the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Apr 2026
Connecting volcanic climate impacts to famine in China using the REACHES database
Abstract. Volcanic eruptions have been linked to historical famines in many parts of the world. In China, reduced temperatures following major eruptions can destabilise the hydroclimate and agricultural production, contributing to subsistence crises and even the downfall of dynasties. This study provides the first long term analysis of the specific connection between volcanic activity and famine in eastern China from 1440 to 1900 CE. Using the REACHES historical climate database, it reconstructs indices measuring temperature, drought, flooding, crop failure and famine. Superposed epoch analysis of these indices reveals a recurring, though regionally distinct, association between eruptions and famine. Famine peaks occur in northern China in the year of an eruption, in central China one to three years later – coinciding with delayed drought and crop failure – and in southern China in the first post-eruption year. Correlation analysis indicates statistically significant relationships between volcanic forcing, hydroclimatic extremes, crop failure and famine. Case studies – including a new assessment of the impacts of the 1809 “unknown” eruption – demonstrate how other factors, such as the El Niño Southern Oscillation, non-volcanic climate processes, price volatility, disease and state relief mediate volcanic impacts. Many of these factors form feedback loop than can delay, amplify or counteract volcanic effects. We conclude that while eruptions may not be the primary drivers of famine in China, they significantly increase the risk of drought, flood, harvest failure and subsequent subsistence crises. The findings demonstrate the capacity of major volcanic events to destabilise food systems through coupled climatic and societal pathways.
- Preprint
(10114 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1228', Anonymous Referee #1, 06 Jul 2026
-
AC1: 'Reply on RC1', Richard Warren, 07 Aug 2026
Thanks to both reviewers for their detailed comments. Responses to general and individual points can be found below in the following format:
Comment
Response
Reviewer 1
The manuscript provides a comprehensive analysis of the link between volcanic eruptions and famine in eastern China during 1440–1900 CE using the REACHES data. The author argues that volcanic activity does not necessarily cause famine, although some of their analyses suggest it could increase the likelihood of droughts and floods, thereby increasing associated agricultural and societal risks, including crop failure. Overall, this study is a valuable contribution to understanding volcano–climate–society relationships. However, I recommend clarifying some aspects of methodology and including additional analyses and discussion. I would also like to note that my background is in climate dynamics, so my review of the context and discussion on agricultural and societal impacts may be limited.
Major comments
- In Section 2.3, it is mentioned that 23 provinces were categorized into three regions based on “climatology, agriculture, and historically concurrent famines.” Defining regions by climatology and agriculture sounds reasonable, but could the inclusion of historical famines systematically bias the assessment of the link between volcanic climate impacts and famines? Do the results vary widely depending on how these three regions are defined?
This is a good question – and I have now done sone quick adjustments with the data to see how adding/removing some of the borderline provinces (i.e those that could be classified as North/Central or South/Central) changes the results – focusing on the superposed epoch analysis. It seems that while the magnitude of SEA results does change (as would be expected), the general pattern of results does not. I know from my original analysis (not included in this paper), which further subdivided the regions between east and west, that results do vary within the selected regions. However the east/west differences were less pronounced than those between the North, Central and South. I therefore chose to use only three regions to prevent the SEA and correlation results from becoming too complex. I am aware this method comes with its own limitations (removing some of the geographical nuances of impacts within the regions) but felt this was a reasonable trade off since the three regions capture the most important general trends. I am also aware that a more granular province by province study will be coming out in the near future that will allow comparison with these regional results.
The famine bias is indeed possible – the regions were partly based on the work by Lin et. al 2020 that identified the co-occurrence of famine in certain regions over multi-decadal periods. The reasoning behind the use of this data was that if famines tends to reoccur over a given area (say the lower Yellow river), I didn’t want to split this area between two regions of analysis and thereby weaken the famine signal. This may have biased the results in the opposite direction – enhancing the famine signals - but given that the famine regions broadly overlapped with regions based purely on climatology and agriculture I felt this was a reasonable trade off to make.
- The analysis of the impact of volcanic eruptions on climate is based solely on the Drought Index from the REACHES database. However, hydroclimatic paleo records are available for this region (e.g., the Great Eurasian Drought Atlas by Cook et al. 2024). I think it would be beneficial to evaluate whether the REACHES-based Drought Index is consistent with other hydroclimatic reconstructions.
Agreed. This would be informative – I will include a comparison between the REACHES drought index and the PDSI measurements in the Great Eurasian Drought Atlas and possibly the precipitation data from the Modern Era Reanalysis (ModE-RA, Valler et al. 2024) as an appendix. I will also consider rerunning the SEA and correlation analysis with the new data if there is a considerable difference in the reconstructions.
- In Lines 161–161, the author mentioned that the summer temperature index is defined as the average of the summer temperature reconstruction from Wang et al. (2024) across each study region, because the reconstruction covers smaller regions than this study. How small was the coverage in Wang et al. (2024) compared to that in this study? I also do not quite understand exactly how the author was able to average across regions when data from some areas are missing.
This does perhaps require further explanation, which I will add to the main text. Wang et al. created temperature indices for 3 regions and 15 subregions of eastern China (see Supplement, page 1). The Wang et al. regions B, C and D broadly correspond to the regions used in this study, but I had to adapt them to follow provincial boundaries (which were historically important because of how they were administered). So, for example, the data from C15 had to be averaged with that from D18 and D17. The Wang et al. subregional indexes are calculated as the mean of all the digitized climate records in a given subregion, so to average three subregions, I calculated the mean of the three subregions, weighted by the number of records in each region.
This gave a representative temperature index for each study region – but, following your previous point, I do feel it would be useful to include an alternative temperature series, such as ModE-RA, in the appendices for comparison.
- This manuscript contributes to an analysis of the Unidentified 1809 event as a case study, but for the other three events, the author relies on previous studies. To demonstrate that the methodology employed in this study would yield results consistent with previous studies, I think the author should repeat the analysis used for the 1809 event (e.g., Fig. 6) for the other events.
Agreed. While I do not think these results should be included in the main text (since they are at least partially based on the same historical data as the presented studies), I am happy to include them as an appendix and discuss the comparison in the main text.
Minor comments
- The introduction could benefit from reorganization. I think adding an explanation of why the author focuses on China and why the period 1440 to 1900 CE is selected would strengthen the context. Additionally, clarifying the type of climate variability described in Lines 49–60 is important, as it is somewhat ambiguous (ENSO is mentioned in the paragraph before but not in Lines 49–60). ENSO and its influence on the climate in China should also be explained in detail in the Introduction, since a decent amount of analysis and discussion is devoted to it in later sections. Furthermore, I recommend adding a paragraph that briefly describes the rest of the manuscript (or the general context of this manuscript) at the end of the Introduction.
Accepted.
- The VICES approach is introduced in Section 2.1, but it is still unclear to me how it is incorporated into the actual analyses. Does it only come into play in Table 5 and Figure 7, or is the inclusion of climatic, environmental, and societal indices (i.e., drought, crop failure, and famine indices) in all analyses considered a part of this VICES approach? I think clarifying this would greatly improve the manuscript and highlight the importance of this study.
This is a helpful point – I will update section 2.1 accordingly.
The aim of the VICES approach is to break down the link between eruptions and human society into its components: volcano-climate, climate-environment (agriculture in this case), climate-society and environment-society and examine each link individually. The coincidence and correlation between measured variables are examined to show the relative importance of these different links (as is done in sections 3.2 and 3.3). These results then need to be explained by examining causal pathways on a case by case basis (section 3.4). For example, to understand why drought correlates with crop failure it is necessary to understand both the effects of both locusts and lack of water.
- I do not think you need to refer to figures/tables that include analysis results in the Data and Methodology section. Also, please make sure that the font size in figures is large enough when the manuscript is printed at a standard A4 size (e.g., Fig. 2) and number the figures and tables in the order they are mentioned in the text.
Accepted.
- I understand this manuscript is based on the REACHES database, but did the other studies mentioned in the text also use a similar index to the Event Index defined by Eq. (1)? This is my first time reviewing a paper that bases analyses on reports rather than numerical observations, and I am curious how common this approach is and whether the results could be largely influenced by how famine and drought are defined.
Unfortunately, such weighted event indices are only sporadically used. Many studies simply used the number or severity of historical records. This can be appropriate if effort is made to ensure consistency in the number of records obtained from each time period/region or the time period/region is very small/short, but in all honesty is problematic when this is not the case. Hence the use of a weighted index for this study, rather than the absolute number of records.
We are actually quite fortunate in the Chinese case when it comes to defining famine and drought. The imperial authorities that recorded and compiled the records behind REACHES had a very clear and consistent definition of drought and famine that was applied across the country throughout this time period (consistency of recording was vital to the administration of agriculture and famine relief systems). We can therefore take all records of a “famine” or “drought” as broadly equivalent, with the caveat that a small minority of reports do use different descriptions or include more detail, as reflected in the magnitude and duration descriptions in REACHES. However, such reports make up a negligible minority of the total in any given category (e.g. drought).
- The author discusses the link between ENSO and famine in China, but some of the links mentioned precede Lag Year 0 (e.g., Line 211). Are such links commonly considered more than a coincidence in studies on famine or other agricultural/environmental contexts?
Most studies would consider it coincidence – and to be fair only significant famine SEA result only just meets the significance criteria. However, the majority of studies only examine the period including and following an EL Niño/La Niña year.
It may be that a result showing an effect preceding a major El Niño may instead be capturing a lagged effect of the preceding La Niña (and vice versa), although a comparison of the El Niño/La Niña results from this study does not support this. There is also some evidence for the effect of a developing El Niño on Chinese agriculture (Li et al. 2020), which this result could be capturing, but one would not really expect such a swift impact on famine. The Indian summer monsoon is known to be suppressed in the summer preceding peak winter El Niño conditions (Park et al. 2010), but whether this is also the case with the East Asian Monsoon is not clear.
In all honesty I think that the variation in the SEA results (particularly when we look at the different ENSO reconstructions) means that the SEA method has little to tell us about the impact of extreme ENSO years in China, other than to confirm what other studies have shown, which is that the impacts are quite variable, but can modulate volcanic impacts. I will now revise the text to make sure that this is explicit in the discussion and the conclusion.
Line comments
Line 10: I recommend including a brief description of the REACHES database.
Accepted.
Lines 16–17: What are considered “non-volcanic climate processes” here?
This would primarily refer to stochastic climate variability and other modes of climate variation (such as the Atlantic Multidecadal Oscillation) not currently linked to volcanic activity. I’ll make sure this is noted explicitly in the text.
Lines 25–26: What periods exactly are meant by “early Chinese history” and “later periods”? I think it is crucial to state specific periods since later in this paragraph it states “a persistent connection between volcanic activity and famine has not been systematically investigated,” which implies a comprehensive analysis, and it is probably better to make sure those “early” and “later” periods are included in this study’s focus 1440–1900 CE.
In this case early Chinese history refers to the Western Han dynasty, 206 BCE to 8CE. The “later periods” are those that come within this studies time period of 1440 to 1900 CE. This is an important point and I will revise the text to make this explicit.
Line 30: It may be worth briefly defining the difference between “climatic” and “environmental” pathways. It is clear once you introduce the VICES approach in Section 2.1, but not here yet.
Given that I already introduce the simple causal pathway from volcanic sulphate aerosols to global temperatures and regional rainfall patterns to agriculture and food prices at the start of this paragraph, I think I would prefer to leave this for the VICES approach section.
Line 49: Again, I recommend including a brief description of the REACHES database if you are mentioning it before the Data and Methodology section. Is REACHES an abbreviation?
Agreed, a brief introduction to the REACHES database (Reconstructed East Asian Climate Historical Encoded Series) would be useful here.
Line 67: Can you clarify what is meant by “perception and meaning” here?
In this case, it refers to how the Chinese people perceived events such as flood and drought (were they random, were they a punishment from god etc.) and how they ascribed meaning to them (did it mean that the ruler had lost the mandate of heaven and that they therefore no longer needed to follow his commands, for example). I will revise the text to make this more explicit.
Line 89: I recommend referring to figures and tables in parentheses instead of a dash [e.g., (see Fig. 1) instead of - see Fig. 1].
Accepted.
Line 96: Did you mean Table 1?
In this case I was referring to the scatterplots in Figs. 3b, 3d and 3f that show the coincidence between major eruptions and unusually high famine indices in all 3 regions.
Lines 102–103: How were these four eruptions chosen as case studies? What is special about these events? I also want to point out that mentioning these events before explaining how major eruptions were chosen is a little bit confusing.
The 1641, 1793 and 1815 eruptions were chosen as case studies because they were the best studied eruptions (within the 1440-1900CE time period) at time of writing. Each was the subject of a detailed study addressing connections between the volcanic eruption, its climate impact and subsequent effects on Chinese society. The 1809 eruption was selected as the fourth case study since it was the largest (as measured by SAOD) eruption in the time period that had not been the subject of such a study.
Thankyou for pointing this out – I will make sure it is explicit in the final text and agree that these should be introduced at the end of the choosing major eruptions section to prevent confusion.
Line 117: What period is used to define the 95th percentile of SST anomalies? Is it the same as the study period of 1440–1900 CE?
It is. I will revise the text to make this explicit.
Line 131: What was considered “a sufficiently large number” of reports?
In this case, the basic rule followed was that provinces selected had reports throughout the time period 1440 to 1900 CE and that there wasn’t any large gaps in reporting (i.e. several consecutive years with no reports). This resulted in regions that broadly followed those used in the Wang et al. 2024 and Lin et al. 2020 studies. I agree this is not clear from the text and will revise it to make it explicit.
Lines 161–169: I assume the author computed post-eruption anomalies relative to some pre-eruption baseline period to perform Superposed epoch analysis. What was the baseline period used in this study? Also, were volcanic years included in the original data used in the Monte Carlo model test?
In this case the baseline is a random sample of 10,000 years taken from 1445 to 1895 CE (5 years is left either side to allow for the SEA lag years). The results for volcanic years are then shown not as an anomaly but as the (mean) absolute values of the index for the 5 years before and after an eruption. These are visually compared in the plots to the mean and percentiles of the random sample.
Volcanic years are therefore included in the Monte Carlo test. Rerunning the SEA excluding volcanic years from the sample gives very similar results, but may be more appropriate for the final manuscript since it enhances the volcanic signal relative to the background sample.
Lines 176–181: Are the case study analyses also based on anomalies relative to some reference period, or on absolute index values?
These are absolute index values. This was partly to allow comparison with similar analysis in the Lin et al. 2020 and Gao et al. 2017 studies, but also to enhance the focus on using the plots to establish links between the climate, environmental and social phenomena, rather than connection to the eruption.
Figure 3: I recommend picking a different color for either the non-volcanic year or Yr 0. Black and grey dots are hard to distinguish, unfortunately.
Accepted.
Figures 4–5: It may be beneficial to include individual events as thinner lines in these Superposed epoch analysis figures.
I did consider this, but the problem is that some single events have very large index values, and so the observation means line is smoothed out – making it harder to see the main result. SEAs showing the individual events could certainly be included as an appendix though.
Table 4: It is just a suggestion, but it may be worth using a heatmap and hatching the insignificant correlations, rather than simply omitting the insignificant coefficient values.
I would prefer to retain the current format as unfortunately in a heatmap format the r=1 correlations between the same variables would dominate, unless omitted.
Figure 6: A more detailed explanation would be appreciated in the caption. For instance, what are the top and bottom rows in each panel? It may also be worth putting lag years next to actual years in CE [e.g., 1809 (Lag Year 0)].
Agreed. I will change the titles and include a note that the top and bottom panels show drought and climate/environmental events, and famine and social events.
Lines 393–394: I would be curious what the results look like if only tropical eruptions are considered. Since there are still 10 events, I think the analysis can be performed reasonably well.
This is an interesting point – I have run a quick assessment of the famine index with the tropical and N Hemisphere eruptions and it does show some interesting differences (see Supplement, page 2).
It seems that the N Hemisphere eruptions have a delayed impact compared to the tropical eruptions and in central China there are considerably more instances of famine. This may be a random artefact of the small sample, but this does agree with the literature that suggest that N Hemisphere eruptions might have a greater impact than tropical.
I will include these graphs as an appendix (along with the SEAs for temperature, flood, drought and crop failure) and include a brief discussion and comparison with the literature in the main text.
Line 341: How are the volcanic impacts on climate (both summer temperature and drought) discussed here consistent with other previous studies that looked at the last millennium? There seems to be some post-eruption drying in northern China and wetting in southern China, according to Tejedor et al. (2021, PNAS), which studied 1000–1850. Could the lack of hydroclimatic impact in southern China mentioned in Line 370 be due to the selection of events?
The temperature results seem to be broadly comparable with previous studies. We see the expected cooling in northern and central eastern China in years 0 and 1 following the eruptions. The lack of a temperature response in southern China may be an artefact of the lack of variation in the summer temperature index for southern China, but a recent review (Sun et al. 2024) found that south eastern China does seen to have a much smaller temperature response to tropical eruptions than the rest of China.
The hydroclimate responses are more at odds with other studies. The increased incidence of drought in northern China in years 1 to 4 following an eruption matches that seen in Tejedor et al. and Sun et al., but central China also experiences increased drought, when most studies show it getting wetter following an eruption. Both this and the lack of a hydroclimate response in southern China could, as you say, be due to the selection of events. I’m hoping the new SEA results (see previous comment) may help to find out if this is the case, since most studies focus only on tropical or N hemisphere eruptions. The lack of agreement could also be due to the fact that hydroclimate responses to eruptions vary considerably based on the reconstruction method (see Sun er al., Figure 6). This is where I feel comparing historical records with historical records may actually have a slight advantage over other reconstruction methods which may add more noise by combining inputs from different sources (i.e. tree ring data and climate models).
Lines 375–389: I enjoyed this paragraph, especially I appreciated the explanation of the different crop types and their sensitivity to climate conditions.
Lines 403–404: The dash and en dash are used inconsistently throughout the text. Please make sure to use a consistent notation!
Noted. This will be corrected.
-
AC1: 'Reply on RC1', Richard Warren, 07 Aug 2026
-
RC2: 'Comment on egusphere-2026-1228', Anonymous Referee #2, 21 Jul 2026
This study addresses an important research question: how major volcanic eruptions may be linked to famine through climate-driven impacts on agricultural production. The topic of volcanic impacts on climate has received considerable attention in recent years, leading to a rapidly growing body of literature examining the severe societal consequences after major volcanic eruptions, including dynastic collapse and social turmoil. A common pitfall in this line of research is the lack of socio-environmental data at appropriate spatial and temporal resolutions, which hinders a thorough examination and often leads to an oversimplification of human societal mechanisms. In this context, I do see the scientific value and potential in this manuscript to help bridge these research gaps. By saying so, I also have several major and minor comments regarding conceptual framing, statistical methodology, and data interpretation that I hope to help the author strengthen the scientific rigidity of the manuscript.
Conceptual rationality and causal claims:
The study adopts the VICES framework to guide its conceptual and methodological design. While this framework acknowledges feedback loops, it remains relatively linear and leans heavily on a top-down approach. Although the VICES framework appears practical, the statistical methods employed—namely correlation analysis and Superposed Epoch Analysis (SEA)—can only establish statistical associations or temporal alignments; they cannot demonstrate direct causality. Furthermore, the spatial distribution maps of climatic, environmental, and societal events (Figure 6) before and after the 1809 eruption indicate that events occur in clusters, suggesting a spatial evolution over time. For example, drought began in North China in 1810, intensified in 1811, and eased slightly in 1812; meanwhile, famine took place in 1811 and peaked in 1812. While this spatiotemporal analysis likely suggests a spatial correlation and a lagged effect (with famine peaking one year after drought), it does not provide empirical proof of a direct causal relationship among these events. Therefore, I strongly suggest the author to reframe the terminology used in the study and tone down the causal claims that can better reflect the level of analysis in the research.Data description:
The abundance of Chinese historical documentary records provides a rich material for studying past climate variability and human responses. Utilizing the REACHES database is a practical approach, and showcasing its utility is valuable for the broader research community. However, because a big portion of readers may be unfamiliar with REACHES database, the manuscript would benefit from a more detailed description of how the records were retrieved in this study, the strength of using the REACHES records, and the potential limitations observed during the research process.Results and proposed causal framework:
I appreciate the transparency and clarity of the results presented in the study and in the Appendices. For instance, Figure 3 clearly illustrates that a large proportion of volcanic eruption years coincided with extreme climate events, which is really interesting. At the same time, there are many extreme event years without volcanic eruptions. This is both expected and important, as volcanic forcing is only one of the external drivers of climate variability as noted in Figure 7. However, figure 7 requires substantial clarification and refinement. For example, what is the non-volcanic climate variability? Does non-volcanic climate variability only influence flood/drought instead of temperature and other environmental processes? Why and how ENSO leads to reduced temperature but not warming temperature? What does temperature mean here, i.e., land surface temperature? Also, soil moisture/cloud feedback can be affected by volcanic eruptions and non-volcanic climate factors. In other words, this is an ambitious figure, aiming to produce causal pathways that are beyond what has been discovered in this study. Moreover, many arrows in the figure and their directionality are still understudied and some of them remain scientific debate such as ENSO’s influence on East Asian monsoon and so on. Therefore, I strongly suggest the author to reframe the figure and focus on the key findings of this study, rather than attempting an all-encompassing conceptual model.ENSO interpretation and statistical methods:
I found ENSO part in this study is confusing. First of all, the SEA statistical results in Appendixes are not convincing and the high/low index classifications in Table 3 are difficult to interpret (how do you define high and low?). Then by looking at the El Nino and La Nina years used in this study (Table 2), it is soon clear to me that many El Nino and La Nina years have short time intervals such as the El Nino years in 1606, 1609, and in 1800, 1804, and then in 1828, 1833, and 1839 and so on. This means that when applying an 11-year window for SEA, these events inevitably overlap across preceding and succeeding years (i.e., 1608 is the succeeding year of 1606 event but a preceding year for 1609 event), introducing significant noise and confounding the signal. Therefore, if the author wishes to retain the ENSO analysis, I strongly recommend applying alternative statistical methods, utilizing narrower time windows, or applying appropriate filtering techniques. Moreover, it is important to clarity the purpose of introducing ENSO in this research. For example, if ENSO is treated as an intermediate factor, you may consider analyzing how many volcanic eruptions co-occurred with ENSO events versus non-ENSO years, and compare their relative environmental and social impacts. Overall, a clear identification of the role of ENSO in this study would help sharpen the analysis.Lagged correlation analysis:
The authors tested lagged correlations between the SAOD volcanic forcing index and various environmental/societal indices, finding none or very weak relationships. This actually adds little value to the main narrative and does not match other results in Figure 5 and 6. I suggest to remove it, otherwise the author has to improve the way of assessment. Lagged correlation is a good approach to assess lagged effect in time however it can also be sensitive depending on the temporal scale (i.e., seasonality) and the parameter in question. By reviewing other figures, I feel the lagged effects for crop failure and famine may associate with severe climate events (i.e., drought) but not volcanic eruption itself. So, there is something for author to reconsider.Minor comment:
The manuscript presents a writing style easy to understand and follow. However, several typos need to be corrected, and certain sections would benefit from a more formal academic writing style. Below I list a few examples:
Line 17-18, ‘Many of these factors form feedback loop than can delay, amplify or counteract volcanic effects’
Line 262, 'r value indicate at best very weak correlations’
Line 319, ‘breading conditions for locusts’
Line 354, ‘eastern China might partly be due variable hydroclimatic impacts’Citation: https://doi.org/10.5194/egusphere-2026-1228-RC2 -
AC2: 'Reply on RC2', Richard Warren, 07 Aug 2026
Thanks to both reviewers for their detailed comments. Responses to general and individual points can be found below in the following format:
Comment
Response
Reviewer 2
This study addresses an important research question: how major volcanic eruptions may be linked to famine through climate-driven impacts on agricultural production. The topic of volcanic impacts on climate has received considerable attention in recent years, leading to a rapidly growing body of literature examining the severe societal consequences after major volcanic eruptions, including dynastic collapse and social turmoil. A common pitfall in this line of research is the lack of socio-environmental data at appropriate spatial and temporal resolutions, which hinders a thorough examination and often leads to an oversimplification of human societal mechanisms. In this context, I do see the scientific value and potential in this manuscript to help bridge these research gaps. By saying so, I also have several major and minor comments regarding conceptual framing, statistical methodology, and data interpretation that I hope to help the author strengthen the scientific rigidity of the manuscript.
Conceptual rationality and causal claims:
The study adopts the VICES framework to guide its conceptual and methodological design. While this framework acknowledges feedback loops, it remains relatively linear and leans heavily on a top-down approach. Although the VICES framework appears practical, the statistical methods employed—namely correlation analysis and Superposed Epoch Analysis (SEA)—can only establish statistical associations or temporal alignments; they cannot demonstrate direct causality. Furthermore, the spatial distribution maps of climatic, environmental, and societal events (Figure 6) before and after the 1809 eruption indicate that events occur in clusters, suggesting a spatial evolution over time. For example, drought began in North China in 1810, intensified in 1811, and eased slightly in 1812; meanwhile, famine took place in 1811 and peaked in 1812. While this spatiotemporal analysis likely suggests a spatial correlation and a lagged effect (with famine peaking one year after drought), it does not provide empirical proof of a direct causal relationship among these events. Therefore, I strongly suggest the author to reframe the terminology used in the study and tone down the causal claims that can better reflect the level of analysis in the research.Accepted. You are correct that the spatial analysis is in effect another form of coincidence/correlation analysis and does not show direct causality (although it does break it down into more detail). This will be made explicit in the main text and the terminology adjusted accordingly. However, this study does draw on modelling and historical studies that can claim to provide more direct causal evidence. I agree however that where such evidence is applied and where it is not is not clear in the text. I will update the manuscript to make this explicit.
Data description:
The abundance of Chinese historical documentary records provides a rich material for studying past climate variability and human responses. Utilizing the REACHES database is a practical approach, and showcasing its utility is valuable for the broader research community. However, because a big portion of readers may be unfamiliar with REACHES database, the manuscript would benefit from a more detailed description of how the records were retrieved in this study, the strength of using the REACHES records, and the potential limitations observed during the research process.Agreed. Section 2.3 will be extended to include this information and the limitations included in the Discussion and Conclusion section.
Results and proposed causal framework:
I appreciate the transparency and clarity of the results presented in the study and in the Appendices. For instance, Figure 3 clearly illustrates that a large proportion of volcanic eruption years coincided with extreme climate events, which is really interesting. At the same time, there are many extreme event years without volcanic eruptions. This is both expected and important, as volcanic forcing is only one of the external drivers of climate variability as noted in Figure 7. However, figure 7 requires substantial clarification and refinement. For example, what is the non-volcanic climate variability? Does non-volcanic climate variability only influence flood/drought instead of temperature and other environmental processes? Why and how ENSO leads to reduced temperature but not warming temperature? What does temperature mean here, i.e., land surface temperature? Also, soil moisture/cloud feedback can be affected by volcanic eruptions and non-volcanic climate factors. In other words, this is an ambitious figure, aiming to produce causal pathways that are beyond what has been discovered in this study. Moreover, many arrows in the figure and their directionality are still understudied and some of them remain scientific debate such as ENSO’s influence on East Asian monsoon and so on. Therefore, I strongly suggest the author to reframe the figure and focus on the key findings of this study, rather than attempting an all-encompassing conceptual model.Thankyou for this feedback. The aim of Figure 7 is to show the probable causal pathways and feedbacks linking eruptions to famines in China and was not meant to be definitive or exhaustive. However, I note that this was not stated in the text and needs to be. I will update the text and reframe the figure accordingly.
The links were identified both from this study’s results and from the literature in the case studies and introduction. However, I entirely agree that some connections are understudied/still under debate and that this needs to be mentioned more explicitly in the text.
I will also address the issues you mention above by restructuring the diagram and adding explanation and clarifications to the accompanying text.
ENSO interpretation and statistical methods:
I found ENSO part in this study is confusing. First of all, the SEA statistical results in Appendixes are not convincing and the high/low index classifications in Table 3 are difficult to interpret (how do you define high and low?). Then by looking at the El Nino and La Nina years used in this study (Table 2), it is soon clear to me that many El Nino and La Nina years have short time intervals such as the El Nino years in 1606, 1609, and in 1800, 1804, and then in 1828, 1833, and 1839 and so on. This means that when applying an 11-year window for SEA, these events inevitably overlap across preceding and succeeding years (i.e., 1608 is the succeeding year of 1606 event but a preceding year for 1609 event), introducing significant noise and confounding the signal. Therefore, if the author wishes to retain the ENSO analysis, I strongly recommend applying alternative statistical methods, utilizing narrower time windows, or applying appropriate filtering techniques. Moreover, it is important to clarity the purpose of introducing ENSO in this research. For example, if ENSO is treated as an intermediate factor, you may consider analyzing how many volcanic eruptions co-occurred with ENSO events versus non-ENSO years, and compare their relative environmental and social impacts. Overall, a clear identification of the role of ENSO in this study would help sharpen the analysis.Accepted.
The aim of the ENSO part of this study was to disentangle the effects of El Ninos/La Ninas from the effects of volcanic eruptions, since they often overlap. Even with my small sample of extreme ENSO years, 6 out of the 13 eruptions overlap to some extent with a major El Nino/La Nina (i.e. one occurred within the 11 year SEA window). Since the sample of eruptions is already quite small, it did not make sense to further reduce it by excluding these. Hence it was important to account for whether ENSO impacts might affect the results and I therefore attempted to isolate the potential impacts of ENSO on my indices using the SEA method. However, as you point out, this was not made explicit in the methodology section, which will now be updated.
The SEA method was chosen to allow the results to be directly comparable with the volcanic eruption SEAs, however it does have limitations since it is very sensitive to changes in the dating of El Ninos/La Ninas. I suspect this may account for the somewhat confusing results in Table 3. As you say though, the sample of extreme ENSO years does need to be filtered to remove events with overlapping time windows, as well as those that overlap with the volcanic events. This will be done and the results, discussion and conclusions updated accordingly.
In this study I do not treat ENSO events as “caused” by volcanic eruptions, since this is still the subject of debate. Instead I treat it as another cause of “background” climate variability. As you say it would therefore be useful to compare the impacts of eruptions that co-occur with El Ninos/La Ninas with those that don’t. Unfortunately however, the sample size of co-occurring eruptions would only be two, not enough for meaningful analysis. I will therefore rely on the results from Liu et al., 2022, who investigated the same question with a much larger sample – although they did only look at the climate impact of tropical eruptions that co-occurred (or didn’t) with El Ninos.
Finally, to address your first point, the high/low classifications in Table 3 were based on the statistical significance test from the SEA methodology (i.e p<0.05 and outside the 95% confidence band). This will be made explicit in the text and I will also consider replacing the table with a textual explanation of the revised ENSO SEA results, as I agree that the table is somewhat confusing. I will also be careful to only draw conclusions where there is some level of agreement between results from the different ENSO reconstructions.
Lagged correlation analysis:
The authors tested lagged correlations between the SAOD volcanic forcing index and various environmental/societal indices, finding none or very weak relationships. This actually adds little value to the main narrative and does not match other results in Figure 5 and 6. I suggest to remove it, otherwise the author has to improve the way of assessment. Lagged correlation is a good approach to assess lagged effect in time however it can also be sensitive depending on the temporal scale (i.e., seasonality) and the parameter in question. By reviewing other figures, I feel the lagged effects for crop failure and famine may associate with severe climate events (i.e., drought) but not volcanic eruption itself. So, there is something for author to reconsider.Accepted. As you say, the results are mainly inconclusive and the most statistically significant results probably simply demonstrate the climate-crop failure-famine links already identified in the SEA and correlation analysis.
Minor comment:
The manuscript presents a writing style easy to understand and follow. However, several typos need to be corrected, and certain sections would benefit from a more formal academic writing style. Below I list a few examples:Noted. Thankyou for identifying these lines – I will correct these and recheck the manuscript to remove any other informal or vague wording.
Line 17-18, ‘Many of these factors form feedback loop than can delay, amplify or counteract volcanic effects’
Line 262, 'r value indicate at best very weak correlations’
Line 319, ‘breading conditions for locusts’
Line 354, ‘eastern China might partly be due variable hydroclimatic impacts’Citation: https://doi.org/10.5194/egusphere-2026-1228-AC2
-
AC2: 'Reply on RC2', Richard Warren, 07 Aug 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 413 | 244 | 23 | 680 | 34 | 35 |
- HTML: 413
- PDF: 244
- XML: 23
- Total: 680
- BibTeX: 34
- EndNote: 35
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “Connecting volcanic climate impacts to famine in China using the REACHES database”
The manuscript provides a comprehensive analysis of the link between volcanic eruptions and famine in eastern China during 1440–1900 CE using the REACHES data. The author argues that volcanic activity does not necessarily cause famine, although some of their analyses suggest it could increase the likelihood of droughts and floods, thereby increasing associated agricultural and societal risks, including crop failure. Overall, this study is a valuable contribution to understanding volcano–climate–society relationships. However, I recommend clarifying some aspects of methodology and including additional analyses and discussion. I would also like to note that my background is in climate dynamics, so my review of the context and discussion on agricultural and societal impacts may be limited.
Major comments
Minor comments
Line comments
Line 10: I recommend including a brief description of the REACHES database.
Lines 16–17: What are considered “non-volcanic climate processes” here?
Lines 25–26: What periods exactly are meant by “early Chinese history” and “later periods”? I think it is crucial to state specific periods since later in this paragraph it states “a persistent connection between volcanic activity and famine has not been systematically investigated,” which implies a comprehensive analysis, and it is probably better to make sure those “early” and “later” periods are included in this study’s focus 1440–1900 CE.
Line 30: It may be worth briefly defining the difference between “climatic” and “environmental” pathways. It is clear once you introduce the VICES approach in Section 2.1, but not here yet.
Line 49: Again, I recommend including a brief description of the REACHES database if you are mentioning it before the Data and Methodology section. Is REACHES an abbreviation?
Line 67: Can you clarify what is meant by “perception and meaning” here?
Line 89: I recommend referring to figures and tables in parentheses instead of a dash [e.g., (see Fig. 1) instead of - see Fig. 1].
Line 96: Did you mean Table 1?
Lines 102–103: How were these four eruptions chosen as case studies? What is special about these events? I also want to point out that mentioning these events before explaining how major eruptions were chosen is a little bit confusing.
Line 117: What period is used to define the 95th percentile of SST anomalies? Is it the same as the study period of 1440–1900 CE?
Line 131: What was considered “a sufficiently large number” of reports?
Lines 161–169: I assume the author computed post-eruption anomalies relative to some pre-eruption baseline period to perform Superposed epoch analysis. What was the baseline period used in this study? Also, were volcanic years included in the original data used in the Monte Carlo model test?
Lines 176–181: Are the case study analyses also based on anomalies relative to some reference period, or on absolute index values?
Figure 3: I recommend picking a different color for either the non-volcanic year or Yr 0. Black and grey dots are hard to distinguish, unfortunately.
Figures 4–5: It may be beneficial to include individual events as thinner lines in these Superposed epoch analysis figures.
Table 4: It is just a suggestion, but it may be worth using a heatmap and hatching the insignificant correlations, rather than simply omitting the insignificant coefficient values.
Figure 6: A more detailed explanation would be appreciated in the caption. For instance, what are the top and bottom rows in each panel? It may also be worth putting lag years next to actual years in CE [e.g., 1809 (Lag Year 0)].
Lines 393–394: I would be curious what the results look like if only tropical eruptions are considered. Since there are still 10 events, I think the analysis can be performed reasonably well.
Line 341: How are the volcanic impacts on climate (both summer temperature and drought) discussed here consistent with other previous studies that looked at the last millennium? There seems to be some post-eruption drying in northern China and wetting in southern China, according to Tejedor et al. (2021, PNAS), which studied 1000–1850. Could the lack of hydroclimatic impact in southern China mentioned in Line 370 be due to the selection of events?
Lines 375–389: I enjoyed this paragraph, especially I appreciated the explanation of the different crop types and their sensitivity to climate conditions.
Lines 403–404: The dash and en dash are used inconsistently throughout the text. Please make sure to use a consistent notation!