Transient simulation of Holocene climatic and isotopic variations across global monsoon regions

Shi, Xiaoxu; Liu, Xiya; Liu, Jiping; Werner, Martin; Yang, Hu; Cruz, Francisco W.; Yang, Chaoyuan; Lohmann, Gerrit

doi:10.5194/egusphere-2026-3894

Preprints

https://doi.org/10.5194/egusphere-2026-3894

Preprints

28 Jul 2026

| 28 Jul 2026

Status: this preprint is open for discussion and under review for Climate of the Past (CP).

Transient simulation of Holocene climatic and isotopic variations across global monsoon regions

Xiaoxu Shi, Xiya Liu, Jiping Liu, Martin Werner, Hu Yang, Francisco W. Cruz, Chaoyuan Yang, and Gerrit Lohmann

Abstract. Speleothem oxygen isotopes provide key insights into Holocene monsoon variability, but the scale-dependent and regionally variable controls on precipitation δ¹⁸O remain poorly resolved. Here, we use a continuous 8.3 ka isotope-enabled transient simulation based on AWI-ESM2-wiso to investigate Holocene climatic and isotopic changes and quantify driving mechanisms for the isotopic variations via a four-predictor multiple linear regression (MLR) model. The simulation successfully reproduces the orbitally driven long-term weakening of Northern Hemisphere monsoons and the bipolar hydroclimatic response to the 8.2 ka freshwater perturbation, characterized by a southward Intertropical Convergence Zone (ITCZ) shift, Northern Hemisphere drying and isotopic enrichment, and Southern Hemisphere moistening and depletion. MLR results show that regional precipitation amount serves as the dominant and temporally stable control on monsoon δ¹⁸O. Secondary predictors including moisture transport, continental recycling, and source temperature exert regionally divergent and time-varying influences. Supplementary large-scale climate indices can improve model performance for specific monsoon domains. El Niño und die Southern Oscillation (ENSO) is the most widespread teleconnection affecting most monsoon domains, while Atlantic Meridional Overturning Circulation (AMOC) and South Atlantic Convergence Zone (SACZ) act as additional key controls over Southwestern South American Monsoon. Spatial correlation analyses further demonstrate that the precipitation amount effect is weak and spatially heterogeneous at local grid scales but becomes robust and uniform at the regional scale. This further indicates that individual speleothem records may misrepresent large-scale monsoon signals due to local spatial noise, while regional proxy ensembles effectively capture the dominant precipitation-isotope relationship.

Received: 01 Jul 2026 – Discussion started: 28 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Xiaoxu Shi, Xiya Liu, Jiping Liu, Martin Werner, Hu Yang, Francisco W. Cruz, Chaoyuan Yang, and Gerrit Lohmann

Status: open (until 29 Oct 2026)

Post a comment Subscribe to comment alert

CC1:
'Comment on egusphere-2026-3894', Giacomo Medici, 11 Aug 2026 reply

General comments

Very good research on paleoclimate and stable isotopes. The specific comments can improve the manuscript.

Specific comments

Lines 18-20. Insert more recent literature on the use of oxygen (δ 18O) or hydrogen (δD) for hydrological and paleo-hydrological studies.
- Medici, G., Marianelli, D., Cornacchia, I., Gori, F., Brandano, M. 2026. Multi-disciplinary approach to paleokarst occurrence in the Eocene–Oligocene succession of the Apulia Carbonate Platform (Salento, Italy). Facies, 72(2), 17.
- Barbieri, M. 2019. Isotopes in hydrology and hydrogeology. Water, 11(2), 291.
Line 72. The general goal of your research is clear. But, what about the specific objectives? You can describe the 3 to 4 specific objectives of your research by using numbers (e.g., i, ii, and iii) at the end of your introduction.
Line 87. Provide more detail on the boundary conditions. You should at least describe the types.
Line 87. Provide reference for the types of boundary conditions of your model.
Line 104. Table 1 needs a link with a map?
Line 233. Some issues with the significant figures after the comma. Please, fit it.
Line 276. Here, I can see a paragraph with references and new concepts. I would keep it as a discussion creating a separate conclusion.

Figures and tables
Table 1. I would couple the table with a figure that shows the study regions.
Figure 3b. Add the associated equations.
Figure 3. Make the legends larger and readable.
Figure 4. Same here, make the legends larger and readable.
Figure 9. Enlarge font size for the content of the two axes.

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3894-CC1
- EC1: 'Reply on CC1', Julien Emile-Geay, 21 Oct 2026 reply
  
  Dear Dr Medici,
  Thank you for offering some constructive feedback on the manuscript, which the authors can choose to incorporate into their revision. However, I find the suggestion to cite your own work (as well as that of a senior colleague of yours at La Sapienza) highly problematic. Without an explanation as to why either of those citations is essential to this manuscript, this reads as a naked attempt to bolster your citation statistics or those of a collaborator, neither of which the intent of a public review and comment process like this one. Please provide a well-argued justification or drop this suggestion altogether. If I am not convinced, I will advise the authors to ignore this suggestion.
  Best regards,
  Julien Emile-Geay
  
  Reply
  
  Citation: https://doi.org/10.5194/egusphere-2026-3894-EC1
RC1: 'Comment on egusphere-2026-3894', Anonymous Referee #1, 29 Sep 2026 reply

Shi et al. present a new transient simulation of the Holocene using the isotope-enabled AWI-ESM2-wiso model. The simulation successfully captures some of the large-scale features associated with orbital and meltwater forcing, such as meridional displacements of the ITCZ and the bipolar seesaw. The authors find generally good agreement with proxy records and conduct a multiple linear regression analysis to determine the principal drivers of δ¹⁸O variability in several core monsoon regions.
Overall, this is an exciting and important new data product for the paleoclimate research community, particularly for understanding the climate drivers of water isotopes. Applying the revisions suggested below would help bolster this manuscript.

General comments:

1. How does this simulation compare with the iTraCE Holocene run (Bao et al., 2025)? That is also a transient, isotope-enabled run featuring orbital, GHG, ice, and meltwater forcing. Are there any notable differences in how these models simulate the effects of these forcings during the Holocene? Is one of the models better correlated with observations?
2. There should be a mention of the Green Sahara, particularly because of the dynamic vegetation turned on in the model. This seems to be a driver of some of the largest simulated anomalies in the Holocene trend maps (Figure 5).
3. It is worth discussing what could account for the remaining variance after applying the predictors to the MLR framework. For instance, there is still ~50% unexplained variance in the SW-SAM region after accounting for forcings and internal modes. And the variance captured by the pooled regression is just a bit higher than that (63%).

Specific comments:
Line 104: Where are these source regions and how were they determined?
Line 107: There is too little detail in this section about the multi-linear regression method, the authors should further elaborate. What regression and fitting methods are employed? How is significance determined? In other words, for those that want to reproduce these results, what do they need to know?
Line 111: Please clarify the amount-weighting scheme. Shouldn't this be monthly precipitation rates and δ¹⁸O that you are summing and then weighting? The result of this equation is the seasonal precipitation rate, but what is going into it is monthly, no? Unless I am misunderstanding.
Lines 240-241: Why do you think these regional coefficient magnitudes are nonstationary?
Lines 345–346: A mechanism is not described here, but rather an observation about the spatial-scale dependence of the amount effect. The authors may wish to replace “This mechanism” with more precise wording.

Table 1: To help the reader, mention in the caption that a map with boxes representing the monsoon study regions is shown in Figure 6.
Figure 2: How do these curves compare with the actual Marcott et al. 2013 reconstruction?
Figure 4: It would be useful to see the seasonal precipitation responses here, especially with the authors’ focus on the monsoon regions.
Figure 6: To assist the reader, mention in the caption the difference between the solid and dashed lines of best fit.
Figure 7: “Observed” as in proxy records? Or is this also simulated data, but not MLR-predicted? It looks like it is all model data.
Figure 10: I don't see the point of restricting the figure to the monsoon regions in 10a-b. But I understand why it is necessary in 10c to keep things restricted to the regions of interest. The authors should consider replotting the data for the whole tropics in a-b, and still have the monsoon region boxes overlaid.

Reference:
Bao, Y., Liu, Z., Thompson, L. G., Mosley-Thompson, E., Wan, L., & Lu, J. (2025). Climate simulations and ice core data highlight the Holocene conundrum over tropical mountains. Communications Earth & Environment, 6(1), 205. https://doi.org/10.1038/s43247-025-02188-2

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3894-RC1
RC2: 'Comment on egusphere-2026-3894', Anonymous Referee #2, 21 Oct 2026 reply

Review of the manuscript
„Transient simulation of Holocene climatic and isotopic variations across global monsoon regions“
by Xiaoxu Shi and colleagues
submitted for publication in Climate of the Past

General
The manuscript investigates a transient simulation of the Holocene based on the isotope enabled Earth System Model AWI-ESM2-wiso for isotopic variations of global monsoon regions. To investigate the importance of different meteorological variables and indices authors setup multiple linear regression models to predict d18o for the various monsoon regions. Results indicate that for some regions especially precipitation is important concerning the amount effect, whereas other regions indicate less clear influence of specific variables. Authors also point to the importance of local effects included in speleothems, the main basis for comparisons.
The manuscript is well written and the data and methods are described to a degree that the reader has necessary background information to follow the line of arguments. Also the model system authors use is state-of-the-art and the experiment setup is carefully designed and implemented. The model shows an overall good performance for model validation against observations and in representing general Holocene near-surface temperature trends for specific seasons and latitudinal bands.
As such, results should be published. Below I listed some points that should be addressed in the revised version, especially related to the setup and interpretation of the MLR models.
Specific
Abstract
The abstract is well written and comprehensively summarizes the main results of the manuscript. Authors might however also include some critical remarks on their conclusions concerning the MLR uncertainties, uncertainties involved in the proxies and maybe also the limits of interpreting results in the overall context on hydrological changes of global monsoons on Holocene time scales.
1 Introduction
l 18 ff: Authors start the intro with a very specific and technical section. I suggest to start-off with a more general paragraph on the monsoon and how it changed throughout the Holocene, based on previous modeling and proxy studies.
A second moment that might also be appreciated relates to the forward modeling of e.g. speleothems, taking into account proxy forward models (e.g. Dee et al. 2015; Hu et al., 2021) Authors could also use this framework as motivation and pre-requisite for their simulation activities using isotope-enabled simulations.
2 Methods
2.1 Model description
The authors should also state here some reference simulations and maybe add a section on the general performance of the model system when compared to observations, specifically for the isotope enabled output for regions where observations are available. This would be helpful to assess the quality of the model and maybe also would give rise to explain some limitations of the modeling approach, especially when interpreting results on Holocene time scales. In addition, a hint to the model validation chapter in 3.1 can also help to address this point.
2.2 Experimental design
Knowing that isotope enabled simulations are very expensive in terms of computing resources, I was wondering if authors could add a section on carrying out an ensemble of simulations and how this could influence conclusions with respect to analyzing only a single simulation based on a single model system.
2.3 Study regions
In addition to Table 1, authors should point to Fig. 6a where the geographical regions are depicted.
2.4 Multi-linear regression framework
l. 115: Using wind_dir as predictor in a monsoonal context is for sure a valid idea. I think using a higher atmospheric level, e.g. winds@850hPa could better represent large-scale conditions of monsoonal systems. At least authors might give some explaining words why they chose this predictor for their subsequent MLR setup.
l. 125: why did the authors proceed with binned 100 year mean values, rather than inter-annual values ?
ll 128 ff: It seems that the MLR uses the same set of predictors, irrespective of the target monsoon region. How about a pre-selection ? In addition, some of the indices might show a high degree of co-linearity. The authors try to address this by limiting the number of additional indices to three. However, also those three indices can show a high amount of co-linearity, weakening the robustness of the according MLR. A cross-correlation matrix among the different predictors prior to the setup of the MLR could help to elucidate which predictors are highly correlated and might not be used at the same time for a specific target region. For example, the AO and SAM will most likely show a high correlation with wind regimes over the Northern and Southern Hemisphere, respectively, and also with other initial predictors related to precipitation and moisture flux.
3 Results
3.1 Model validation
l.169 Here authors should mention complications when comparing d18O from speleothems with d18O from precipitation in the context of forward modeling. For instance, which kind of factors control the amount of d18O that is finally recorded into the speleothem compared to the one that is included in the d18O of precipitation ?. (e.g. due to soil and other depletion/enrichment processes). Are the processes similar or different for the various regions and proxy archives in terms of speleothems. In addition, also some words concerning the time scale of robustly comparing results between isotope enabled models and speleothems would be helpful (i.e. the skillful time scale).
3.2 The 8.2 ka event
For the 8.2 ka event the routing of rivers and the according pathway of the fresh water masses seems to be critical (c.f. also Younger Dryas event). Is the resolution of the FESOM model high enough to allow for a realistic pathway along the coast of North America (southward displacement due to Labrador current) or does the fresh water directly enter the core regions of the NADW production sites, leading to changes in the AMOC and subsequent consequences ?
3.3 Holocene climate trend
Are the trends linear ? Maybe the authors could also present a Figure representing trends based on different time chunks and/or running trends for the different latitudinal bands for different temporal filters.
3.4 Multiple linear regression analysis
3.4.1 Pooled regression and the dominance of the amount effect
l. 217: Are the four predictors statistically independent? Again, a cross-correlation matrix among the predictors will help to elucidate the amount of co-linearity and might help to construct more robust MLR models for the different monsoon regions (and might also help to explain some of the effects described in the performance and regression slopes of the different MLR setups). The (near-) independence of predictors used for the MLR is essential for a physically robust interpretation of the according regression slopes/coefficients.
l. 233: How should the model reproduce short-lived oscillations when the MLR was constructed based on 100 year mean values ?
3.4.2 Contribution from climate indices
The inclusion of additional predictors should be taken out with care in order to not artificially increase the performance metrics due to over-fitting because of the presence of co-linear predictors.
4 Discussion and conclusions
The discussion and conclusions are well structured, summarize the main results and also put new results in context of current literature and hypotheses. I suggest however to add a dedicated paragraph on the limitations of the model system, the difficulties interpreting and comparing d18O based on modeled precipitation vs. d18o based on speleothem proxies.
In addition also some critical words on the physical interpretation of the individual predictors in the MLR based models should be mentioned, including a discussion on the effect of using co-linear predictors for the MLR models.
Figure and Tables:
Fig 1: The Figures show the absolute values for the last 500 years of the simulation. As such it is difficult to see any model bias compared to observations. Maybe authors could provide difference maps using a gridded re-analysis product for the pre-industrial era (e.g Franke et al., 2017; Steiger et al., 2018).
Fig. 4: The areas showing stat. significant differences (i.e. alpha =5%) should be marked.
Fig. 5: The stippling is hardly visible.
Fig. 7: Please include a dashed line at „0“ for better reference in all subplots.
References:
Dee, S., Emile-Geay, J., Evans, M.N. et al. (2015): PRYSM: An open-source framework for PRoxY System Modeling, with applications to oxygen-isotope systems, J. Adv. Model. Earth Syst., 7, 1220–1247, doi:10.1002/2015MS000447.
Franke, J., Brönnimann, S., Bhend, J. et al. (2017): A monthly global paleo-reanalysis of the atmosphere from 1600 to 2005 for studying past climatic variations. Sci Data 4, 170076, https://doi.org/10.1038/sdata.2017.76.
Hu, J., Dee, S. G., Corinne I. Wong, C.I. et al. (2021): Assessing proxy system models of cave dripwater δ18O variability, Quaternary Science Reviews, 254, https://doi.org/10.1016/j.quascirev.2021.106799 Steiger, N., Smerdon, J., Cook, E. et al. (2018): A reconstruction of global hydroclimate and dynamical variables over the Common Era. Sci Data 5, 180086. https://doi.org/10.1038/sdata.2018.86.
Steiger, N., Smerdon, J., Cook, E. et al. (2018): A reconstruction of global hydroclimate and dynamical variables over the Common Era. Sci Data 5, 180086. https://doi.org/10.1038/sdata.2018.86.

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3894-RC2

Xiaoxu Shi, Xiya Liu, Jiping Liu, Martin Werner, Hu Yang, Francisco W. Cruz, Chaoyuan Yang, and Gerrit Lohmann

Data sets

AWI-ESM-2.1-wiso Holocene transient simulation: seasonal-mean output for MLR analysis of monsoon δ¹⁸O Xiaoxu Shi https://zenodo.org/records/21068859

Xiaoxu Shi, Xiya Liu, Jiping Liu, Martin Werner, Hu Yang, Francisco W. Cruz, Chaoyuan Yang, and Gerrit Lohmann

Viewed

Total article views: 471 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
289	143	39	471	27	31

HTML: 289
PDF: 143
XML: 39
Total: 471
BibTeX: 27
EndNote: 31

Views and downloads (calculated since 28 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	95	34	7	136
Aug 2026	56	36	11	103
Sep 2026	106	56	12	174
Oct 2026	32	17	9	58

Cumulative views and downloads (calculated since 28 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	95	34	7	136
Aug 2026	56	36	11	103
Sep 2026	106	56	12	174
Oct 2026	32	17	9	58

Viewed (geographical distribution)

Total article views: 439 (including HTML, PDF, and XML) Thereof 439 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 21 Oct 2026

Short summary

Cave stalagmites record the oxygen isotope of past rainfall and are used to reconstruct how monsoons changed, but due to the complexity of isotope impacting factors, what these records tell us is debated. We performed a 8300-year simulation with an isotope-enabled climate model, and show that rainfall amount is the key stable driver across global monsoons, while temperature and moisture pathways vary with region and time. Single-site records are noisy, but regional averages recover the signal.


Total:	0
HTML:	0
PDF:	0
XML:	0