Synoptic Moisture Intrusion Provided Heavy Isotope Precipitations in Inland Antarctica During the Last Glacial Maximum

Stable water isotopes in inland Antarctic ice cores are powerful paleoclimate proxies; however, their relationship with dynamical atmospheric circulations remains controversial. Using a water isotope climate model (MIROC5‐iso), we assessed the influence of the Last Glacial Maximum (LGM; ∼21,000 years ago) sea surface temperatures (SST) and sea ice (SIC) on Antarctic precipitation isotopes (δ18Op) through atmospheric circulation. The results revealed that the synoptic circulation mostly maintained southward moisture transport, reaching inland Antarctica. The steepened meridional SST gradient in the mid‐latitudes increased δ18Op in inland Antarctica with the enhanced baroclinic instability and synoptic moisture transport. In contrast, expanded SIC distribution decreased δ18Op over Antarctica by enhanced preferential removal of heavy isotopes during vapor transport due to the increased transport distance and enhanced surface cooling. These findings propose to use Antarctic ice cores to describe the southern hemisphere atmospheric circulation, represented by the westerly jets, during the LGM and other past climates.


Introduction
Ratios of stable isotopologues of water, H 2 16 O, H 2 18 O, and HD 16 O, expressed hereafter in the usual δ notation (i.e., δ 18 O, with respect to V-SMOW scale; Dansgaard, 1964), from Antarctic ice cores have been widely used to study Earth's climate variations for the past several hundred thousand years.For example, the obtained δ 18 O values made it possible to estimate temperature changes between the Last Glacial Maximum (LGM) and today (Augustin et al., 2004;Kawamura et al., 2017).
LGM is well known as an extremely cold climate period characterized by a low atmospheric CO 2 level (approximately 180 ppm) and extensive ice sheets in the northern hemisphere (NH).To simulate the climate conditions during this period allows to evaluate the climate models used for future climate projections (Kageyama et al., 2018).Estimating the temperature changes between LGM and today has been a mainstay in paleoclimatology for decades due to its connection with the Earth's climate sensitivity (ECS) (Kageyama et al., 2018).Changes in Antarctic climatological mean surface air temperature (T a ) (ΔT a ; Δ denotes changes between climates), derived from the δ 18 O values, have significantly contributed to constraining the ECS (Masson-Delmotte et al., 2006a, 2006b).
Although reconstructing the past T a from the ice core δ 18 O values is critical for the climate sciences, the relationship between T a and δ 18 O involves considerable uncertainties.To reconstruct the past Antarctic T a from the δ 18 O values, the isotopic thermometer assumption, which regards an observed present-day spatial T a /δ 18 O slope as a surrogate for the temporal slope at a given site, has been traditionally used (Dahe et al., 1994;Dansgaard, 1964;Lorius & Merlivat, 1977;Lorius et al., 1979;Satow et al., 1999).Process-based reconstruction of the past Antarctic T a was also carried out using simple one-dimensional Lagrangian isotope models or water isotope-enabled general circulation models (iso-GCMs).Such process-based studies suggested that various climatic factors possibly affect the Antarctic T a /δ 18 O slope; the moisture origin (Uemura et al., 2018), sea ice extension (Lee et al., 2008), ice sheet topography (Werner et al., 2018), surface inversion layer depth (Buizert et al., 2021), and precipitation seasonality (Erb et al., 2018).
Ice cores are expected to reflect precipitation-weighted δ 18 O p (Krinner & Werner, 2003;Sime & Wolff, 2011;Werner et al., 2018).If the influence of EPEs on Antarctic δ 18 O p is significant, T a reconstructed from the δ 18 O in Antarctic ice cores (δ 18 O ice ) should be weighted toward EPEs and deviate from the climatological mean T a .However, the roles of SH mid-latitude atmospheric circulations or EPEs in determining δ 18 O or T a /δ 18 O slope have not yet been fully explored for past climates, including LGM.In this study, we conducted multiple LGM experiments by an isotope-enabled atmospheric GCM (iso-AGCM), namely MIROC5-iso (Okazaki & Yoshimura, 2019a), with different sea surface boundary conditions.These experiments enabled us to test two different states of SH mid-latitude atmospheric circulations that are linked with sea surface temperature (SST; e.g., Nakamura et al., 2008) and sea ice concentration (SIC; e.g., Sime et al., 2013) during the LGM.We applied two recent sea surface reconstructions (Paul et al., 2021;Sherriff-Tadano et al., 2023) as boundary conditions for MIROC5-iso.
The remainder of this paper is organized as follows.In Section 2, the model, experimental settings, observational data set, and analysis method are discussed.In Section 3, we evaluated the simulated LGM climates in Antarctica and analyzed the processes ruling the δ 18 O p in Antarctica by investigating the differences between the simulated LGM experiments.In Section 4, we summarized our main findings concerning the combined effects of SST and SIC on δ 18 O p and discussed issues that need to be resolved.

Isotope-Enabled Atmospheric General Circulation Model
The atmospheric component of the fifth version of the Model for Interdisciplinary Research on Climate (MIROC; Watanabe et al., 2010) is based on a three-dimensional primitive equation in the hybrid σ-p coordinate, with spectral truncation adopted for horizontal discretization.Version MIROC5-iso, where water isotopes in the atmosphere and land surface parts were implemented by Okazaki andYoshimura (2017, 2019a) was used in this study.The resolution of the MIROC5-iso was set to a horizontal spectral truncation of T42 (approximately 280 km) and 40 vertical layers.Okazaki andYoshimura (2017, 2019a) and Kino et al. (2021) discuss the detailed parameterizations of the models and their applications for global and Antarctic present-day conditions.

Experimental Design
Four experiments were performed using MIROC5-iso (Table S1 in Supporting Information S1).A Pre-Industrial (PI) simulation was set up following the "piControl" experimental design in the Coupled Model Intercomparison Project-Phase 6 (CMIP6; Eyring et al., 2016).The mean SST and SIC fields (monthly averaged from 1870 to 1899) were obtained from the Atmospheric Modeling Intercomparison Project-Phase 2 (AMIP2; Taylor et al., 2000).Three LGM experiments were designed using the PMIP4 protocol (Kageyama et al., 2017).For the elevation and distribution of the ice sheet, the GLAC-1D reconstruction at year 21 ka (Briggs et al., 2014;Ivanovic et al., 2016;Tarasov et al., 2012) was used.The land-sea mask was extended according to the ice sheet.

Geophysical Research Letters
10.1029/2024GL108191 The boundary conditions of the land surface were the same as those in the PI simulation but masked by the LGM ice sheet.The δ 18 O of seawater was set to a globally uniform value (+1‰), according to Werner et al. (2018).
The LGM simulations differ in the prescribed SST and SIC to force MIROC5-iso.

Proxy Data for Model Evaluation
Ten Antarctic ice core records were used for evaluation.Each site location is shown in Figure 1a.For EDML, Dome B, Vostok, Dome C, Taylor Dome, Talos, WDC, and Byrd, Δδ 18 O ice and Δδ 18 O ice -inferred ΔT a for LGM minus PI compiled by Werner et al. (2018) were employed.For Δδ 18 O ice of the South Pole, we used the results of Steig et al. (2021).
Δδ 18 O data from speleothems (Comas-Bru et al., 2019, 2020b) and ice cores (Kawamura et al., 2007;Landais et al., 2013;Uemura et al., 2018) were used to evaluate the simulated LGM climates at the global scale.For speleothems, Δδ 18 O in the calcite was obtained from the Speleothem Isotope Synthesis and Analysis version 2 (SISALv2) data set (Comas-Bru et al., 2020a).The speleothem values of Δδ 18 O were converted in drip water as performed by Cauquoin et al. (2019), using the respective experiments and method outlined by Tremaine et al. (2011).

Analysis Method
To investigate the impact of atmospheric variations across multiple timescales on moisture transport to Antarctica, we decoupled the contributions of synoptic (associated with EPEs), intra-to inter-seasonal (e.g., the Southern Annular Mode: SAM), and climatological scale meridional moisture transports by following Newman et al. (2012).First, the climatological mean state was determined from daily time-series.The residuals of daily time-series anomalous from the daily climatology were further decoupled to low-frequency and synoptic.The low frequency was determined using a Lanczos low-pass filter, which passes periods greater than 10-day, and the residual represented periods less than 10-day (synoptic state) using the SciPy library.Such filtering was applied to specific humidity (q) and meridional wind speed (v) in the model vertical layers.We finally obtained the vertically integrated moisture transport (Q) in each time scale as follows: /g, and

10.1029/2024GL108191
Here, p s is the surface pressure, and g is the gravity constant.The overbar represents the time mean, and the subscripts m, LF, and syn are the climatological mean, the low frequency, and the synoptic time scales, respectively.

Evaluation of Last Glacial Maximum Climate Simulations in MIROC5-Iso
We globally assessed the simulated δ 18 O p in LGM_G and LGM_M (Figures S2 and S3 in Supporting Information S1).In this section, all references to LGM_G and LGM_M are relative to PI; Δ denotes changes from PI to LGM.We confirmed that the spatial distributions of Δδ 18 O p of both LGMs (LGM_G and LGM_M) were similar in global perspective, with polar amplification in both NH and SH, consistent with previous studies (Cauquoin  In the more detailed assessment of the simulated Antarctic Δδ 18 O p , the changes in isotopic ratios exhibited different spatial distributions among LGM_G, LGM_M, and the ice cores (Figures 1a and 1b and Figure S4a in Supporting Information S1).In some ice core sites, degrees of discrepancy between Δδ 18 O p and Δδ 18 O ice varied between LGM_G and LGM_M (Figure S4a in Supporting Information S1).At EDML, Vostok, and Dome C, LGM_M was more consistent with the ice cores than LGM_G, whereas LGM_G overestimated the δ 18 O reduction for ∼1-3‰ in these sites.At WDC and Byrd, LGM_G was in a good agreement with the ice cores, whereas LGM_M underestimated the δ 18 O reduction by ∼3‰ in these sites.In summary, LGM_G and LGM_M simulated by MIROC5-iso were generally in better agreement with isotopic observation in West and East Antarctica, respectively.
Our results provide a reasonable basis for discussing Antarctic δ 18 O p in the LGM, despite the need to consider model dependencies in future work.For certain ice core sites, our Δδ 18 O p values differed from those of the LGM experiments by Cauquoin et al. (2023) carried out with ECHAM6-wiso.For example, the δ 18 O p reduction at Dome C was larger than that at Dome Fuji in MIROC5-iso and the ice core records, whereas the opposite was modeled in ECHAM6-wiso.Still, the perspective differences in MIROC5-iso experiments, Δδ 18 O p values in LGM_G were lower than the ones in LGM_M at most Antarctic ice core sites, were common in those of ECHAM6-wiso.
Simulated ΔT a at the 10 ice core sites in both LGMs, in contrast to Δδ 18 O p , showed only slight variations (Figure S4b in Supporting Information S1) and was less sensitive to ocean boundary conditions than Δδ 18 O p (Figure S4a in Supporting Information S1).The simulated ΔT a at these sites showed significant discrepancies with Δδ 18 O iceinferred ΔT a , highlighting the challenges in reconstructing temperatures from isotopic records in polar ice cores.Decreases in the simulated T a were less pronounced than those in Δδ 18 O ice -inferred ΔT a at EDML, Dome Fuji, Dome B, Vostok, and Dome C and more pronounced at Taylor Dome, Talos, WDC, and Byrd.It is difficult to identify the cause of the discrepancies between the model and the ice core reconstructions.This is because uncertainties such as bias and spatial representativeness exist in both the model and the data.In any case, our results suggest that ΔT a and δ 18 O p at these Antarctic sites have different determination processes.The temporal ΔT a / δ 18 O p slope should vary from site to site, cautioning against reconstructing ΔT a from Δδ 18 O ice using the same ΔT a /Δδ 18 O slope values over a wide area of Antarctica.

Role of the Atmospheric Circulation in Determining Δδ 18 O p in Inland Antarctica During the Last Glacial Maximum
In this section, we discuss the processes contributing to δ 18 O p values in Antarctica, in relation to the southern midhigh latitude atmospheric circulations.We paid attention to which time-scale atmospheric circulation is important because several previous studies on modern climate suggested the substantial contribution of synoptic-scale phenomena on Antarctic δ 18 O p .According to the time-scale decomposition analysis for PI, LGM_G, LGM_M, and LGM_Mw/Gice, southward moisture transport toward inland Antarctica was primarily driven by Q syn , followed by Q LF , with minimal contribution from Q m (Figure S5 in Supporting Information S1).This finding is consistent with the modern climate ones of reanalysis data (Newman et al., 2012).These results highlight the importance of understanding changes in Q syn across the experiments, as they pertain to the moisture transportation processes toward inland Antarctica, hence inland Antarctic δ 18 O p .
To this end, we analyzed the individual effects of SST and SIC on the southern mid-high latitude climate, including Q syn , and Antarctic δ 18 O p .This was accomplished by comparing LGM_Mw/Gice sensitivity experiments with LGM_G and LGM_M.

SST Substitution Impacts (LGM_G Minus
LGM_Mw/Gice) We investigated the effects of SST substitution from LGM_Mw/Gice to LGM_G on inland Antarctic δ 18 O p in relation to the southern mid-latitude atmospheric circulation.As a feature of the substituted SST, the meridional SST gradient steepened (Figure S6 in Supporting Information S1), with a SST increase in north of ∼40°S and a decrease in south of ∼40°S (Figure 1e).The steepened meridional SST gradient coincided with the enhancement of the Eady growth rate as an indicator of the baroclinic instability (Figure S7 in Supporting Information S1) and with a strengthening of the westerly subpolar and subtropical jets (Figure 1e).The strengthened subtropical jets, which link to the storm track along the subarctic SST front (Nakamura et al., 2013), may induce the observed increase in southward Q syn (Figure 1g) and a slight warming (Figure S8a in Supporting Information S1) in the South Pacific, where the south of subarctic SST fronts located at ∼55°S (Figure S6 in Supporting Information S1).These changes in atmospheric circulation coincided with ∼5% and ∼2‰ increases in precipitable water (Figure 1g) and δ 18 O of precipitable water (δ 18 O prw ; Figure S9a in Supporting Information S1) in inland Antarctica, resulting in a ∼1‰ increase in δ 18 O p (Figure 1c).
The spatially heterogeneous Δδ 18 O p values (Figure 1c) were attributed to these atmospheric circulation patterns.The Δδ 18 O p in inner Antarctica was relatively high in ∼120-180°E, which was downstream of the Q syn intrusion to Antarctica (Figure 1g).The increase in southward Q syn in the South Pacific suggests that the changes in the frequency and/or intensity of the precipitation pattern of the flow from West Antarctica (Dittmann et al., 2016;Schlosser et al., 2017) may be a crucial response to the SST substitution.In contrast to inland Antarctica, the Δδ 18 O p over sea ice in the South Atlantic and Indian Ocean decreased by ∼0.5-3% (Figure 1c).Coincident decreases in southward Q syn (Figure 1g), δ 18 O prw (Figure S9a in Supporting Information S1), and precipitable water (Figure 1g) over sea ice suggest that strengthening of the subpolar jet (Figure 1e) may prevent oceanic vapor containing relatively high δ 18 O prw to present in the polar side of the jet, resulted in the decrease in δ 18 O p .
The Q syn toward Antarctica and various blocking patterns related to EPEs (Wang et al., 2023), along with variations in their occurrence frequency, would explain the relationship between EPEs and Δδ 18 O p . Figure 1j represents the climatological δ 18 O p without daily precipitation weighting (hereafter referred to as δ 18 O a , see Text S2 in Supporting Information S1).Because δ 18 O a should underestimate the contributions from days with EPEs, the differences between Δδ 18 O p and Δδ 18 O a can be considered as indicative of the impact of EPEs.In contrast to Δδ 18 O p (Figure 1i), the small absolute values of Δδ 18 O a indicate that the effects of SST substitution were mostly related to the EPEs.Practically, the impact of the EPEs was pronounced especially in the South Pacific sector of Antarctica (120°E 60°W; Figure S10a in Supporting Information S1).Such spatial heterogeneity indicates the contribution of the EPEs to Δδ 18 O p was uneven among sites.The overall results of the SST substitution suggest that simulating the reasonable precipitation intensity, frequency, and related atmospheric circulation states was crucial to estimating the contribution of the EPEs to Δδ 18 O p at each Antarctic ice core site.
We should note that Q m and Q LF would play indirect roles in determining Antarctic δ 18 O p .Signs of changes in southward Q m over Antarctica (Figure S11b in Supporting Information S1) were opposite to those of Q syn (Figure 1g).The southward Q m decreased in ∼60-180°W where the southward Q syn increased, and vice versa in the remaining longitudes.Such relationship suggests that Q m plays a compensatory role for Q syn in Antarctica, notably in ∼60-180°W.Q LF showed little change over Antarctica (Figure S11a in Supporting Information S1).These minor changes suggested low-frequency atmospheric circulation variations played a minor role in moisture transportation toward inland Antarctica.However, it is important to consider that the low-frequency variations may indirectly influence the moisture fluxes via synoptic-scale circulations, as suggested by Kino et al. (2021).The role of Q LF related to the large-scale circulation patterns, such as the SAM and the Pacific-South Atlantic patterns, in determining inland Antarctic δ 18 O p during LGM is an important future issue.

SIC Substitution Impacts (LGM_Mw/Gice Minus LGM_M)
We investigated the effects of the SIC substitution from LGM_M to LGM_Mw/Gice, specifically SIC expansion in SH (Figure 1f), to inland Antarctic δ 18 O p , with respect to atmospheric circulations.Compared to the SST substitution case, changes in atmospheric circulations, namely westerly jets (Figure S12 in Supporting Information S1) and Q syn (Figures S13 in Supporting Information S1), due to the SIC substitution were less evident.
The SIC substitution rather resulted in a ∼5°C T a decrease over sea ice (Figure S8b in Supporting Information S1).The surface cooling coincided with a ∼10% and ∼4% decrease in precipitable water (Figure 1h) and δ 18 O prw (Figure S9b in Supporting Information S1), respectively, resulting in a ∼5‰ decrease in δ 18 O p over sea ice-covered areas (Figure 1d).These changes were consistent with the sea ice expansion effect on δ 18 O p , described by Lee et al. (2008).To resume, the SIC substitution changed the mean isotopic environment around Antarctica.The T a reduction and transport distance extension promoted isotopic fractionation during the moisture transport from the ice-free ocean to Antarctica, resulting in a ∼3‰ average decrease in δ 18 O p over Antarctica (Figure 1d).
Unlike the SST substitution case, the isotope ratios were substantially reduced in δ 18 O a (Figure 1j), supporting that the SIC substitution changed the mean isotopic environment around Antarctica.The substantial Δδ 18 O a suggests that EPEs alone did not control the δ 18 O p reduction.The effects of SIC substitution on the EPEs were evidenced by the more pronounced variations in Δδ 18 O p (Figure 1i) than Δδ 18 O a among the sites (Figure 1j).Such spatially uneven influences of EPEs can partly be explained by the sea level pressure anomaly due to the SIC substitution (Figure 1f).The SIC expansion from LGM_M to LGM_Mw/Gice was zonally non-uniform, particularly evident in the region of ∼0-120°E (Figure 1f).This localized SIC expansion induced the highpressure anomaly (Figure 1f) that eased the stagnation of Rossby waves.Such blocking can influence oceanic inflows associated with EPEs, thus inducing the evident Δδ 18 O p minus Δδ 18 O a around Dome Fuji and Dome B (Figure S10b in Supporting Information S1).

Influence of Sea Surface Conditions on Southward Moisture Transport and δ 18 O p in Inland Antarctica: Discussion and Perspectives
Our results revealed that the sea surface conditions influence southward moisture transport and δ 18 O p in inland Antarctica through atmospheric circulations.Figure 2 illustrates the main findings of our study.Our three LGM experiments indicate that the synoptic circulations in the southern mid-latitudes, overlooked in studies utilizing the classical one-dimensional isotope model approach, are crucial in understanding δ 18 O p in inland Antarctica during the LGM.The impacts of the SST and SIC substitutions differed among sites, particularly with respect to EPEs.Appropriate sea surface conditions, atmospheric circulations, and Antarctic precipitation intensities and frequencies are essential for adequately representing δ 18 O ice in iso-GCMs.The spatial relationship of Δδ 18 O ice across Antarctic ice core sites could be a reliable indicator of these atmospheric circulations.Our study further suggests that investigating Antarctic ice cores with iso-GCMs can aid in more precisely constraining the behavior of the SH atmospheric circulations, represented by westerlies, which affect the SO circulation that traps atmospheric CO 2 in the deep ocean (Toggweiler et al., 2006), during LGM.
Certain biases in MIROC5-iso and methodological limitations underscore the need for enhanced modeling to constrain LGM climate.MIROC5-iso overestimated summer precipitation intensity at Dome Fuji in the modern climate, which could be attributed to its low spatial resolution (Kino et al., 2021) Such biases would affect Δδ 18 O p and its relationship with ΔT a and partly explain the spatial inconsistency between MIROC5-iso in T42 and ECHAM6-wiso in T63 (Cauquoin et al., 2023).Detailed comparisons among iso-GCMs are necessary to optimize the constraints of the LGM climate.The westerly jets of MIROC5 are biased equatorward in the modern climate (Watanabe et al., 2010).This bias in the modern climate may be associated with the features of mid-latitude circulations in the LGM, according to Sime et al. (2016).Internal variability should also be discussed.A combination of multiple moisture transport patterns, including blocking, is crucial to determine δ 18 O p (Dittmann et al., 2016;Schlosser et al., 2017;Wang et al., 2023) and precipitation amounts (Wang et al., 2023) in Antarctica; 30-year analyses may be insufficient to evaluate the frequency of each pattern.
For further exploration of the link between the sea surface conditions and δ 18 O p in inland Antarctica, a colored moisture analysis (e.g., Yoshimura et al., 2004) would be effective.Such analysis would allow us to estimate from which latitudes or ocean basins the precipitation water isotopes originated.It is also worthwhile to examine changes in moisture origins in the modern (e.g., Dittmann et al., 2016;Schlosser et al., 2017) and past climates.A related limitation of this study is the lack of assessment of D-excess, which is the second-order index defined as the departure from the simple relationship between δ 18 O and deuterium isotope ratio (δD).Combination of these isotope indices has been utilized to decompose source and site temperature contributions on Antarctic ice cores (Uemura et al., 2012(Uemura et al., , 2018) ) and to estimate potential oceanic source regions (Uemura et al., 2018).Through the comprehensive analyses discussed above in the synoptic timescale, we will finally be able to understand how the use of iso-GCMs versus simple one-dimensional models or the traditional isotopic thermometer assumption leads to systematic differences in the reconstruction of past temperatures inferred from water isotope signals in Antarctic ice cores.
Further experiments forced by idealized SST and SIC (e.g., zonally uniform) will be performed to isolate and examine more precisely the impacts of SST gradients on atmospheric dynamics, leading to a better understanding of the mechanisms linking sea surface conditions to Antarctic δ 18 O p .Further investigation is required to determine the effects of changes in ice sheet elevation (Werner et al., 2018), precipitation seasonality (Erb et al., 2018), and the inversion layer depths (Buizert et al., 2021) on δ 18 O p in inland Antarctica.The global reproducibility of LGM climates is also important to optimize Antarctic temperature estimations, as δ 18 O p in inland Antarctica is associated with the broader SH.The two sets of LGM sea surface conditions used in this study were not the end members of PMIP models (Kageyama et al., 2021).Multiple LGM simulations would be required to constrain the LGM Antarctic temperatures.

Data Availability Statement
• Ice core data used for Figures S2 and S3 in Supporting Information S1 are reported in Cauquoin et al. (2019).Ice core data used for Figure 1, except for the South Pole, are available in Table 1 of Werner et al. (2018).Data for the South Pole is available in Steig et al. (2020).The SISAL speleothem data set is available in Comas-Bru 10.102910. /2024GL108191 et al. (2020a)).The GLOMAP data set is available in Paul et al. (2020).The SST and SIC outputs from MIROC4m-AOGCM are available from the authors of Sherriff-Tadano et al. ( 2023).• The code of the isotopic version MIROC5-iso is available upon request on Okazaki and Yoshimura (2019b).
• The source codes and data used in this study are available in Kino et al. (2024aKino et al. ( , 2024b)).

Figure 1 .
Figure 1.Simulated climatological anomalies.(a) δ 18 O of precipitation (δ 18 O p ) changes from Pre-Industrial (PI) to LGM_G in 50-90°S.(b) As in panel (a), but for PI to LGM_M.(c)-(d) Changes due to sea surface temperature (SST) and sea ice concentration (SIC) substitutions: (c) from LGM_Mw/Gice to LGM_G; (d) fromLGM_Mw/Gice to LGM_M.(e) As in panel (c), but for SST (shades) and 250 hPa zonal wind (solid and thin purple contours in every 4 and 1 m/s, respectively) in 20-90°S.(f) As in panel (d), but for SIC (shades) and sea level pressure (SLP; solid and thin purple contours in every 2 and 0.5 hPa, respectively) in 20-90°S.(g) As in panel (e), but for precipitable water (shades) and synoptic-scale southward moisture transport (Q syn ; solid and thin purple contours in every 5 and 1 kg/m/s).(h) As in panel (f), but for precipitable water (shades).(i) Changes in δ 18 O p at 10 Antarctic sites due to the SST (red), SIC (blue), and both SST and SIC (from LGM_M to LGM_G; black) substitutions.(j) As in panel (i), but for δ 18 O p without daily precipitation weighting (δ 18 O a , see Text S2 in Supporting Information S1).For panels (e)-(h), positive anomalies for zonal winds, SLP, and Q syn are depicted with solid contours, respectively, while negative anomalies are dashed.For precipitable water in panels (g)-(h), changes are presented as ratios, while for the remaining variables and panels, they are presented as anomalies.For panels (a)-(d), (f), and (h), 15% SIC in PI, GLOMAP, and MIROC (thin solid, thick dashed, and thick dotted black contours, respectively) are denoted.For (a)-(h), each panel denotes 10 Antarctic ice core sites with circles; for (a)-(b), these are overlaid with Δδ 18 O ice .Refer to Section 2.2 for specifics on the sites.

Figure 2 .
Figure 2. Schematic view of the processes ruling the Δδ 18 O p in inland Antarctica during Last Glacial Maximum (LGM) proposed by this study.The red and blue colors represent the key processes, typically in the South Pacific, associated with the substitution of SST (LGM_G minus LGM_Mw/Gice) and sea ice (LGM_Mw/Gice minus LGM_M), respectively.The upward and downward arrows represent the increases and decreases of the values of the variables, respectively.
Risi et al., 2010).MIROC5-iso reasonably reproduced Δδ 18 O of both LGMs with a root mean square error of ∼3.1‰ between the modeled and observed values(Figures S2b and S3b in Supporting Information S1).From a regional perspective, degrees of discrepancies between the modeled and observed Δδ 18 O values varied based on the sea surface boundary conditions.The positive Δδ 18 O p in the tropics (+1-2‰) was significant in LGM_G (FigureS2ain Supporting Information S1), resulting in a better model-data agreement for tropical speleothems than in LGM_M (FigureS3ain Supporting Information S1).For Antarctica, the simulated Δδ 18 O p values in both LGMs were reasonable compared to observed Δδ 18 O ice values (Figures1a and 1band FigureS4ain Supporting Information S1).In the simulations, the respective Δδ 18 O p values of both LGMs were similar at Dome Fuji, South Pole, and Dome B and consistent with the Δδ 18 O ice at Dome Fuji and Dome B. At Taylor Dome, Talos, and South Pole, observed Δδ 18 O ice values were substantially different from Δδ 18 O p of both LGMs (FigureS4ain Supporting Information S1).Δδ 18 O p at Taylor Dome and Talos may relate to the substantial elevation increases (more than 500 m) in the Ross Sea of the GLAC-1D data set.At South Pole, the δ 18 O reduction was underestimated by 2.8-3.0‰ in both LGMs, compared to the ice core value.Potential reasons for the substantial reduction of Δδ 18 O p values at South Pole are discussed in Section 4.
Schlosser et al. (2017)f changes in δ18O at South Pole from PI to LGM, reported in Section 3.1, might be partly due to unrealistic changes in moisture conversion and precipitation.In LGM_G, despite high Δδ 18 O p values derived from daily precipitation weighting (Figures 1a), Δδ 18 O a value was close to Δδ 18 O ice at South Pole (FigureS4ain Supporting Information S1).The similar values of δ 18 O a and Δδ 18 O ice in LGM_G suggest that overly frequent EPEs accompanied by abrupt warming contributed to the excessively high Δδ 18 O p at South Pole.However, in LGM_M, the frequency of EPEs accompanied by abrupt warming could not explain the too high Δδ 18 O p at South Pole in LGM_M (Figure1b); the differences between Δδ 18 O a and Δδ 18 O ice values at South Pole ( 3.7 and 6.5‰, respectively) were substantial.To explain the model-data discrepancy at South Pole, we should explore changes in the frequency and intensity of EPEs and their relation to circulation patterns, which has been classified into several patterns including EPEs associated with abrupt warming and radiative cooling byDittmann et al. (2016)andSchlosser et al. (2017).