea receive a small increase in monsoon precipitation (Fig. 9b, c), but little change in the spatial extent of the monsoon is simulated relative to 49ka_control. Reduced DJF SPCZ precipitation is simulated, corresponding to the warming anomalies over the area (Fig. 8b, c).

During the austral winter (JJA), significant drying is simulated along the east coast of Australia, extending to central Australia in all simulations (Fig. 9d–f). This drying trend is due to slightly increased pressure across Australia and New Zealand (Fig. 6b, d, f). In HS simulation, the JJA STR is shifted southward by 1° to ∼ 30° S, associated with a displaced descending branch of the southern HC. The southward shift of the STR suppresses the mid-latitude frontal systems associated with low pressure that brings the cold south-westerly winds and precipitation into southern Australia in winter, as occurs under modern climate, which then leads to drying conditions in southeastern Australia in HS (Fig. 9d). The drying pattern over southern Australia in JJA is consistent across all experiments, which may be due to the southward shift of the SH westerlies in all simulations (Fig. 7a). The STR position and intensity remains unchanged in the slowdown simulations in both seasons.

In this section, we present a diagnostic of the role of energetic constraints in shaping the atmospheric response to AMOC weakening through simulations of HS at 49 ka, as well as how that leads to the SH mid-latitude changes and hydroclimate response around Australasia through changes in the cross-equatorial Hadley Cells.

The northward oceanic heat transport is reduced by ∼ 77 % at 30° N in the HS simulation (Fig. S2), which generates a significant interhemispheric energy imbalance that is primarily compensated by the atmospheric response, particularly in DJF (Fig. S4) – through a southward shift of the annual ITCZ, and a reorganization of the HC strength and width. The winter (DJF) northern HC strengthens and increases in width, with a southward shift in its ascending branch (ITCZ position), while the southern wintertime (JJA) HC weakens but increases in width due to a small northward shift in the JJA ITCZ position at 850 hPa. The northward shift in the JJA ITCZ reduces the cross-equatorial heat transport, which explains the smaller contributions to atmospheric compensation in JJA. Note that the simulated change in the southern JJA width is different from responses under the Last Interglacial and PI boundary conditions using the same model (Saini et al., 2025a). This difference may be due to the different background states between warm interglacial and cold glacial climates, with more investigation needed in the future. Nevertheless, our findings of an enhanced northern HC and weaker southern HC in response to AMOC weakening are consistent with previous modelling studies (Chiang et al., 2014; Chiang and Friedman, 2012; Lee et al., 2011; Saini et al., 2025a).

Changes in the subtropical high surface pressure (STR) are associated with variations in the descending branch of the HC. An increase in DJF MSLP is simulated in the NH subtropics due to enhanced descending branch of the northern HC, while the opposite weakening of the STR is seen in the SH (Fig. 10). A strong decrease in surface pressure accompanied by anomalous cyclonic circulation is simulated over North Pacific in JJA, associated with significantly increased precipitation over the region (Fig. S6f). The upper-level SH subtropical jet weakens in JJA as a result of weakened southern winter HC. Global mean SH mid-latitude westerly winds at 850 hPa strengthen and migrate southwards in both JJA and DJF, which further leads to strengthened upwelling of the cold deep waters across the Antarctic circumpolar latitudes and drives surface Southern Ocean cooling (Anderson et al., 2009). The strengthening of the SH westerly winds may also explain the increase in atmospheric CO₂ concentration during Heinrich 5 inferred from proxy records via the Southern Ocean (Menviel et al., 2014a; Wendt et al., 2024).

The STR over the Australian region is weaker and moves southwards in both seasons. The austral summer (DJF) STR sits south of the Australian continent during the HS (Fig. 10a), which allows tropical moisture from the expanded Indo-Australian monsoon domain to reach southern Australia, leading to significantly enhanced precipitation across the Australian continent south of ∼ 10° S. A southward shift in the winter (JJA) STR reduces the fronts that pass over southern Australia, leading to drier conditions over southern Australia along with the enhanced and southwards displacement of the SH westerlies (Fig. 10b).

The simulated atmospheric circulation response to AMOC shutdown agrees with previous modelling studies, suggesting weaker South Pacific jet, stronger SH westerlies (Bard and Rickaby, 2009; Chiang et al., 2014; Lee et al., 2011), and increased SH monsoon precipitation (Cheng et al., 2012; Zhang et al., 2023) due to AMOC weakening. The results of this study can provide an example of the climate response to a strong AMOC reduction that can be compared with available proxy records to better understand the processes at play.

In this study, we use normalised temperature and precipitation response to the AMOC decrease to quantify the linearity of climatic changes. The global area-averaged precipitation and temperature anomalies show linear changes per Sv of AMOC decrease in AMOC slowdown simulations at 49 ka, with roughly -0.003 mm d⁻¹ and -0.05 °C per Sv change for both the 0.2 and 0.3 Sv experiments, respectively (see more details in Table S1). However, crossing the threshold of AMOC shutdown between 0.3 and 0.4 Sv triggers a different response in precipitation and temperature patterns which do not continue this linear response either globally or regionally (Fig. 3). The Australian temperature field between the slowdown and shutdown experiments shows relatively linear changes (Fig. 8), but this is not the case for hydroclimates, particularly during the DJF monsoon season (Fig. 9). The magnitude of global annual changes in precipitation and temperature in the shutdown experiment are 1.3 times as large per Sv relative to the slowdown simulations (Fig. 3 for spatial patterns), with larger changes in the SH (Table S1). This appears to be due to the non-linear responses in the large-scale atmospheric circulations such as the Hadley Cell, leading to different responses in the climatic processes (e.g. pressure system, westerly winds, etc.) between AMOC slowdown and AMOC shutdown.

We attribute this non-linearity between D–O and HS climate responses primarily to a threshold in the ocean heat transport response between the modest reductions associated with AMOC slowdown and the much larger reductions due to AMOC shutdown. The northward ocean heat transport in D–O stadial simulations is reduced by less than 25 % at 30° N (Fig. S2), which imposes a weak energetic constraint on the atmosphere. This results in minimal compensation by the atmosphere and consequently insignificant anomalies in the Hadley circulation and regional atmospheric responses. This study provides a possible location of the threshold in hosing strength between linear weakening and nonlinear shutdown, but it is important to note that the location is likely to differ between models and climate states both in the real world and in simulations. Comparison of simulations from different climate models would also provide evidence of the robustness of model responses to AMOC weakening under glacial conditions. Future studies performing longer simulations would be useful to explore slower responses, such as changes in sea ice extent and Southern Ocean temperatures.

Our experimental setup was designed to assess the multi-centennial-scale impact of an AMOC shutdown and an AMOC slowdown on the climate system. The experiments were designed to obtain the AMOC slowdown and shutdown states with the smallest freshwater input. Although the input fluxes are idealized and significantly higher than current estimates (Zhou and McManus, 2024), the experimental design is consistent with the MIS3 protocol (Malmierca-Vallet et al., 2023), and similar to the recent North Atlantic Hosing Model Intercomparison Project (NaHOsMIP), in which 0.3 Sv is added in the North Atlantic for more than 100 model years to simulate the climate response to an AMOC shutdown in CMIP models (Ben-Yami et al., 2024; Diamond et al., 2025; Jackson et al., 2023). The difference in the input fluxes arises from the different sensitivity of AMOC to freshwater perturbations across models. In ACCESS-ESM1.5, the AMOC response to North Atlantic freshwater fluxes depends on both the magnitude and duration of the meltwater input (Du et al., 2025). In our experimental design, a pulse of 0.3 Sv for a few hundred years does not lead to an AMOC shutdown under 49 ka boundary conditions (Fig. 1). This highlights the issue of this model's sensitivity to freshwater forcing.

In this study, the 49ka_0.3Sv experiment is obtained from a two-step increase in freshwater input of 0.25 Sv for 100 model years followed by 0.3 Sv of 200 years after the 49ka_0.2Sv experiment. Nevertheless, this setup should not significantly affect the large-scale climate response given the length of the simulations performed here (500 years), nor should the exact location of the meltwater input in the subpolar North Atlantic, based on previous experiments with different North Atlantic meltwater input location, duration and magnitudes (Pontes and Menviel, 2024; Du et al., 2025; Saini et al., 2025a). Future studies should perform a thorough examination of the impact of meltwater input location on the climate response.

Lastly, this study uses the last 50 years of each hosing experiment to assess the climate responses. When using the 30-year running-mean values for the last 150 years in each experiment, the results are consistent with our 50-year average (Fig. S7). We note that only one model was used, and a future study could examine the linearity in multiple models.

We now compare our model simulations with available proxy records for D–O and HS during MIS3 (Menviel et al., 2020), focusing on temperature and precipitation anomalies in 49ka_shutdown versus the 49ka_0.3Sv experiment (as representative of HS and D–O stadial respectively in Fig. 11). Our model simulation is generally consistent with these proxy records, as well as with the proxy benchmarks summarized by Izumi et al. (2023) in their Tables 1 and 3, with colder and drier conditions over the NH, a weakened NH summer monsoon, and stronger South American and South African monsoons. However, the monsoon responses from our simulations in the South American and South African domains are more complex (Fig. S8). Moreover, our simulated monsoon changes for HS5 (Fig. S8a) are consistent with proxy records indicating weaker Indian (Deplazes et a. 2013) and East Asian summer monsoons (Zhang et al., 2024), as well as southward displacement and strengthening of the IASM (Carolin et al., 2013; Ayliffe et al., 2013; Scroxton et al., 2022). Both D–O and HS simulations appear to be consistent with high-resolution proxy records of surface moisture from Australia in HS1, suggesting wetter conditions during North Atlantic cooling events (Denniston et al., 2013, 2017; Treble et al., 2017). In future work, we plan to compare our simulation of Heinrich and D–O stadials with HS5 records in more detail.