Simulated Cycles of East Asian Temperature and Precipitation Over the Past 425 ka

Records from a wide range of geological archives covering the last few glacial‐interglacial cycles show large inconsistencies in the East Asian summer monsoon variability, which severely hampers our understanding of the evolution and potential mechanisms of the regional East Asian climate on orbital timescales. Here, we examine the simulated temperature and precipitation in East Asia based on a series of equilibrium simulations conducted for the past 425 ka, and we investigate the sensitivity of temperature and precipitation to potential forcings. Our simulations show that, in East Asia, the seasonal mean temperature is dominated by a ∼20‐kyr cycle, and the annual mean temperature (AMT) is dominated by a ∼100‐kyr cycle, which is consistent with previous modeling efforts and geological reconstructions. Additional sensitivity experiments indicate that the greenhouse gas concentration, in combination with the ice volume, is the dominant force for the variations of AMT in East Asia on orbital timescales. For the precipitation in East Asia, our equilibrium simulations and additional sensitivity experiments, together with comprehensive model‐data intercomparison analysis, suggest that the cycles of simulated annual mean precipitation over East Asia are highly model‐dependent, although the dominant ∼20‐kyr cycle in summer precipitation appears to be a robust feature. Overall, the results highlight the large model uncertainty with regard to the relative roles of forcings in hydroclimate variations in East Asia on orbital time scales. There is, therefore, an urgent need to implement more realistic precipitation schemes in models in order to decrease the model spread in simulated precipitation.

2 of 13 et al., 2001) after the mid-Pleistocene (∼1 Ma). However, interpretations of these records are often controversial, in particular when some records are used as a proxy for variations of the East Asian summer monsoon (EASM). For example, the EASM interpreted from the magnetic susceptibility and chemical weathering indices from loess in North China (NC) shows a dominant ∼100-kyr cycle (Figure 2b; Guo et al. [2000]; Hao et al. [2012]). In contrast, if stalagmites δ 18 O in the Yangtze River Valley (YRV) are interpreted as a proxy of EASM, monsoon intensity would follow a dominant ∼20-kyr cycle (Figure 2c; e.g., Cheng et al. [2016]; Y. Wang et al. [2008]). Therefore, these paleoclimate records may reflect different aspects of the East Asian monsoon climate, such as temperature or precipitation, and not only the monsoon intensity.
In the past decade, many modeling studies have investigated the variation of East Asian monsoon climate (e.g., Liu et al., 2014;Lu et al., 2013;Wen et al., 2016). However, the cycles of simulated climate variables -and in particular precipitation -are complex and sometimes ambiguous. Although these simulations show a consistent ∼20-kyr cycle of simulated summer precipitation in East Asia (e.g., Li et al., 2013;Lyu et al., 2021;Weber & Tuenter, 2011;Xie & Liu, 2020), they also demonstrate a remarkable model-spread in the cycle of simulated annual precipitation (Clemens et al., 2018;Li et al., 2013;Li, Liu, Zhao., 2017;Lyu et al., 2021;Sun et al., 2019;Thomas et al., 2016). For example, the annual precipitation in NC according to the simulation with CCSM 3.0 shows a strong dominance of the ∼20-kyr cycle (Li et al., 2013;Li, Liu, Zhao., 2017) while the ∼100-kyr cycle is more or less absent. Simulations with HadCM3, on the other hand, show that the ∼100-kyr cycle is also dominant in addition to the ∼20-kyr cycle (Lyu et al., 2021;Sun et al., 2019).
In summary, there are presently large uncertainties associated with reconstructions of East Asian summer monsoon intensity based on different geological records, and many discrepancies are found in the simulated precipitation from current transient modeling studies over the past glacial-interglacial cycles. Here, to shed some further light on the above-mentioned proxy data and model uncertainty, we analyze a series of equilibrium experiments over the past 425 ka simulated with the low-resolution version of the Norwegian Earth System Model (NorESM-L). These experiments have included variations in orbital parameters, ice sheets, and greenhouse gas levels over the  (Berger, 1978); (c) CO 2 concentrations (Luthi et al., 2008); (d) LR04 δ 18 O (Lisiecki & Raymo, 2005). 3 of 13 past 425 ka (Dai et al., 2021;Yan et al., 2021;Zhang et al., 2020), and can well simulate the East Asian climate change during glacial-interglacial cycles. For example, a tripolar precipitation pattern in monsoonal China was well presented by using these equilibrium experiments, which is supported by some available geological proxies (Dai et al., 2021). Furthermore, using the Community Earth System Model (CESM) and NorESM-L, we conduct several sensitivity experiments, to compare the sensitivity of temperature and precipitation in East Asia to individual forcings (including orbital parameters, ice sheets, and greenhouse gas levels). By using two different Earth System Models we can also estimate model-induced dependencies.
This article is organized as follows: Section 2 introduces the methods, including the model introduction, related experimental design as well as the analytical method. Section 3 describes the simulation results. Section 4 presents the discussion and conclusions.

Model Introduction
NorESM-L is a fully coupled Earth System Model, which includes the atmosphere (CAM4), land (CLM4), sea ice (CICE4), ocean (MICOM), and coupler (CLP7) components. The ocean component uses the horizontal resolution corresponding to a nominal grid size of 3° and the vertical resolution of 32 layers. The atmosphere component uses the horizontal resolution of T31, corresponding roughly to a grid size of approximately 3.75°, and the vertical resolution of 26 levels. The atmosphere component includes the physical parameterizations of convection, clouds, surface processes, and turbulent mixing (Neale et al., 2013), and simulates the temperature and precipitation in East Asia realistically (e.g., Zhang et al., 2017). The land component follows the atmospheric spatial grid, while the sea-ice module uses the ocean horizontal grid. Details of NorESM-L are introduced in Zhang et al. (2012).
In addition to NorESM-L, CESM version 1.2.2, a widely used model (e.g., Goldner et al., 2013), is applied for carrying out additional sensitivity experiments. CESM is an earth system model developed by National Center for Atmospheric Research. CESM has the atmosphere (CAM4), land (CLM4), sea ice (CICE4), ocean (POP2), and coupler (CLP7) components. In this study, the POP2 runs at a horizontal resolution of nearly 1° × 0.5° and a vertical resolution of 60 levels; the CAM4 runs at a horizontal resolution of 1.9° × 2.5° and a vertical resolution of 26 levels.
There are two main differences between the NorESM-L and CESM 1.2.2 simulations. (a) The resolutions in the atmosphere component CAM4 adopted in the two models are different. The horizontal resolution used in our NorESM-L simulations is lower than in our CESM simulations, meaning that the two models employ different cloud physics and precipitation parameterization schemes. (b) The other difference is the ocean component, and this leads to different ocean feedback.

Experimental Design
Based on the relative maxima, minima, and median points of July insolation at 65°N during the past 425 ka, we selected 77-time slices in total for model simulations (Figure S1a in Supporting Information S1; Berger [1978]). For each time slice, we use NorESM-L to carry out the equilibrium simulation forced with the orbital parameters (Berger, 1978), greenhouse gas CO 2 (Luthi et al., 2008), and CH 4 (Loulergue et al., 2008), as well as the simulated Northern Hemisphere ice volumes (Details can be seen in Supporting Information; Zhang et al. [2020]) at the selected time points. In addition, the PI simulation is conducted according to the standardized experimental design from the Paleoclimate Modeling Intercomparison Project Phase 3 (PMIP3; see the PMIP3 website at https://pmip3.lsce.ipsl.fr/). In our simulations, the vegetation cover in East Asia remains the same as in the pre-industrial (PI) experiment. Detailed information on the 77-time slices used and the experimental flow can be found in previous studies (Dai et al., 2021;Yan et al., 2021;Zhang et al., 2020).
Relative to the often-used reconstructed ice sheets from the ICE6G dataset (Peltier et al., 2015), the simulated Northern Hemisphere ice sheets by Zhang et al. (2020) differ by including the fast waning and waxing of an ice sheet on Northeast Siberia-Beringia. Additionally, the simulated Northern Hemisphere ice sheets underestimate the total volume of Northern Hemisphere ice sheets, compared to ICE6G. Here we apply simulated Northern Hemisphere ice sheets as primary boundary condition/forcing. However, to evaluate the impact of ice sheet uncertainty, we also analyze the output of 21 time slice simulations initiated with the reconstructed ice sheets from the ICE6G dataset (Peltier et al., 2015) over the past 126 ka (Figure S1b of Supporting Information S1; also see Zhang et al. [2020]).
To investigate the East Asian climate sensitivity to precession, obliquity, CO 2 concentrations, and ice sheets, we additionally design a set of sensitivity experiments (Table 1). To address the climate response to precession, two experiments are performed by using a high precession value corresponding to the longitude of perihelion is about 90° and a low precession value corresponding to the longitude of perihelion is about 270°, with all other conditions being identical to the PI experiment. In the same way, we perform two experiments to address the climate sensitivities to the obliquity; one with a high obliquity value (Obliquity = 24.5°) and one with a low obliquity value (Obliquity = 22.1°). To address the climate sensitivities to the atmospheric CO 2 concentration, we conduct an experiment with the CO 2 concentration set at 184.5 ppmv (a low level at the Last Glacial Maximum [LGM]), whereas the PI experiment uses a level of 280 ppmv. Finally, we design two experiments to address the climate sensitivities to the ice sheets. One is the LGM experiment following the PMIP3 guideline, and the other is similar to the LGM experiment, but with modern ice sheets used. To assess potential model dependency, we use NorESM-L and CESM to carry out all these sensitivity experiments. Each experiment is run for 500 model  Table 1 Earth's Orbital Parameters (Berger, 1978), CO 2 Concentration (Petit et al., 1999), and Ice Sheet (Peltier et al., 2015) in the CESM and NorESM-L Sensitivity Experiments years to reach quasi-equilibrium, and the outputs of the last 100 model years are analyzed.

Study Area and Time Scales
We use the indices of East Asian summer monsoon intensity (EASMI) following the definition from Zhu and Wang (2001) and East Asian winter monsoon intensity (EAWMI) following the definition from Gong et al. (2002). These two indices are calculated based on sea level pressure, and their detailed calculation formulas can be seen in the Supporting Information S1. Moreover, we choose three areas in East Asia (marked in Figure 2a) for the calculation of regional averaged temperature and precipitation. These areas include North China (NC) between 35° and 43°N and 101°-116°E, the YRV between 27° and 33°N and 110°-120°E, and South China (SC) between 6° and 24°N and 112-120°E, with the South China Sea (SCS) included. We use the 78 equilibrium experiments (77 paleo-simulations and a PI control experiment) to compose climate time series during the past 425 ka. The average temporal resolution of these series is ∼5.5 ka, which is suitable for the spectrum analysis on the orbital time scale. Here we carry out the spectral analyses for the simulated elements with significance tests employed (Li et al., 2019).

Simulated Cycles in East Asian Climate
Our simulations demonstrate a remarkable seasonal contrast-between summer and winter-in the dominating cycles of simulated East Asian monsoon intensity during the past four glacial-interglacial cycles ( Figure 3). The ∼20-kyr cycle is dominant for the EASMI, while the ∼40-kyr cycle is less strong and the ∼100-kyr cycle does not exist ( Figure 3a). However, the ∼100-kyr cycle is dominant in the EAWMI, and the ∼40 and ∼20-kyr cycles are less strong (Figure 3b). This contrast does not appear in the cycles of simulated seasonal mean temperature. The simulated mean temperatures for all four seasons are dominated by the ∼20-kyr cycle ( Figure 4). However, the simulated annual mean temperature (AMT) in East Asia is mainly dominated by the ∼100-kyr cycle, and not the ∼20-kyr cycle ( Figure 6a). In addition to the ∼100-kyr cycle, the ∼40 and ∼20-kyr cycles also appear in the AMT over NC and SC, while there is no ∼40-kyr cycle in the AMT over the YRV.
The simulated seasonal and annual mean precipitation (AMP) also show distinct differences with respect to dominating orbital cycles. The seasonal mean precipitation is almost entirely dominated by the ∼20-kyr cycle in East Asia (except for the summer precipitation in the YRV; Figure 5). However, the cycles for AMP show large regional differences (Figure 6b). The AMP shows the distinct ∼100, ∼40, and ∼20-kyr cycles in NC, but only the ∼20-kyr cycle (without the ∼100-kyr cycle) in SC. Our simulations, therefore, suggest that the importance of the ∼100 and ∼40-kyr cycles decreases, and the importance of the ∼20-kyr cycle increases, gradually from north to south in the East Asian monsoon region.

Correlations Between Climate Indices and Forcings
Although the one-variable linear regression analysis method only can not show the certain physical relationship between two factors, it remains to be effective for us to employ this method to distinguish the relative importance of the potential four forcings (precession, obliquity, greenhouse gases, and ice volume) to climate factors. Here, we conducted correlation analyses between the simulated climate indices (EASMI, EAWMI, AMT, and AMP) and the four forcings. Overall, these linear regression analyses assume the fact that the dominating cycles of simulated climate indices always follow the key cycles of their potential forcing.   The simulated EASMI has the highest correlation with precession, followed by obliquity (Figures S2a and S2b of Supporting Information S1). Atmospheric CO 2 concentration and ice volume are not correlated to EASMI. These correlation analyses coincide with the dominating ∼20-kyr cycle of simulated EASMI. In contrast, the simulated EAWMI has the highest correlation with the ice volume and atmospheric CO 2 concentration, while no correlations are found for the precession and the obliquity (Figures S2c and S2d of Supporting Information S1). This relationship is also in agreement with the fact that the simulated EAWMI is dominated by the ∼100-kyr cycle.
The simulated AMT in East Asia has the highest correlation with the atmospheric CO 2 concentration (Figures 7a-7c; Figures S3 and S5 of Supporting Information S1), as well as with the ice volume, though the correlation to ice volume is dependent on the region ( Figure S3 of Supporting Information S1). Relatively higher correlations between the AMT and the ice volume exist in NC compared to YRV/SC. The simulated AMT also shows a higher correlation to the ICE6G ice volume (Peltier et al., 2015), when compared to our simulated ice volume which includes some underestimations (Figure 7a-7c; Figures S3 and S5 of Supporting Information S1; Zhang et al. [2020]). These linear regression analyses thus consistently show the larger importance of the CO 2 concentration and the ice volume, rather than obliquity and precession, in controlling the AMT. This strongly supports the suggestion that the ∼100-kyr cycle is dominating the simulated AMT in East Asia (Figure 6a).
Compared to the simulated AMT, the simulated AMP in East Asia shows more variable correlations with the four forcings (Figure 7d and 7f; Figures S4 and S6 of Supporting Information S1). For example, our simulations show that in NC, the AMP has the highest correlation with the ice volume and atmospheric CO 2 concentration, in agreement with the cycles for the simulated AMP in NC, which show the highest spectral power on the ∼100 and ∼40-kyr cycles (Figure 6b). In the YRV, the simulated AMP has the highest correlation with the precession, followed by some correlation to obliquity, but shows no correlations with the atmospheric CO 2 concentration and the ice volume. Again confirming the cyclicity of the simulated AMP in the YRV with the highest spectral power on the ∼20 and ∼40-kyr cycles (Figure 6b). In SC, the simulated AMP only has a high correlation with the precession, again agreeing well with the dominance of the ∼20-kyr cycle in the simulated AMP in SC (Figure 6b).

Evaluation of Sensitivity Simulations
The above linear regression analyses can be further assessed with climate sensitivity experiments. We carried out a total of eight sensitivity experiments (listed in Table 1) spanning full ranges in precession, obliquity, atmospheric CO 2 concentration, and ice volume. We define the sensitivity factors ΔS T (unit: °C/W) and ΔS P (unit: mm/W), the ratio of the AMT and AMP changes to variations in shortwave radiation at the top of the atmosphere (∆FSNT), respectively. The calculation formula is as follows: ΔS T is a temperature sensitivity index (unit: °C/W), defined as the ratio of the annual temperature change (ΔT, unit: °C) to the short-wave radiation energy change at the top of the atmosphere (ΔFSNT, unit: W). ΔS P is a precipitation sensitivity index (unit: mm/W), defined as the ratio of the annual precipitation change (ΔP, unit: mm) to the short-wave radiation energy change at the top of the atmosphere (ΔFSNT, unit: W). For example, imagine that a variation in precession and obliquity both cause a 1W increase in radiation at the top of the atmosphere. If the precession variation leads to a temperature response of 1°C and the obliquity variation in the temperature response of 2°C, this means that the temperature is more sensitive to the obliquity than to the precession. The advantage of this sensitivity analysis is that we can avoid comparing the absolute changes in temperature or precipitation caused by the various forcings with different magnitudes. Combined with this analysis we can further compare the impact of the various forcings on temperature and precipitation, which is more convincing than only using the elementary correlation analysis.  Figure S3 of Supporting Information S1). (d)∼(f) are for the simulated AMP (refer to correlation plots in Figure S4 of Supporting Information S1). From top to bottom: North China (NC), Yangtze River Valley (YRV), and South China (SC). The light orange colors are for the simulations using simulated northern hemisphere ice sheet (Zhang et al. [2020]; n = 78); and the pink colors are for the simulations using ICE6G northern hemisphere ice sheet (Peltier et al. [2015]; n = 22). Here, the LR04 ice volume is indicated by LR04 δ 18 O (Lisiecki & Raymo, 2005), and the simulated northern hemisphere ice sheet index is from Zhang et al. (2020). The ICE6G northern hemisphere ice sheet index is obtained according to Peltier et al. (2015).

Sensitivity Indices With NorESM-L
The sensitivity analyses show that the ΔS T due to the atmospheric CO 2 concentration is the largest in East Asia (Figures 8a-8c). This high sensitivity, which is consistent with the regression relationships between the AMT and atmospheric CO 2 concentration, suggests again that the CO 2 concentration is the dominant forcing for the simulated AMT with NorESM-L in East Asia. The sensitivity analyses highlight some more complicated sensitivities of AMP to the four forcings (Figures 8d-8f). The largest ΔS P is caused by the atmospheric CO 2 concentration in NC, by the obliquity in the YRV, and by the precession in SC. These sensitivities reveal some differences compared to the linear regression analyses. In NC, both the sensitivity and the linear regression analyses highlight the importance of atmospheric CO 2 concentration and ice volume as compared to the orbital parameters. However, the sensitivity analyses suggest a stronger AMP sensitivity to the atmospheric CO 2 concentration than to ice volume in NC (Figure 8d), whereas the linear regression analyses suggest the ice volume is more important (Figure 7d). In the YRV, both the sensitivity and linear regression analyses show the prominence of the orbital parameters. However, the sensitivity analyses suggest that the larger AMP sensitivity is to obliquity rather than to precession in the YRV (Figure 8e), whereas the linear regression analyses suggest the opposite (Figure 7e). In SC, both the sensitivity and the linear regression analyses demonstrate the key role of the precession in forcing the AMP in the NorESM-L simulations (Figures 7f and 8f).

Sensitivity Indices With CESM
In order to test to what degree the simulated AMT and AMP sensitivities to the four forcings are model-dependent, we repeat the sensitivity experiments with CESM. The CESM simulations indicate that the ΔS T due to the atmospheric CO 2 concentration remains the dominant forcing in East Asia, consistent with what was found in the NorESM-L simulations (Figures 8a-8c). This consistency between NorESM-L and CESM models provides further indication that the atmospheric CO 2 concentration is the dominant forcing for the changes of the AMT in East Asia on orbital time scales. However, the CESM simulations show remarkable differences in terms of AMP sensitivities, when compared to the NorESM-L simulations. The CESM simulations show that the largest ΔS P in East Asia is caused by the ice volume, rather than the atmospheric CO 2 concentration or the orbital configurations (Figures 8d-8f).
In NC, both models show the larger importance of the atmospheric CO 2 concentration/ice volume over the orbital configurations. However, CESM suggests the ice volume is more important than the atmospheric CO 2 concentration in controlling the AMP in NC, whereas NorESM-L suggests the opposite. In the YRV and SC, CESM suggests that the ice volume and the CO 2 concentration play more dominant roles in controlling the AMP when compared to the orbital configurations. This result is completely different from the sensitivities revealed in the NorESM-L simulations. Therefore, the comparison between the AMP sensitivities revealed in the CESM and the NorESM-L simulations demonstrates that the cycles in the simulated AMP are highly model dependent.

Discussion and Summary
When we compare our simulations to previous modeling studies, we find that models consistently demonstrate the greenhouse gas levels or the ice volumes play a more important role than the orbital configurations in controlling the AMT in East Asia. The interglacial simulations with LOVECLIM (Yin & Berger, 2012), CESM (Colleoni et al., 2014), and HadCM3 (Sun et al., 2019), as well as the last deglaciation simulation with CCSM3 (Liu et al., 2009), also show the high importance of greenhouse gas forcings. Based on these simulations one can reliably conclude that the glacial-interglacial AMT is dominated by the ∼100-kyr cycle in East Asia. Recently, a previous study also confirmed that the decrease in atmospheric CO 2 concentration and increase of the Northern Hemisphere ice volume is the major reason for the AMT decline during the LGM, while the waning of boreal summer insolation played a less important role (Cao et al., 2019). Despite the nonlinear response of the AMT to potential forcings, both our study and previous experiments demonstrate that the effects of CO 2 concentrations and ice sheets are always more important than the orbital configurations. However, it remains difficult to distinguish the relative contribution of the greenhouse gas concentrations and the ice volumes in modulating the ∼100-kyr cycles due to the large model spread (Figures 8a-8c). For example, the HadCM3 suggests that the greenhouse gas concentrations are more important (Sun et al., 2019), whereas our simulations using the ICE6G ice volume (Peltier et al., 2015) suggest that the greenhouse gas concentrations and the ice volumes play equally important roles (Figures 7a-7c). The previous research suggests that on the orbital time scale, the AMT in East Asia is forced by the meridional thermal gradient, which is highly related to the high-northern-latitude temperature indicated by northern hemisphere ice volume forced primarily by greenhouse gas levels (Liu et al., 2009;Lu et al., 2013).
Our study, together with previous modeling studies, indicates that it is challenging, at least from a modeling point of view, to converge on what is the dominant cycle of AMP in East Asia, largely due to irreconcilable model discrepancies. For instance, the simulation based on CCSM 3.0 demonstrates that the AMP in NC shows the dominant ∼20-kyr cycle, rather than the dominant ∼100-kyr cycle (Li et al., 2013;Li, Liu, Zhao., 2017). Simulations with HadCM3, on the other hand, show that the ∼100-kyr cycle is also dominant in addition to the ∼20-kyr cycle for the AMP in NC (Lyu et al., 2021;Sun et al., 2019). The simulated AMP in the YRV illustrates the dominant ∼69-kyr cycle, then followed by the ∼29-kyr cycle according to transient simulation with CCSM 3.0 (Clemens et al., 2018). One possible reason for these model discrepancies is the different experimental designs in previous individual studies. For example, some previous simulations only considered orbital parameters (e.g., Chen et al., 2011;Li et al., 2013), or generated the outputs by interpolating the simulation of various scenarios during glacial-interglacial cycles (Lyu et al., 2021;Sun et al., 2019), which may cause uncertainties about the simulations. Moreover, not only experimental designs, but also precipitation schemes in models lead to differences (e.g., Clemens et al., 2018;Xie & Liu, 2020). As revealed in our study, although the same sensitivity experiments are carried out, different precipitation parameterization schemes were employed due to the different resolutions for the atmospheric module in NorESM-L and CESM, thus may cause remarkably different responses in the AMP in East Asia (Figure 8d-8f; Figure S8 of Supporting Information S1; Gao et al. [2006]). In addition, different cloud and ocean feedback could also be one possible factor for different responses in the AMP in East Asia in different models (Tian & Jiang, 2013).
However, previous transient simulations (Li et al., 2013;Xie et al., 2020) and Gaussian emulator based on HadCM3 simulations (Lyu et al., 2021), as well as our composited snapshot simulations, consistently demonstrate the dominance of the ∼20-kyr cycle for the seasonal precipitation in East Asia ( Figure 5). Insolation modulated by the orbital parameters, in particular the precession, is the dominant forcing for the seasonal precipitation in East Asia.
Compared to the paleo-temperature reconstructions from East Asia, the cycles of simulated AMT in our study display a good agreement with the cycles revealed in various reconstructions. The paleo-temperature reconstructions from the loess sediments in the Chinese Loess Plateau (Figure 2f; Lu et al. [2019]; Tang et al. [2017]; Thomas et al. [2016]) and the marine sediments in the SCS (Figure 2g; Li et al. [2009]; Thomas et al. [2014]; Wei et al. [2007]) illustrate the high spectral power of the ∼100-kyr cycle, but relatively low spectral power of the ∼40 and ∼20-kyr cycles. This agreement further confirms that the greenhouse gas levels or the ice volumes, rather than the orbital parameters, are the dominant forcing in modulating AMT during the past glacial-interglacial cycles. It seems that most paleo-temperature reconstructions reflect the annual mean signal. They are not necessarily indicators of a seasonal signal (e.g., Lu et al., 2019), since the ∼20-kyr cycle is dominant in the seasonal mean temperature in East Asia (Figure 4). Although the stalagmites δ 18 O in East Asia show a dominant 20-kyr cycle (Figure 2c), its paleoclimatic implication is still under debate.
In relation to the paleo-temperature reconstructions from East Asia, our study suggests that the current paleo-precipitation derived from the geological records remains complicated, and maybe contain some uncertainties. Although some geochemical proxies (δ 13 C, 10 Be, Sr/Ca, and δDwax) from loess sediments are all used to indicate precipitation on the Chinese Loess Plateau, they show remarkably different cyclicity (Figures 2d  and 2e) and different trends to some extent ( Figure S9 of Supporting Information S1). Many climate processes can influence geochemical proxies recorded in sediments. For instance, Sun et al. (2015) found that the δ 13 C value from the loess is affected by integrated processes, such as precipitation, evaporation, and relative humidity during the vegetation growing season, which may cause the trend of loess δ 13 C value is not entirely same with other proxies during the Marine Isotopic Stages 6a ( Figure S9 of Supporting Information S1). Unfortunately, our study suggests that modeling results provide little help in constraining the uncertainties in the interpretations of precipitation reconstructions from geochemical proxies. However, it can be stated with relatively high confidence that the AMP in East Asia, at least in certain regions such as the Chinese Loess Plateau, should include a strong ∼100-kyr cycle, in addition to the ∼20-kyr cycle (Lyu et al., 2021;Sun et al., 2019;Thomas et al., 2016).
Our study further reveals the advantage and limitations of current models in simulating East Asian monsoon climate on the orbital time scale. Validation for the AMT and AMP over East Asia shows that models simulate temperature overall more realistic than precipitation (Figures S7 and S8 of Supporting Information S1). The simulated precipitation in East Asia is highly model-dependent. In this sense, it remains difficult to obtain consistent results in the periodic investigation of AMP in East Asia, whereas the simulated cycles for the AMT will become more consistent and convincing in the glacial-interglacial experiments with different climate models. Our findings further point out which forcings have the greatest impact on the East Asian temperature, but conducting the quantitative atmosphere dynamic analysis in the future is also necessary to further figure out how the dominant forcing affects the temperature in detail. Meanwhile, investigating the source of uncertainty of the simulated precipitation cycle in East Asia is also an important issue but highly depends on carrying out suitable types of modeling experiments.
In summary, we use a series of equilibrium simulations to examine cycles in the simulated temperature and precipitation in East Asia during the past 425 ka. Our simulations demonstrate that the seasonal temperature is dominated by the ∼20-kyr cycle and the AMT is dominated by the ∼100-kyr cycle in East Asia, in good agreement with previous modeling and reconstruction studies. Our sensitivity analyses further demonstrate that the greenhouse gas concentrations and the ice volumes play key roles in modulating the AMT in East Asia on orbital timescales. Furthermore, our study demonstrates that the cycles revealed in the simulated AMP are highly model-dependent, though the dominant ∼20-kyr cycle in summer precipitation remains a robust feature over East Asia. Therefore, there is an urgent need to improve the ability of precipitation simulations in the future, particularly for winter precipitation.