the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Climatic impacts on mortality in pre-industrial Sweden
Abstract. Climate variability and change, as well as extreme weather events, have notable impacts on human health and mortality. In historical times, the effect of climate on health and mortality was presumably stronger than today, owing to that nutrition status was meditated through climatic impacts on food production along with factors such as poor housing and healthcare. Despite this, climatic impacts on mortality in the past remain poorly understood. This study aims to improve the understanding of historical climate effects on mortality, using annual mortality records and meteorological data from Sweden between 1749 and 1859. The analysis includes the entire population as well as subgroups based on sex and age. A statistically significant negative correlation was found between winter and spring temperatures and mortality (i.e., lower temperatures = higher mortality and vice versa). We demonstrate that colder winters and springs were linked to higher mortality levels, not only for the same year but also the following year. Conversely, no statistically significant associations were observed between summer or autumn temperatures and mortality, and only weak associations existed with precipitation. The impact of winter– spring temperature on mortality was most pronounced for the same year in southern Sweden and during the 19th century, but stronger for the following year in central Sweden and during the 18th century. These findings call for further research, especially investigating specific diseases and additional contributing factors to the observed increase in mortality following cold winter and spring conditions in Sweden during the late pre-industrial period.
- Preprint
(9414 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on cp-2023-92', Anonymous Referee #1, 25 Dec 2023
The comment was uploaded in the form of a supplement: https://cp.copernicus.org/preprints/cp-2023-92/cp-2023-92-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Fredrik Charpentier Ljungqvist, 23 Jan 2024
Response to reviewer #1 of Chen et al.
Comment: It includes an excellent review of the literature. I suggest to add some information from the public health point of view, including a description of the annual cycle of mortality rate whose maximum should coincide with the coldest months of the year. I found an article that claims that for cohorts born in 1800 the risk of dying during the winter season was almost twice that of dying during summer (Ledberg, 2020).
Reply: Thank you for your positive evaluation of our review of the literature. You bring up two aspects, first how mortality varies over the calendar year, and next that this variation has declined with later-born cohorts, which is the conclusion of the suggested article by Ledberg (2020). This is very interesting and likely the result of many factors, for example, better housing, and better overall health especially concerning respiratory function (reduced smoking).
We have, as the reviewer suggests, now added some additional information about the annual cycle of mortality rate, relevant from a public health point of view, and also included a sentence in the revised manuscript summarising the main results of Ledberg (2020). Ledberg (2020) is interesting, although it should be pointed out that Ledberg (2020) only covers the period after 1860. Currently there are no monthly national series for Sweden for the 18th century, and there could be large regional differences. For example, if the impact of harvests were larger in earlier periods, deaths may have spread more during the whole year – as could also be the effect of epidemics and wars.
The annual cycle of mortality is, however, very complex. Not only does the effect vary by region, but also seems to vary for the occurrence of the disease and prognosis of the disease. For example, findings from Spain in The Lancet Regional Health – Europe (https://www.thelancet.com/journals/lanepe/article/PIIS2666-7762(23)00176-X/fulltext) shows a strong seasonal fluctuation in in-hospital mortality from respiratory diseases, and while hospital admissions were higher during the cold season, the maximum incidence of inpatient mortality was during the summer and was strongly related to high temperatures.
We also added the following text in the Introduction section to provide more information on the annual cycle of mortality:
“The total number of deaths follow an annual cycle, with higher mortality during the winter months in the extra-tropical Northern Hemisphere, a pattern also observed in contemporary Sweden. Many, but not all, causes of deaths display varying seasonal patterns. For example, data from the United States show that pneumonia, influenza, chronic obstructive pulmonary disease (COPD), and other respiratory diseases, peak during the winter months. A similar pattern, though less pronounced, is observed for cardiovascular diseases. Conversely, deaths from traffic accidents peak during summer months, while suicide and cancers show small or no seasonality. However, the differences vary by region and by baseline mortality level. In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates.”Comment: I suggest to indicate the sizes of the three age groups: 0–14 years, 15–64 years, over 65 years.
Reply: Thank you for pointing this out. Information of the sample size is now added for each of the three age groups in Table 3 of the revised manuscript.
Comment: It is not clear the purpose of including Fig. 1 showing the seasonal histograms of the Uppsala temperature record. As this information is not used later on in the analysis, I suggest to eliminate it.
Reply: We respectfully disagree on this point. We think that the seasonal temperature distribution and their temporal changesare of significant interest to provide readers with valuable insight to understand the climate of (central) Sweden. We motivate the inclusion of Uppsala temperature record in the following way: “The Uppsala temperature record is presented because it is the longest available record in present-day Sweden covering the studied period.” However, if the editor deems it necessary, we are open to the possibility of excluding Fig. 1, the Uppsala temperature record, from the article.
Comment: In line 145 the study period is defined as 1750–1859. However in other parts of the text is mentioned as 1749–1859 (lines 6, 95, 111, 135, 138, 158, figure caption of Fig. 1).
Reply: We agree with the need for clarification, and will incorporate this into the revision. Information from 1749 is only included to calculate excess mortality for 1750.
Comment: It is a general consensus that in temperate climates mortality rate is highest during winter, let say from November to March. Then, as the data for mortality rate is only available on an annual basis, the causes for an anomalously high value of mortality for a certain year can be attributed to anomalously cold weather either at the beginning of the year (January–February), at the end of the year (November–December) or both at the beginning or at the end of the year. This difficulty for the interpretation of the results is not mentioned when presenting the available data.
Reply: Thank you for bringing this issue up. In the original manuscript, we only included this difficulty for the interpretation in the Discussion section. We have now, as the reviewer suggests, also added a short commentary on this in the presentation of the data: “Annual mortality data may not capture the seasonal nuances of mortality rates in temperate climates, where peak mortality typically occurs during the winter months, roughly from November to March. This limitation makes it challenging to attribute anomalously high mortality rates during a specific year solely to the effects of exceptionally cold conditions during certain months (e.g., January–February or November–December). Despite this limitation, our study provides valuable insights into the potential connection between climate and mortality. If adverse climatic conditions have an immediate effect on mortality, we would expect correlations without time lags in our annual data. However, even in annual data, correlations with time lags may emerge if there is a delayed effect between climate conditions and mortality. Future research using more granular monthly mortality data could offer a more detailed understanding of these dynamics.”
Comment: The correlation technique is generally used to test an hypothesis for a relationship between two variables. What it would be the hypothesis for a relationship between the mean regional temperature for a single month and the annual rate of mortality that are behind the correlations presented in Tables 1, 2, 3, 4, A1 and A2? Is it plausible to consider that mean temperature during a single month will have an impact on the annual rate of mortality? This question if particularly valid for months at the end of the year, considering that to a large extent the mortality rate is determined by the deaths that occurred during the previous months.
I suggest to use a more precise language when extracting conclusions from correlations relatively low in magnitude, although statistically significant. For example, from correlations presented in Table 1 it is mentioned in line 190 “Colder winters and springs were associated with higher mortality and vice versa”. In fact, the correlations of the order of –0.3 explain only 9% of the mortality variance. A scatter diagram (which is not presented) would illustrate the weakness of the link among the two variables. Later on in the text it is shown that the relatively low magnitude of the correlations derives from the fact that the relationship between temperature and mortality rate during winter and spring does not persist during the entire study period, due to geographical changes in its intensity (see Fig. 7).Reply: We appreciate this suggestion and will strive to use more precise and careful language when formulating our conclusions in the revised manuscript. Although we acknowledge the relatively low magnitude of correlations, we consider fluctuations of ~10% in mortality attributed to climate as rather noteworthy. The effect of spring temperature on excess death is demonstrated in the scatter diagram (below).
Comment: I suggest to eliminate the discussion about the relationship between the rate of mortality and the PDSI index. Correlations presented in Table A2 are mostly not statistically significant and explain at most 4% of the variance of mortality. Furthermore, it is mentioned in line 151–152 that precipitation measurements can be considered unreliable prior to the late 19th century in Sweden. Regarding this, in the discussion section there is another example of a statement that overrates the results that were obtained. Line 307: “While our findings demonstrate the influence of a wet autumn on mortality in both western and eastern Sweden…”. This statement is not supported by the results presented in Table A2, showing very low and mostly not statistically significant correlations between mortality and PDSI data for March, April and May. As in the case of temperature, it is possible that the positive correlations are stronger for certain periods, but this analysis was not performed.
Reply: We will reconsider, after receiving the second review and hearing the editor’s decision, whether we should include or exclude hydroclimate (PDSI) from the study.
Comment: There are several references to results showing the impact of winter and spring temperature on mortality (i.e. lines 255, 270, 282, 293). However, surprisingly, most of the correlations shown in Tables 1–4, A1 and A2 for December, January and February are not statistically significant. It is more correct to indicate that the results reveal some impact of low temperature on mortality during late winter and spring. Same objection is valid for statements in the Abstract and in the Conclusions.
Reply: We will in the revised manuscript make changes, as suggested by the reviewer, emphasising the effect of temperature on late winter and spring rather than winter and spring.
Comment: Lines 367–369: “… we found that the southern-most regions experienced the greatest impact of temperature on mortality during the same year. Conversely, central Sweden exhibited the strongest temperature effect on mortality in the following year”
Comment: In my opinion this conclusion does not summarize in an adequate way the results presented in Fig. 7. This figure shows significant geographical changes of these impacts during the study period. Thus, in the eastern – central part of the country anomalously low temperature during FMA tended to be associated with above average mortality rate in the period 1805–1859, but not in the period before. Do the author have and hypothesis for this?. On the other hand, while during the period 1750–1804 relatively large mortality rate tended to prevail in the whole region the year following an anomalously cold FMA, this relationship was restricted to the centraleastern portion of the country afterward. Do the authors consider that improved food and nutrition security did not reach that part of the country during that period (1804–1859)?Reply: We will add the following clarification in the Conclusion section of the revised manuscript: “This suggests that in central and northern Sweden, the influence of temperature on mortality was primarily attributed to the adverse effects of cold weather on agricultural harvests, while this was not the case in southern Sweden. To establish causation definitively, future studies should incorporate data on regional harvest and monthly mortality.”
Comment: Lines 371–372: “Among adults, colder conditions in April had the most adverse effect on mortality both for the same year and the following year”. I question this statement referring to the impact of temperature during a single month, considering that Table 3 shows correlations similar in magnitude to that in April for the periods February–April, March–April and March April and May.
Reply: We agree with the reviewer, and we have hence now removed this sentence.
Citation: https://doi.org/10.5194/cp-2023-92-AC1
-
AC1: 'Reply on RC1', Fredrik Charpentier Ljungqvist, 23 Jan 2024
-
CC1: 'Comment on cp-2023-92', Bertil Forsberg, 20 Mar 2024
Climate impacts on mortality in pre-industrial Sweden
T.T. Chen, R- Edvinsson, K. Modig, H.W. Linderholm, F. C Ljungqvist
Recommendation: To be published after some changes.
1.1 Climate-mortality relationships: past and present
It would be useful here with a clear separation between observed seasonal and annual relationships and the known effects of hot and cold days and short episodes of extreme temperatures on the daily mortality, including the different lag structure as a result of different mechanisms. There are a few studies covering quite long periods e.g. Åström DO, Forsberg B, Edvinsson S, Rocklöv J. Acute fatal effects of short-lasting extreme temperatures in Stockholm, Sweden: evidence across a century of change. Epidemiology. 2013 Nov;24(6):820-9. doi: 10.1097/01.ede.0000434530.62353.0b. PMID: 24051892.I lack a discussion about harvesting effects, how high mortality among vulnerable (elderly) could be followed by lower mortality and weaker effects of exposure. This effect can be visible also with moderate increases in deaths, see Rocklöv J, Forsberg B, Meister K. Winter mortality modifies the heat-mortality association the following summer. Eur Respir J. 2009 Feb;33(2):245-51. doi: 10.1183/09031936.00037808. Epub 2008 Sep 17. Erratum in: Eur Respir J. 2009 Apr;33(4):947. PMID: 18799511.
In the last paragraphs discussing the attenuated effects of cold, one could also expect vaccination of elderly and lower air pollution levels to modify the relationships.
2.3 Statistical methods
It would be interesting to know if the expected number of deaths in the oldest group (65+) would differ much (due to harvesting) if you use a baseline/trend without the last year and mortality from last year as a separate variable, or if annual numbers are too crude.4.1 Discussion
The lack of an effect the following year in the oldest group could be a harvesting effect (see above!).
Could cold winter-spring months also have resulted in a lack of wood fuel to heat homes?
Citation: https://doi.org/10.5194/cp-2023-92-CC1 -
AC2: 'Reply on CC1', Fredrik Charpentier Ljungqvist, 21 May 2024
Comment: Climate-mortality relationships: past and present
It would be useful here with a clear separation between observed seasonal and annual relationships and the known effects of hot and cold days and short episodes of extreme temperatures on the daily mortality, including the different lag structure as a result of different mechanisms. There are a few studies covering quite long periods e.g. Åström DO, Forsberg B, Edvinsson S, Rocklöv J. Acute fatal effects of short-lasting extreme temperatures in Stockholm, Sweden: evidence across a century of change. Epidemiology. 2013 Nov;24(6):820-9. doi: 10.1097/01.ede.0000434530.62353.0b. PMID: 24051892.
I lack a discussion about harvesting effects, how high mortality among vulnerable (elderly) could be followed by lower mortality and weaker effects of exposure. This effect can be visible also with moderate increases in deaths, see Rocklöv J, Forsberg B, Meister K. Winter mortality modifies the heat-mortality association the following summer. Eur Respir J. 2009 Feb;33(2):245-51. doi: 10.1183/09031936.00037808. Epub 2008 Sep 17. Erratum in: Eur Respir J. 2009 Apr;33(4):947. PMID: 18799511.
In the last paragraphs discussing the attenuated effects of cold, one could also expect vaccination of elderly and lower air pollution levels to modify the relationships.
Reply: Thank you for your comment. It is, unfortunately, not possible to make a “separation between observed seasonal and annual relationships” as we for the period only have access to annual mortality data. The data available from Tabellverket is only showing annual deaths per parish. We will, however, cite the suggested reference by Åström et al. (2013) as well as to the Discussion add a paragraph about the “harvesting effect” among the elderly not at least very well-known from the recent Covid-19 pandemic. We agree that both vaccination of elderly and lower air pollution levels would alter the temperature–mortality relationship but this is hardly relevant for a study only covering the pre-1860 period.
Comment: Statistical methods
It would be interesting to know if the expected number of deaths in the oldest group (65+) would differ much (due to harvesting) if you use a baseline/trend without the last year and mortality from last year as a separate variable, or if annual numbers are too crude.
Reply: While we fully acknowledge that this would be an interesting issue to further investigate, this would both complicate the calculations and make an already long article too long. Thus, we decide that this is an issue to instead further investigate in future studies. We will, however, consider to add one or two sentence about this issue in the Discussion section of the revised article.
Comment: Discussion
The lack of an effect the following year in the oldest group could be a harvesting effect (see above!).
Could cold winter–spring months also have resulted in a lack of wood fuel to heat homes?
Reply: We will, as stated above, add a sentence or two about the possible “harvesting effect” among the elderly in the Discussion section. Regarding a possible lack of wood fuel to heat homes during cold winter–spring months we are not aware of that a lack of fuel in Sweden during this period should have been a major issue although parts of Skåne had a relative lack of wood locally. In the absence of clear evidence for this in the historical literature, we hesitate to include this possibility in any discussion.
Citation: https://doi.org/10.5194/cp-2023-92-AC2
-
AC2: 'Reply on CC1', Fredrik Charpentier Ljungqvist, 21 May 2024
-
RC2: 'Comment on cp-2023-92', Anonymous Referee #2, 30 Apr 2024
This study presents an analysis of the correlation between climate and mortality in Sweden during the period 1749-1859. Through extensive data processing, the authors succeed in showing that cold winter-springs correlate with increased mortality, both in the same year and the following year. In contrast, summer and autumn temperatures and precipitation showed only weak correlations with mortality. Age-wise, children (0-14 years) showed increased mortality in the year following a year of cold winter-spring, while the over-65 age group experienced higher mortality in the same year as a cold winter-spring and, finally, the 15-65 age group had higher mortality both in the same year and in the year following a year of cold winter-spring. Geographically, mortality increased in southern Sweden in the same year as a cold winter-spring, while in east-central Sweden mortality increased mainly in the year following a cold winter-spring. Over time, the correlation between a cold winter-spring and mortality in the following year decreases, while the correlation between a cold winter-spring and mortality in the same year emerges after 1790 but weakens towards the end of the study period.
Of course, a historian is not satisfied with simply reading signals from historical source material as a purely methodological test, but also wants to see if the results “make sense” in the historical and social context. The plausibility of such reasoning is at the same time an indicator of the quality of the correlations demonstrated. The article also discusses why one variable affects the other. Cold winter-springs are said to affect mortality in two ways: firstly, cold winter-springs result in less food production, leading to malnutrition and increased mortality, and secondly, cold weather can increase the risk of disease spread due to overcrowding or reduced resistance to respiratory diseases.
A critical point is that the current discussion about the effects of historical climate conditions on mortality could be clearer and more in-depth. One reason is that, despite the processing of large amounts of data, the evidence does not provide clear-cut answers. One alternative, of course, would have been for the authors to stop at the ambition of establishing various correlations with the study. In this case, the third question about how climate “contribute to" mortality should be changed to “correlate with” as there are several factors that are not corrected for in the analysis. But such a pure methodological test risks being rather bland.
Of the two different links between climate and mortality highlighted by the authors, one of them goes through the intermediary of harvest. This is where the discussion can become clearer. In his thesis from 2006, Daniel Larsson - who, unlike this study, worked with monthly data for mortality - interpreted the increased mortality during the winter-spring as a sign of food shortages due to a poorer harvest the previous year. This seasonal peak in mortality declined over time, which Larsson saw as a direct consequence of an improved supply situation. This study instead looks at annual mortality rates but monthly climate data. Thus, we don't know what time of year people died. However, the increased mortality in the year following a year with a cold winter-spring could reasonably be linked to a poor harvest in between. The fact that the correlation weakens over time could also be interpreted as a general improvement in the livelihood situation, which is exactly what happens during the period.
A crux, however, is that we lack the central intermediate stage of harvest in this study, while much of the reasoning about climate variations and mortality revolves around the nutritional situation. The authors could have been more precise in their discussion if they had included a measure of harvest in the calculations. Yield ratios are difficult to reconstruct and are also subject to uncertainties, but such data series are not lacking in the scientific community. There is a vagueness in the argumentation when they skip one of the more important intermediate stages.
In this context, it is worth noting how literature is used. To build the cold spring-winter - poorer harvest - higher mortality link, two works are highlighted, both of which claim that the temperature during late winter and early spring played “a critical role” in cereal production in south-central Sweden (Edvinsson et al. (2009) and Holopainen et al. (2012)). However, Skoglund's thesis from 2023 shows no correlation between cold winter-springs and harvests in southernmost Sweden, but instead that harvests varied most with summer temperature and precipitation (summer drought). Spring-winter temperature as a factor for harvest was more important in northern Sweden, where the length of the growing season was critical. When different studies seem to achieve contradictory results, there is a risk of “cherry-picking” - choosing the studies that confirm your own results. I do not claim that the authors do this here. However, the heterogeneity of the research literature on the issue does seem to demonstrate how complex the relationship between climate variability and crop yields appears to be. Simple, unambiguous answers are not given. The study might do well to enter more clearly into a dialogue with the various results available in the literature. (In the context of Skoglund (2023), one might also emphasize the need for combined precipitation and temperature data in the calculation model to distinguish a hot but humid summer from a summer with drought).
As for the increased mortality in the same year as a year with a cold winter-spring, the lack of monthly mortality data makes analyses more difficult to make. We cannot know whether the increased mortality occurred at the same time as the cold winter-spring or in the autumn, after the harvest, when mortality could hypothetically be related to the nutritional situation (due to a poor harvest). The authors themselves seem to see this mortality (in the same year as a cold winter-spring) mainly as an effect of the cold weather itself. They mention increased risk of spreading infectious diseases as people spend more time indoors due to the cold and reduced resistance to respiratory infections due to cold weather. I think there should be given more room for an account of the scientific debate and an explanatory model showing that cold weather really leads to disease to support such an interpretation. Some space should also be given to address the geographical pattern in this context. According to the study, populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter-spring. Perhaps an interpretation of the correlation could be more clearly based on this seemingly paradoxical relationship?
One reflection is that in the study of how climatic variations affect mortality, the authors find relatively weak evidence that poor harvests increased mortality. According to the study, the mortality of the adult population seems to be more correlated with cold winter-springs the same year than the year before. It is, if I understand correctly, the mortality of children (in east-central Sweden) that clearly correlates with cold winter-spring in the previous year. One problem here, as mentioned, is that the authors do not measure harvests. This leads to somewhat vague analysis. The relationship between weather and harvest (and mortality) is likely to be more complex than can be captured in a computational model like the one used here. It is not that the results suggest that nutrition was not important, but that this study cannot capture this relationship to more than a small extent.
Furthermore, a limitation of the study is that it works only with correlations and not with regressions, which would have given us an opportunity to assess more clearly the strength of the impact of the reported relationships on mortality. To assess more clearly how much mortality is affected by the factors studied, such as temperature in winter-spring, an appropriate measure could have been to include a scatterplot. By showing the mortality rate against the temperature variation, such a plot would allow a visual assessment of the strength of the relationship.
I also have a comment on Figure 6. One curve shows the correlation over time between cold springs and mortality in the following year (31-year moving window). This curve shows a strong correlation that starts to weaken towards the end of the study period. This is logical. As the livelihood situation in the country improves based on new cultivation and other factors, this category of mortality declines. The second curve shows the relationship between a cold winter-spring and mortality in the same year. This curve shows a very weak correlation until around 1790. After that, the correlation is strong until around 1825 and then weakens. This curve is difficult to interpret. Readers may be suspicious that some strong observations in the early 19th century are behind this stronger correlation. The pattern is also difficult to explain. The authors suggest that this correlation was not visible during the first period studied (1750-1790) due to a generally higher mortality level then. In other words, we would be dealing with a pattern that we can detect statistically only when mortality starts to fall. This could indeed be true. But shouldn't the correlation continuously strengthen in such a case? Why does the correlation become weaker again after about 1825?
All in all, this is a solid and very interesting study, as evidenced by the fact that it gives rise to many thoughts and questions. It is a shortcoming to omit the intermediate factor of harvest, but I still think the authors get away with it based on the tight structure and a certain awareness of the problem. However, there is a lot of noise in a study that skips such a step. But what comes through the noise is still clear. A cold winter-spring is likely to result in high mortality in the following year and implies a poorer nutritional status. The geographical pattern of mortality in the same and following year, respectively, after a cold winter-spring is remarkable and I believe still awaits explanation. The temporal curve of the relationship between mortality and a cold winter-spring in the same year is also still somewhat puzzling to me. In any case, I hope to see the article published. I believe it will generate new studies that will help researchers refine their formulations, approaches and methods.
Citation: https://doi.org/10.5194/cp-2023-92-RC2 -
AC3: 'Reply on RC2', Fredrik Charpentier Ljungqvist, 21 May 2024
Comment: This study presents an analysis of the correlation between climate and mortality in Sweden during the period 1749–1859. Through extensive data processing, the authors succeed in showing that cold winter–springs correlate with increased mortality, both in the same year and the following year. In contrast, summer and autumn temperatures and precipitation showed only weak correlations with mortality. Age-wise, children (0–14 years) showed increased mortality in the year following a year of cold winter–spring, while the over-65 age group experienced higher mortality in the same year as a cold winter–spring and, finally, the 15–65 age group had higher mortality both in the same year and in the year following a year of cold winter–spring. Geographically, mortality increased in southern Sweden in the same year as a cold winter–spring, while in east-central Sweden mortality increased mainly in the year following a cold winter–spring. Over time, the correlation between a cold winter–spring and mortality in the following year decreases, while the correlation between a cold winter–spring and mortality in the same year emerges after 1790 but weakens towards the end of the study period.
Of course, a historian is not satisfied with simply reading signals from historical source material as a purely methodological test, but also wants to see if the results “make sense” in the historical and social context. The plausibility of such reasoning is at the same time an indicator of the quality of the correlations demonstrated. The article also discusses why one variable affects the other. Cold winter–springs are said to affect mortality in two ways: firstly, cold winter–springs result in less food production, leading to malnutrition and increased mortality, and secondly, cold weather can increase the risk of disease spread due to overcrowding or reduced resistance to respiratory diseases.
Reply: Thank you for a very thorough and positive summery of our article. We highly appreciate the long and detailed review of our work.
Comment: A critical point is that the current discussion about the effects of historical climate conditions on mortality could be clearer and more in-depth. One reason is that, despite the processing of large amounts of data, the evidence does not provide clear-cut answers. One alternative, of course, would have been for the authors to stop at the ambition of establishing various correlations with the study. In this case, the third question about how climate “contribute to” mortality should be changed to “correlate with” as there are several factors that are not corrected for in the analysis. But such a pure methodological test risks being rather bland.
Reply: While we acknowledge that our study is not able to provide clear-cut answers with regard to the causality between changes in temperature and changes in mortality patterns, we agree with the reviewer that to only describe our finding in terms of “correlate with” would make the article and its findings too bleak. In the revision, we will thus – in line with the wish of the reviewer – deepen the climate–mortality relationship discussion both in the background section and in the discussion section. In the Introduction section, we will among other things add:
“The total number of deaths follow an annual cycle, with higher mortality during the winter months in the extra-tropical Northern Hemisphere, a pattern also observed in contemporary Sweden (Statistics Sweden, 2020). Many, but not all, causes of deaths display varying seasonal patterns. For example, data from the United States show that pneumonia, influenza, chronic obstructive pulmonary disease (COPD), and other respiratory diseases, peak during the winter months. A similar pattern, though less pronounced, is observed for cardiovascular diseases. Conversely, deaths from traffic accidents peak during summer months, while suicide and cancers show small or no seasonality (Rau et al., 2017). However, the differences vary by region and by baseline mortality level. In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates (Eurowinter Group, 1997).”
Furthermore, we note that it cannot exist any endogeneity in the present case as temperature may well influence mortality whereas mortality cannot influence temperature.
Comment: Of the two different links between climate and mortality highlighted by the authors, one of them goes through the intermediary of harvest. This is where the discussion can become clearer. In his thesis from 2006, Daniel Larsson – who, unlike this study, worked with monthly data for mortality – interpreted the increased mortality during the winter–spring as a sign of food shortages due to a poorer harvest the previous year. This seasonal peak in mortality declined over time, which Larsson saw as a direct consequence of an improved supply situation. This study instead looks at annual mortality rates but monthly climate data. Thus, we don’t know what time of year people died. However, the increased mortality in the year following a year with a cold winter–spring could reasonably be linked to a poor harvest in between. The fact that the correlation weakens over time could also be interpreted as a general improvement in the livelihood situation, which is exactly what happens during the period.
A crux, however, is that we lack the central intermediate stage of harvest in this study, while much of the reasoning about climate variations and mortality revolves around the nutritional situation. The authors could have been more precise in their discussion if they had included a measure of harvest in the calculations. Yield ratios are difficult to reconstruct and are also subject to uncertainties, but such data series are not lacking in the scientific community. There is a vagueness in the argumentation when they skip one of the more important intermediate stages.
Reply: While the reviewer correctly emphasizes the importance of harvest data in order to better understand historical climate–mortality relationships, several issues compromise its reliability and representativeness. Frequently, yield ratios are collected from individual farms, which may not accurately reflect entire regions and tend to be biased towards the most prosperous ones. Tithes are frequently used as an indicator of harvest volume; however, research indicates that in Sweden, tithes captured only 25 to 50 percent of the actual harvests, with some being fixed amounts. Even when tithes vary, their use as an indicator of harvest quality is problematic; they tend to exaggerate fluctuations since, in poor harvest years, tithes payments are drastically reduced. Moreover, harvest data do not necessarily correlate with food availability, particularly in smaller regions which might offset poor local yields by importing grain from elsewhere. Consequently, the link between harvest data and mortality rates is complex and not always direct.
Nevertheless, taken the reviewer’s concern into consideration, we have decided to add two new paragraphs to the discussion section including what the reviewer seems to ask for:
“The following-year mortality response to climate variability is presumed to mainly occur by affecting nutritional situation through the intermediary of harvest outcome and, to a lesser extent, livestock mortality. We acknowledge that the exclusion of harvest data, as a central intermediate stage between climate and mortality, constitute a major limitation in the present study. However, we leave the complex issue of harvest–mortality relationships to further studies due to several issues compromising the reliability and representativeness of 18th and 19th century Swedish harvest data. First, yield ratio series were typically collected from individual farms, which may not accurately reflect entire regions and may not even be locally representative as they tended to be biased towards the most prosperous farms or manors (Slicher van Bath, 1963). Second, tithes are frequently used as an indicator of harvest volume (Leijonhufvud, 2001); however, tithes in Sweden captured only 25 to 50 percent of the actual harvests, with some being fixed amounts, and for the study period only small, and over time decreasing, portions of Sweden still paid taxes in tithes reflecting the actual harvest variations (Hallberg et al., 2016). Third, harvest data do not necessarily translate into actual food availability as grain and other food sources could be, and were, imported from other regions (Edvinsson, 2012), or from aboard (Åmark, 1915), although the presumed climatic effect on following-year mortality reasonably must be interpreted as a result of local to regional harvest outcome.
At present, a reconstruction of Swedish harvests (Edvinsson, 2009) is available for our study period only until 1820 and it has, as a national average, limited explanatory value for studying regional harvest–mortality relationships. This harvest reconstruction, however, indicates an increasing per capita harvest from about 1810, and with setbacks already by the 1790s, signifying increased food security. The per capita harvests, when also including potatoes, were about 15–20% higher around 1820 than they had been prior to around 1790 (Edvinsson, 2009). This agrees well with our finding that the relationship between winter–spring temperature and the following year’s mortality becoming insignificant during the 1820s. Furthermore, we note that this concurs with the findings by (Larsson, 2006). In his work with mortality data from selected parishes, (Larsson, 2006) also interpreted increased mortality during the winter–spring season as indicative of malnutrition due to previous year harvest failures, and he found that these seasonal mortality peaks declining over time as a result on an improved food supply situation.”
Comment: In this context, it is worth noting how literature is used. To build the cold spring–winter – poorer harvest – higher mortality link, two works are highlighted, both of which claim that the temperature during late winter and early spring played “a critical role” in cereal production in south-central Sweden (Edvinsson et al. (2009) and Holopainen et al. (2012)). However, Skoglund’s thesis from 2023 shows no correlation between cold winter–springs and harvests in southernmost Sweden, but instead that harvests varied most with summer temperature and precipitation (summer drought). Spring–winter temperature as a factor for harvest was more important in northern Sweden, where the length of the growing season was critical. When different studies seem to achieve contradictory results, there is a risk of “cherry-picking” – choosing the studies that confirm your own results. I do not claim that the authors do this here. However, the heterogeneity of the research literature on the issue does seem to demonstrate how complex the relationship between climate variability and crop yields appears to be. Simple, unambiguous answers are not given. The study might do well to enter more clearly into a dialogue with the various results available in the literature. (In the context of Skoglund (2023), one might also emphasize the need for combined precipitation and temperature data in the calculation model to distinguish a hot but humid summer from a summer with drought).
Reply: We fully agree with the reviewer that climate–harvest relationships are complex, and often far from clear-cur, but we do not agree that the presentation of the literature about climate–harvest relationships have the character of “cherry-picking”. We are indeed very familiar with the thesis by Martin Skoglund from 2023 and the articles published in it and subsequently by him (the corresponding author was his co-supervisor). His results, based on different climate data (mainly from Lund), showing a relative lack of a winter and spring temperature signal but a considerable spring and summer drought signal, apply to the province of Skåne in southern-most present-day Sweden. Ongoing, still unpublished results, also reveal that the same applies to the province of Halland just northwest of Skåne. However, other research – including assessments conducted by agrarian historian Lotta Leijonhufvud – has shown that the late winter and early spring temperatures indeed plays a major role for harvests in south-central Sweden (around Lake Mälaren) as opposed to in southern-most Sweden (Skåne). Indeed, we think that one very likely reason for the lagged mortality response to late winter–early spring temperature in south-central Sweden, as opposed to the lack of such a lagged response in southern-most Sweden, actual is a direct result of harvest not being sensitive to late winter–early spring temperature in southern-most Sweden while they are more sensitive to it in south-central Sweden (with a shorter growing season making the timing of its onset more critical). We will in the revision of the article further clarify this geographical difference as it appears from the reviewer’s comment that this was not written clearly enough.
Comment: As for the increased mortality in the same year as a year with a cold winter–spring, the lack of monthly mortality data makes analyses more difficult to make. We cannot know whether the increased mortality occurred at the same time as the cold winter–spring or in the autumn, after the harvest, when mortality could hypothetically be related to the nutritional situation (due to a poor harvest). The authors themselves seem to see this mortality (in the same year as a cold winter–spring) mainly as an effect of the cold weather itself. They mention increased risk of spreading infectious diseases as people spend more time indoors due to the cold and reduced resistance to respiratory infections due to cold weather. I think there should be given more room for an account of the scientific debate and an explanatory model showing that cold weather really leads to disease to support such an interpretation. Some space should also be given to address the geographical pattern in this context. According to the study, populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter–spring. Perhaps an interpretation of the correlation could be more clearly based on this seemingly paradoxical relationship?
Reply: The lack of monthly mortality data indeed poses, as we have outlined in our response to reviewer #1, a challenge for our study. Currently there are no monthly national series for Sweden for the 18th century and if, for example, the impact of harvests was larger in earlier periods, deaths may have spread more during the whole year. We will in the revised article added an entire new paragraph in the Introduction section to provide more information on the annual cycle of mortality. We have also decided, in the revision, to add a short commentary on this in the presentation of the data:
“Annual mortality data may not capture the seasonal nuances of mortality rates in temperate climates, where peak mortality typically occurs during the winter months, roughly from November to March. This limitation makes it challenging to attribute anomalously high mortality rates during a specific year solely to the effects of exceptionally cold conditions during certain months (e.g., January–February or November–December). Despite this limitation, our study provides valuable insights into the potential connection between climate and mortality. If adverse climatic conditions have an immediate effect on mortality, we would expect correlations without time lags in our annual data. However, even in annual data, correlations with time lags may emerge if there is a delayed effect between climate conditions and mortality. Future research using more granular monthly mortality data could offer a more detailed understanding of these dynamics.”
Regarding the fact that populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter–spring we indeed think that the higher vulnerability to colder winter and early springs temperatures in regions with generally warmer winters and early springs can be related to poorer housing and heating facilities in these regions (e.g., in Skåne). We also now mention this in the Introduction that such factors play into mortality in contemporary times: “In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates (Eurowinter Group, 1997).” Regions with generally mild winters paradoxically also today tend to be those with the highest vulnerability to prolonged cold snaps. We will add an extra new sentence about this in the Discussion section in the revised manuscript.
Comment: One reflection is that in the study of how climatic variations affect mortality, the authors find relatively weak evidence that poor harvests increased mortality. According to the study, the mortality of the adult population seems to be more correlated with cold winter–springs the same year than the year before. It is, if I understand correctly, the mortality of children (in east-central Sweden) that clearly correlates with cold winter–spring in the previous year. One problem here, as mentioned, is that the authors do not measure harvests. This leads to somewhat vague analysis. The relationship between weather and harvest (and mortality) is likely to be more complex than can be captured in a computational model like the one used here. It is not that the results suggest that nutrition was not important, but that this study cannot capture this relationship to more than a small extent.
Reply: See our reply above with regard to harvests.
Comment: Furthermore, a limitation of the study is that it works only with correlations and not with regressions, which would have given us an opportunity to assess more clearly the strength of the impact of the reported relationships on mortality. To assess more clearly how much mortality is affected by the factors studied, such as temperature in winter–spring, an appropriate measure could have been to include a scatterplot. By showing the mortality rate against the temperature variation, such a plot would allow a visual assessment of the strength of the relationship.
Reply: This study primarily examines the correlations between mortality and climate (mainly temperature), rather than their direct impacts, which are more challenging to ascertain. The correlations identified are not particularly strong (only about 10% explained variance) – suggesting that relying solely on regression coefficients could be misleading. About 90% of the mortality variations are not related to climate (at least not to the climate variables under investigation here) and regressions could lead to uncertain or biased results we they majority of the mortality variations are related to factors not under investigation. While correlations can suggest certain trends, any regression analysis must also take into account relevant factors explaining more of the mortality variations to provide a more comprehensive understanding. This is not feasible in this case.
We are unsure whether it is really necessary to provide scatter-plots in the article and would like to leave it up to the editor to decide that considering that our article already is rather extensive and contains numerous figures. However, we have now created two scatter-plots as the reviewer wishes: one for spring temperature and mortality the same year and one for spring temperature and mortality the following year.
Comment: I also have a comment on Figure 6. One curve shows the correlation over time between cold springs and mortality in the following year (31-year moving window). This curve shows a strong correlation that starts to weaken towards the end of the study period. This is logical. As the livelihood situation in the country improves based on new cultivation and other factors, this category of mortality declines. The second curve shows the relationship between a cold winter–spring and mortality in the same year. This curve shows a very weak correlation until around 1790. After that, the correlation is strong until around 1825 and then weakens. This curve is difficult to interpret. Readers may be suspicious that some strong observations in the early 19th century are behind this stronger correlation. The pattern is also difficult to explain. The authors suggest that this correlation was not visible during the first period studied (1750–1790) due to a generally higher mortality level then. In other words, we would be dealing with a pattern that we can detect statistically only when mortality starts to fall. This could indeed be true. But shouldn’t the correlation continuously strengthen in such a case? Why does the correlation become weaker again after about 1825?
Reply: We agree with the reviewer that the interpretation of Figure 6 is far from clear-cut and already discusses this in the article. The most likely explanation, in our opinion, behind the weakening relationship between late winter–early spring temperature and same-year mortality after about 1825 is generally improved living condition, including nutrition, during the course of the 19th century. The general mortality level (mainly caused by infectious diseases) decreases during the 19th century along with reduced inter-annual fluctuations in death rates (as well as in birth rates). For example, smallpox becomes virtually extricated (with mandatory smallpox vaccination from 1816). We will add a sentence or two about this in the revised article. Furthermore, we will add the following clarification in the Conclusion section of the revised manuscript: “This suggests that in central and northern Sweden, the influence of temperature on mortality was primarily attributed to the adverse effects of cold weather on harvests, while this was not the case in southern Sweden. To establish causation definitively, future studies should incorporate data on both regional harvests and monthly mortality.”
Comment: All in all, this is a solid and very interesting study, as evidenced by the fact that it gives rise to many thoughts and questions. It is a shortcoming to omit the intermediate factor of harvest, but I still think the authors get away with it based on the tight structure and a certain awareness of the problem. However, there is a lot of noise in a study that skips such a step. But what comes through the noise is still clear. A cold winter–spring is likely to result in high mortality in the following year and implies a poorer nutritional status. The geographical pattern of mortality in the same and following year, respectively, after a cold winter–spring is remarkable and I believe still awaits explanation. The temporal curve of the relationship between mortality and a cold winter–spring in the same year is also still somewhat puzzling to me. In any case, I hope to see the article published. I believe it will generate new studies that will help researchers refine their formulations, approaches and methods.
Reply: Thank you again for an accurate summery of our work and for the encouraging words. We appreciate not only that the reviewer wants to see our article published, but also that he or she agree with us that additional research is needed to better constrain and understand the causal links between changes in climate variables and changes in mortality.
Citation: https://doi.org/10.5194/cp-2023-92-AC3
-
AC3: 'Reply on RC2', Fredrik Charpentier Ljungqvist, 21 May 2024
-
AC4: 'Comment on cp-2023-92', Fredrik Charpentier Ljungqvist, 21 Jul 2024
We have now provided detailed responses to all the reviewer’s comments on our manuscript individually since some time ago. As far as we can deem, we have answered each of the reviewers’ suggestions in detail. We have already almost finished the working with revising the article according to the reviewer’s suggestions.
Citation: https://doi.org/10.5194/cp-2023-92-AC4
Status: closed
-
RC1: 'Comment on cp-2023-92', Anonymous Referee #1, 25 Dec 2023
The comment was uploaded in the form of a supplement: https://cp.copernicus.org/preprints/cp-2023-92/cp-2023-92-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Fredrik Charpentier Ljungqvist, 23 Jan 2024
Response to reviewer #1 of Chen et al.
Comment: It includes an excellent review of the literature. I suggest to add some information from the public health point of view, including a description of the annual cycle of mortality rate whose maximum should coincide with the coldest months of the year. I found an article that claims that for cohorts born in 1800 the risk of dying during the winter season was almost twice that of dying during summer (Ledberg, 2020).
Reply: Thank you for your positive evaluation of our review of the literature. You bring up two aspects, first how mortality varies over the calendar year, and next that this variation has declined with later-born cohorts, which is the conclusion of the suggested article by Ledberg (2020). This is very interesting and likely the result of many factors, for example, better housing, and better overall health especially concerning respiratory function (reduced smoking).
We have, as the reviewer suggests, now added some additional information about the annual cycle of mortality rate, relevant from a public health point of view, and also included a sentence in the revised manuscript summarising the main results of Ledberg (2020). Ledberg (2020) is interesting, although it should be pointed out that Ledberg (2020) only covers the period after 1860. Currently there are no monthly national series for Sweden for the 18th century, and there could be large regional differences. For example, if the impact of harvests were larger in earlier periods, deaths may have spread more during the whole year – as could also be the effect of epidemics and wars.
The annual cycle of mortality is, however, very complex. Not only does the effect vary by region, but also seems to vary for the occurrence of the disease and prognosis of the disease. For example, findings from Spain in The Lancet Regional Health – Europe (https://www.thelancet.com/journals/lanepe/article/PIIS2666-7762(23)00176-X/fulltext) shows a strong seasonal fluctuation in in-hospital mortality from respiratory diseases, and while hospital admissions were higher during the cold season, the maximum incidence of inpatient mortality was during the summer and was strongly related to high temperatures.
We also added the following text in the Introduction section to provide more information on the annual cycle of mortality:
“The total number of deaths follow an annual cycle, with higher mortality during the winter months in the extra-tropical Northern Hemisphere, a pattern also observed in contemporary Sweden. Many, but not all, causes of deaths display varying seasonal patterns. For example, data from the United States show that pneumonia, influenza, chronic obstructive pulmonary disease (COPD), and other respiratory diseases, peak during the winter months. A similar pattern, though less pronounced, is observed for cardiovascular diseases. Conversely, deaths from traffic accidents peak during summer months, while suicide and cancers show small or no seasonality. However, the differences vary by region and by baseline mortality level. In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates.”Comment: I suggest to indicate the sizes of the three age groups: 0–14 years, 15–64 years, over 65 years.
Reply: Thank you for pointing this out. Information of the sample size is now added for each of the three age groups in Table 3 of the revised manuscript.
Comment: It is not clear the purpose of including Fig. 1 showing the seasonal histograms of the Uppsala temperature record. As this information is not used later on in the analysis, I suggest to eliminate it.
Reply: We respectfully disagree on this point. We think that the seasonal temperature distribution and their temporal changesare of significant interest to provide readers with valuable insight to understand the climate of (central) Sweden. We motivate the inclusion of Uppsala temperature record in the following way: “The Uppsala temperature record is presented because it is the longest available record in present-day Sweden covering the studied period.” However, if the editor deems it necessary, we are open to the possibility of excluding Fig. 1, the Uppsala temperature record, from the article.
Comment: In line 145 the study period is defined as 1750–1859. However in other parts of the text is mentioned as 1749–1859 (lines 6, 95, 111, 135, 138, 158, figure caption of Fig. 1).
Reply: We agree with the need for clarification, and will incorporate this into the revision. Information from 1749 is only included to calculate excess mortality for 1750.
Comment: It is a general consensus that in temperate climates mortality rate is highest during winter, let say from November to March. Then, as the data for mortality rate is only available on an annual basis, the causes for an anomalously high value of mortality for a certain year can be attributed to anomalously cold weather either at the beginning of the year (January–February), at the end of the year (November–December) or both at the beginning or at the end of the year. This difficulty for the interpretation of the results is not mentioned when presenting the available data.
Reply: Thank you for bringing this issue up. In the original manuscript, we only included this difficulty for the interpretation in the Discussion section. We have now, as the reviewer suggests, also added a short commentary on this in the presentation of the data: “Annual mortality data may not capture the seasonal nuances of mortality rates in temperate climates, where peak mortality typically occurs during the winter months, roughly from November to March. This limitation makes it challenging to attribute anomalously high mortality rates during a specific year solely to the effects of exceptionally cold conditions during certain months (e.g., January–February or November–December). Despite this limitation, our study provides valuable insights into the potential connection between climate and mortality. If adverse climatic conditions have an immediate effect on mortality, we would expect correlations without time lags in our annual data. However, even in annual data, correlations with time lags may emerge if there is a delayed effect between climate conditions and mortality. Future research using more granular monthly mortality data could offer a more detailed understanding of these dynamics.”
Comment: The correlation technique is generally used to test an hypothesis for a relationship between two variables. What it would be the hypothesis for a relationship between the mean regional temperature for a single month and the annual rate of mortality that are behind the correlations presented in Tables 1, 2, 3, 4, A1 and A2? Is it plausible to consider that mean temperature during a single month will have an impact on the annual rate of mortality? This question if particularly valid for months at the end of the year, considering that to a large extent the mortality rate is determined by the deaths that occurred during the previous months.
I suggest to use a more precise language when extracting conclusions from correlations relatively low in magnitude, although statistically significant. For example, from correlations presented in Table 1 it is mentioned in line 190 “Colder winters and springs were associated with higher mortality and vice versa”. In fact, the correlations of the order of –0.3 explain only 9% of the mortality variance. A scatter diagram (which is not presented) would illustrate the weakness of the link among the two variables. Later on in the text it is shown that the relatively low magnitude of the correlations derives from the fact that the relationship between temperature and mortality rate during winter and spring does not persist during the entire study period, due to geographical changes in its intensity (see Fig. 7).Reply: We appreciate this suggestion and will strive to use more precise and careful language when formulating our conclusions in the revised manuscript. Although we acknowledge the relatively low magnitude of correlations, we consider fluctuations of ~10% in mortality attributed to climate as rather noteworthy. The effect of spring temperature on excess death is demonstrated in the scatter diagram (below).
Comment: I suggest to eliminate the discussion about the relationship between the rate of mortality and the PDSI index. Correlations presented in Table A2 are mostly not statistically significant and explain at most 4% of the variance of mortality. Furthermore, it is mentioned in line 151–152 that precipitation measurements can be considered unreliable prior to the late 19th century in Sweden. Regarding this, in the discussion section there is another example of a statement that overrates the results that were obtained. Line 307: “While our findings demonstrate the influence of a wet autumn on mortality in both western and eastern Sweden…”. This statement is not supported by the results presented in Table A2, showing very low and mostly not statistically significant correlations between mortality and PDSI data for March, April and May. As in the case of temperature, it is possible that the positive correlations are stronger for certain periods, but this analysis was not performed.
Reply: We will reconsider, after receiving the second review and hearing the editor’s decision, whether we should include or exclude hydroclimate (PDSI) from the study.
Comment: There are several references to results showing the impact of winter and spring temperature on mortality (i.e. lines 255, 270, 282, 293). However, surprisingly, most of the correlations shown in Tables 1–4, A1 and A2 for December, January and February are not statistically significant. It is more correct to indicate that the results reveal some impact of low temperature on mortality during late winter and spring. Same objection is valid for statements in the Abstract and in the Conclusions.
Reply: We will in the revised manuscript make changes, as suggested by the reviewer, emphasising the effect of temperature on late winter and spring rather than winter and spring.
Comment: Lines 367–369: “… we found that the southern-most regions experienced the greatest impact of temperature on mortality during the same year. Conversely, central Sweden exhibited the strongest temperature effect on mortality in the following year”
Comment: In my opinion this conclusion does not summarize in an adequate way the results presented in Fig. 7. This figure shows significant geographical changes of these impacts during the study period. Thus, in the eastern – central part of the country anomalously low temperature during FMA tended to be associated with above average mortality rate in the period 1805–1859, but not in the period before. Do the author have and hypothesis for this?. On the other hand, while during the period 1750–1804 relatively large mortality rate tended to prevail in the whole region the year following an anomalously cold FMA, this relationship was restricted to the centraleastern portion of the country afterward. Do the authors consider that improved food and nutrition security did not reach that part of the country during that period (1804–1859)?Reply: We will add the following clarification in the Conclusion section of the revised manuscript: “This suggests that in central and northern Sweden, the influence of temperature on mortality was primarily attributed to the adverse effects of cold weather on agricultural harvests, while this was not the case in southern Sweden. To establish causation definitively, future studies should incorporate data on regional harvest and monthly mortality.”
Comment: Lines 371–372: “Among adults, colder conditions in April had the most adverse effect on mortality both for the same year and the following year”. I question this statement referring to the impact of temperature during a single month, considering that Table 3 shows correlations similar in magnitude to that in April for the periods February–April, March–April and March April and May.
Reply: We agree with the reviewer, and we have hence now removed this sentence.
Citation: https://doi.org/10.5194/cp-2023-92-AC1
-
AC1: 'Reply on RC1', Fredrik Charpentier Ljungqvist, 23 Jan 2024
-
CC1: 'Comment on cp-2023-92', Bertil Forsberg, 20 Mar 2024
Climate impacts on mortality in pre-industrial Sweden
T.T. Chen, R- Edvinsson, K. Modig, H.W. Linderholm, F. C Ljungqvist
Recommendation: To be published after some changes.
1.1 Climate-mortality relationships: past and present
It would be useful here with a clear separation between observed seasonal and annual relationships and the known effects of hot and cold days and short episodes of extreme temperatures on the daily mortality, including the different lag structure as a result of different mechanisms. There are a few studies covering quite long periods e.g. Åström DO, Forsberg B, Edvinsson S, Rocklöv J. Acute fatal effects of short-lasting extreme temperatures in Stockholm, Sweden: evidence across a century of change. Epidemiology. 2013 Nov;24(6):820-9. doi: 10.1097/01.ede.0000434530.62353.0b. PMID: 24051892.I lack a discussion about harvesting effects, how high mortality among vulnerable (elderly) could be followed by lower mortality and weaker effects of exposure. This effect can be visible also with moderate increases in deaths, see Rocklöv J, Forsberg B, Meister K. Winter mortality modifies the heat-mortality association the following summer. Eur Respir J. 2009 Feb;33(2):245-51. doi: 10.1183/09031936.00037808. Epub 2008 Sep 17. Erratum in: Eur Respir J. 2009 Apr;33(4):947. PMID: 18799511.
In the last paragraphs discussing the attenuated effects of cold, one could also expect vaccination of elderly and lower air pollution levels to modify the relationships.
2.3 Statistical methods
It would be interesting to know if the expected number of deaths in the oldest group (65+) would differ much (due to harvesting) if you use a baseline/trend without the last year and mortality from last year as a separate variable, or if annual numbers are too crude.4.1 Discussion
The lack of an effect the following year in the oldest group could be a harvesting effect (see above!).
Could cold winter-spring months also have resulted in a lack of wood fuel to heat homes?
Citation: https://doi.org/10.5194/cp-2023-92-CC1 -
AC2: 'Reply on CC1', Fredrik Charpentier Ljungqvist, 21 May 2024
Comment: Climate-mortality relationships: past and present
It would be useful here with a clear separation between observed seasonal and annual relationships and the known effects of hot and cold days and short episodes of extreme temperatures on the daily mortality, including the different lag structure as a result of different mechanisms. There are a few studies covering quite long periods e.g. Åström DO, Forsberg B, Edvinsson S, Rocklöv J. Acute fatal effects of short-lasting extreme temperatures in Stockholm, Sweden: evidence across a century of change. Epidemiology. 2013 Nov;24(6):820-9. doi: 10.1097/01.ede.0000434530.62353.0b. PMID: 24051892.
I lack a discussion about harvesting effects, how high mortality among vulnerable (elderly) could be followed by lower mortality and weaker effects of exposure. This effect can be visible also with moderate increases in deaths, see Rocklöv J, Forsberg B, Meister K. Winter mortality modifies the heat-mortality association the following summer. Eur Respir J. 2009 Feb;33(2):245-51. doi: 10.1183/09031936.00037808. Epub 2008 Sep 17. Erratum in: Eur Respir J. 2009 Apr;33(4):947. PMID: 18799511.
In the last paragraphs discussing the attenuated effects of cold, one could also expect vaccination of elderly and lower air pollution levels to modify the relationships.
Reply: Thank you for your comment. It is, unfortunately, not possible to make a “separation between observed seasonal and annual relationships” as we for the period only have access to annual mortality data. The data available from Tabellverket is only showing annual deaths per parish. We will, however, cite the suggested reference by Åström et al. (2013) as well as to the Discussion add a paragraph about the “harvesting effect” among the elderly not at least very well-known from the recent Covid-19 pandemic. We agree that both vaccination of elderly and lower air pollution levels would alter the temperature–mortality relationship but this is hardly relevant for a study only covering the pre-1860 period.
Comment: Statistical methods
It would be interesting to know if the expected number of deaths in the oldest group (65+) would differ much (due to harvesting) if you use a baseline/trend without the last year and mortality from last year as a separate variable, or if annual numbers are too crude.
Reply: While we fully acknowledge that this would be an interesting issue to further investigate, this would both complicate the calculations and make an already long article too long. Thus, we decide that this is an issue to instead further investigate in future studies. We will, however, consider to add one or two sentence about this issue in the Discussion section of the revised article.
Comment: Discussion
The lack of an effect the following year in the oldest group could be a harvesting effect (see above!).
Could cold winter–spring months also have resulted in a lack of wood fuel to heat homes?
Reply: We will, as stated above, add a sentence or two about the possible “harvesting effect” among the elderly in the Discussion section. Regarding a possible lack of wood fuel to heat homes during cold winter–spring months we are not aware of that a lack of fuel in Sweden during this period should have been a major issue although parts of Skåne had a relative lack of wood locally. In the absence of clear evidence for this in the historical literature, we hesitate to include this possibility in any discussion.
Citation: https://doi.org/10.5194/cp-2023-92-AC2
-
AC2: 'Reply on CC1', Fredrik Charpentier Ljungqvist, 21 May 2024
-
RC2: 'Comment on cp-2023-92', Anonymous Referee #2, 30 Apr 2024
This study presents an analysis of the correlation between climate and mortality in Sweden during the period 1749-1859. Through extensive data processing, the authors succeed in showing that cold winter-springs correlate with increased mortality, both in the same year and the following year. In contrast, summer and autumn temperatures and precipitation showed only weak correlations with mortality. Age-wise, children (0-14 years) showed increased mortality in the year following a year of cold winter-spring, while the over-65 age group experienced higher mortality in the same year as a cold winter-spring and, finally, the 15-65 age group had higher mortality both in the same year and in the year following a year of cold winter-spring. Geographically, mortality increased in southern Sweden in the same year as a cold winter-spring, while in east-central Sweden mortality increased mainly in the year following a cold winter-spring. Over time, the correlation between a cold winter-spring and mortality in the following year decreases, while the correlation between a cold winter-spring and mortality in the same year emerges after 1790 but weakens towards the end of the study period.
Of course, a historian is not satisfied with simply reading signals from historical source material as a purely methodological test, but also wants to see if the results “make sense” in the historical and social context. The plausibility of such reasoning is at the same time an indicator of the quality of the correlations demonstrated. The article also discusses why one variable affects the other. Cold winter-springs are said to affect mortality in two ways: firstly, cold winter-springs result in less food production, leading to malnutrition and increased mortality, and secondly, cold weather can increase the risk of disease spread due to overcrowding or reduced resistance to respiratory diseases.
A critical point is that the current discussion about the effects of historical climate conditions on mortality could be clearer and more in-depth. One reason is that, despite the processing of large amounts of data, the evidence does not provide clear-cut answers. One alternative, of course, would have been for the authors to stop at the ambition of establishing various correlations with the study. In this case, the third question about how climate “contribute to" mortality should be changed to “correlate with” as there are several factors that are not corrected for in the analysis. But such a pure methodological test risks being rather bland.
Of the two different links between climate and mortality highlighted by the authors, one of them goes through the intermediary of harvest. This is where the discussion can become clearer. In his thesis from 2006, Daniel Larsson - who, unlike this study, worked with monthly data for mortality - interpreted the increased mortality during the winter-spring as a sign of food shortages due to a poorer harvest the previous year. This seasonal peak in mortality declined over time, which Larsson saw as a direct consequence of an improved supply situation. This study instead looks at annual mortality rates but monthly climate data. Thus, we don't know what time of year people died. However, the increased mortality in the year following a year with a cold winter-spring could reasonably be linked to a poor harvest in between. The fact that the correlation weakens over time could also be interpreted as a general improvement in the livelihood situation, which is exactly what happens during the period.
A crux, however, is that we lack the central intermediate stage of harvest in this study, while much of the reasoning about climate variations and mortality revolves around the nutritional situation. The authors could have been more precise in their discussion if they had included a measure of harvest in the calculations. Yield ratios are difficult to reconstruct and are also subject to uncertainties, but such data series are not lacking in the scientific community. There is a vagueness in the argumentation when they skip one of the more important intermediate stages.
In this context, it is worth noting how literature is used. To build the cold spring-winter - poorer harvest - higher mortality link, two works are highlighted, both of which claim that the temperature during late winter and early spring played “a critical role” in cereal production in south-central Sweden (Edvinsson et al. (2009) and Holopainen et al. (2012)). However, Skoglund's thesis from 2023 shows no correlation between cold winter-springs and harvests in southernmost Sweden, but instead that harvests varied most with summer temperature and precipitation (summer drought). Spring-winter temperature as a factor for harvest was more important in northern Sweden, where the length of the growing season was critical. When different studies seem to achieve contradictory results, there is a risk of “cherry-picking” - choosing the studies that confirm your own results. I do not claim that the authors do this here. However, the heterogeneity of the research literature on the issue does seem to demonstrate how complex the relationship between climate variability and crop yields appears to be. Simple, unambiguous answers are not given. The study might do well to enter more clearly into a dialogue with the various results available in the literature. (In the context of Skoglund (2023), one might also emphasize the need for combined precipitation and temperature data in the calculation model to distinguish a hot but humid summer from a summer with drought).
As for the increased mortality in the same year as a year with a cold winter-spring, the lack of monthly mortality data makes analyses more difficult to make. We cannot know whether the increased mortality occurred at the same time as the cold winter-spring or in the autumn, after the harvest, when mortality could hypothetically be related to the nutritional situation (due to a poor harvest). The authors themselves seem to see this mortality (in the same year as a cold winter-spring) mainly as an effect of the cold weather itself. They mention increased risk of spreading infectious diseases as people spend more time indoors due to the cold and reduced resistance to respiratory infections due to cold weather. I think there should be given more room for an account of the scientific debate and an explanatory model showing that cold weather really leads to disease to support such an interpretation. Some space should also be given to address the geographical pattern in this context. According to the study, populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter-spring. Perhaps an interpretation of the correlation could be more clearly based on this seemingly paradoxical relationship?
One reflection is that in the study of how climatic variations affect mortality, the authors find relatively weak evidence that poor harvests increased mortality. According to the study, the mortality of the adult population seems to be more correlated with cold winter-springs the same year than the year before. It is, if I understand correctly, the mortality of children (in east-central Sweden) that clearly correlates with cold winter-spring in the previous year. One problem here, as mentioned, is that the authors do not measure harvests. This leads to somewhat vague analysis. The relationship between weather and harvest (and mortality) is likely to be more complex than can be captured in a computational model like the one used here. It is not that the results suggest that nutrition was not important, but that this study cannot capture this relationship to more than a small extent.
Furthermore, a limitation of the study is that it works only with correlations and not with regressions, which would have given us an opportunity to assess more clearly the strength of the impact of the reported relationships on mortality. To assess more clearly how much mortality is affected by the factors studied, such as temperature in winter-spring, an appropriate measure could have been to include a scatterplot. By showing the mortality rate against the temperature variation, such a plot would allow a visual assessment of the strength of the relationship.
I also have a comment on Figure 6. One curve shows the correlation over time between cold springs and mortality in the following year (31-year moving window). This curve shows a strong correlation that starts to weaken towards the end of the study period. This is logical. As the livelihood situation in the country improves based on new cultivation and other factors, this category of mortality declines. The second curve shows the relationship between a cold winter-spring and mortality in the same year. This curve shows a very weak correlation until around 1790. After that, the correlation is strong until around 1825 and then weakens. This curve is difficult to interpret. Readers may be suspicious that some strong observations in the early 19th century are behind this stronger correlation. The pattern is also difficult to explain. The authors suggest that this correlation was not visible during the first period studied (1750-1790) due to a generally higher mortality level then. In other words, we would be dealing with a pattern that we can detect statistically only when mortality starts to fall. This could indeed be true. But shouldn't the correlation continuously strengthen in such a case? Why does the correlation become weaker again after about 1825?
All in all, this is a solid and very interesting study, as evidenced by the fact that it gives rise to many thoughts and questions. It is a shortcoming to omit the intermediate factor of harvest, but I still think the authors get away with it based on the tight structure and a certain awareness of the problem. However, there is a lot of noise in a study that skips such a step. But what comes through the noise is still clear. A cold winter-spring is likely to result in high mortality in the following year and implies a poorer nutritional status. The geographical pattern of mortality in the same and following year, respectively, after a cold winter-spring is remarkable and I believe still awaits explanation. The temporal curve of the relationship between mortality and a cold winter-spring in the same year is also still somewhat puzzling to me. In any case, I hope to see the article published. I believe it will generate new studies that will help researchers refine their formulations, approaches and methods.
Citation: https://doi.org/10.5194/cp-2023-92-RC2 -
AC3: 'Reply on RC2', Fredrik Charpentier Ljungqvist, 21 May 2024
Comment: This study presents an analysis of the correlation between climate and mortality in Sweden during the period 1749–1859. Through extensive data processing, the authors succeed in showing that cold winter–springs correlate with increased mortality, both in the same year and the following year. In contrast, summer and autumn temperatures and precipitation showed only weak correlations with mortality. Age-wise, children (0–14 years) showed increased mortality in the year following a year of cold winter–spring, while the over-65 age group experienced higher mortality in the same year as a cold winter–spring and, finally, the 15–65 age group had higher mortality both in the same year and in the year following a year of cold winter–spring. Geographically, mortality increased in southern Sweden in the same year as a cold winter–spring, while in east-central Sweden mortality increased mainly in the year following a cold winter–spring. Over time, the correlation between a cold winter–spring and mortality in the following year decreases, while the correlation between a cold winter–spring and mortality in the same year emerges after 1790 but weakens towards the end of the study period.
Of course, a historian is not satisfied with simply reading signals from historical source material as a purely methodological test, but also wants to see if the results “make sense” in the historical and social context. The plausibility of such reasoning is at the same time an indicator of the quality of the correlations demonstrated. The article also discusses why one variable affects the other. Cold winter–springs are said to affect mortality in two ways: firstly, cold winter–springs result in less food production, leading to malnutrition and increased mortality, and secondly, cold weather can increase the risk of disease spread due to overcrowding or reduced resistance to respiratory diseases.
Reply: Thank you for a very thorough and positive summery of our article. We highly appreciate the long and detailed review of our work.
Comment: A critical point is that the current discussion about the effects of historical climate conditions on mortality could be clearer and more in-depth. One reason is that, despite the processing of large amounts of data, the evidence does not provide clear-cut answers. One alternative, of course, would have been for the authors to stop at the ambition of establishing various correlations with the study. In this case, the third question about how climate “contribute to” mortality should be changed to “correlate with” as there are several factors that are not corrected for in the analysis. But such a pure methodological test risks being rather bland.
Reply: While we acknowledge that our study is not able to provide clear-cut answers with regard to the causality between changes in temperature and changes in mortality patterns, we agree with the reviewer that to only describe our finding in terms of “correlate with” would make the article and its findings too bleak. In the revision, we will thus – in line with the wish of the reviewer – deepen the climate–mortality relationship discussion both in the background section and in the discussion section. In the Introduction section, we will among other things add:
“The total number of deaths follow an annual cycle, with higher mortality during the winter months in the extra-tropical Northern Hemisphere, a pattern also observed in contemporary Sweden (Statistics Sweden, 2020). Many, but not all, causes of deaths display varying seasonal patterns. For example, data from the United States show that pneumonia, influenza, chronic obstructive pulmonary disease (COPD), and other respiratory diseases, peak during the winter months. A similar pattern, though less pronounced, is observed for cardiovascular diseases. Conversely, deaths from traffic accidents peak during summer months, while suicide and cancers show small or no seasonality (Rau et al., 2017). However, the differences vary by region and by baseline mortality level. In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates (Eurowinter Group, 1997).”
Furthermore, we note that it cannot exist any endogeneity in the present case as temperature may well influence mortality whereas mortality cannot influence temperature.
Comment: Of the two different links between climate and mortality highlighted by the authors, one of them goes through the intermediary of harvest. This is where the discussion can become clearer. In his thesis from 2006, Daniel Larsson – who, unlike this study, worked with monthly data for mortality – interpreted the increased mortality during the winter–spring as a sign of food shortages due to a poorer harvest the previous year. This seasonal peak in mortality declined over time, which Larsson saw as a direct consequence of an improved supply situation. This study instead looks at annual mortality rates but monthly climate data. Thus, we don’t know what time of year people died. However, the increased mortality in the year following a year with a cold winter–spring could reasonably be linked to a poor harvest in between. The fact that the correlation weakens over time could also be interpreted as a general improvement in the livelihood situation, which is exactly what happens during the period.
A crux, however, is that we lack the central intermediate stage of harvest in this study, while much of the reasoning about climate variations and mortality revolves around the nutritional situation. The authors could have been more precise in their discussion if they had included a measure of harvest in the calculations. Yield ratios are difficult to reconstruct and are also subject to uncertainties, but such data series are not lacking in the scientific community. There is a vagueness in the argumentation when they skip one of the more important intermediate stages.
Reply: While the reviewer correctly emphasizes the importance of harvest data in order to better understand historical climate–mortality relationships, several issues compromise its reliability and representativeness. Frequently, yield ratios are collected from individual farms, which may not accurately reflect entire regions and tend to be biased towards the most prosperous ones. Tithes are frequently used as an indicator of harvest volume; however, research indicates that in Sweden, tithes captured only 25 to 50 percent of the actual harvests, with some being fixed amounts. Even when tithes vary, their use as an indicator of harvest quality is problematic; they tend to exaggerate fluctuations since, in poor harvest years, tithes payments are drastically reduced. Moreover, harvest data do not necessarily correlate with food availability, particularly in smaller regions which might offset poor local yields by importing grain from elsewhere. Consequently, the link between harvest data and mortality rates is complex and not always direct.
Nevertheless, taken the reviewer’s concern into consideration, we have decided to add two new paragraphs to the discussion section including what the reviewer seems to ask for:
“The following-year mortality response to climate variability is presumed to mainly occur by affecting nutritional situation through the intermediary of harvest outcome and, to a lesser extent, livestock mortality. We acknowledge that the exclusion of harvest data, as a central intermediate stage between climate and mortality, constitute a major limitation in the present study. However, we leave the complex issue of harvest–mortality relationships to further studies due to several issues compromising the reliability and representativeness of 18th and 19th century Swedish harvest data. First, yield ratio series were typically collected from individual farms, which may not accurately reflect entire regions and may not even be locally representative as they tended to be biased towards the most prosperous farms or manors (Slicher van Bath, 1963). Second, tithes are frequently used as an indicator of harvest volume (Leijonhufvud, 2001); however, tithes in Sweden captured only 25 to 50 percent of the actual harvests, with some being fixed amounts, and for the study period only small, and over time decreasing, portions of Sweden still paid taxes in tithes reflecting the actual harvest variations (Hallberg et al., 2016). Third, harvest data do not necessarily translate into actual food availability as grain and other food sources could be, and were, imported from other regions (Edvinsson, 2012), or from aboard (Åmark, 1915), although the presumed climatic effect on following-year mortality reasonably must be interpreted as a result of local to regional harvest outcome.
At present, a reconstruction of Swedish harvests (Edvinsson, 2009) is available for our study period only until 1820 and it has, as a national average, limited explanatory value for studying regional harvest–mortality relationships. This harvest reconstruction, however, indicates an increasing per capita harvest from about 1810, and with setbacks already by the 1790s, signifying increased food security. The per capita harvests, when also including potatoes, were about 15–20% higher around 1820 than they had been prior to around 1790 (Edvinsson, 2009). This agrees well with our finding that the relationship between winter–spring temperature and the following year’s mortality becoming insignificant during the 1820s. Furthermore, we note that this concurs with the findings by (Larsson, 2006). In his work with mortality data from selected parishes, (Larsson, 2006) also interpreted increased mortality during the winter–spring season as indicative of malnutrition due to previous year harvest failures, and he found that these seasonal mortality peaks declining over time as a result on an improved food supply situation.”
Comment: In this context, it is worth noting how literature is used. To build the cold spring–winter – poorer harvest – higher mortality link, two works are highlighted, both of which claim that the temperature during late winter and early spring played “a critical role” in cereal production in south-central Sweden (Edvinsson et al. (2009) and Holopainen et al. (2012)). However, Skoglund’s thesis from 2023 shows no correlation between cold winter–springs and harvests in southernmost Sweden, but instead that harvests varied most with summer temperature and precipitation (summer drought). Spring–winter temperature as a factor for harvest was more important in northern Sweden, where the length of the growing season was critical. When different studies seem to achieve contradictory results, there is a risk of “cherry-picking” – choosing the studies that confirm your own results. I do not claim that the authors do this here. However, the heterogeneity of the research literature on the issue does seem to demonstrate how complex the relationship between climate variability and crop yields appears to be. Simple, unambiguous answers are not given. The study might do well to enter more clearly into a dialogue with the various results available in the literature. (In the context of Skoglund (2023), one might also emphasize the need for combined precipitation and temperature data in the calculation model to distinguish a hot but humid summer from a summer with drought).
Reply: We fully agree with the reviewer that climate–harvest relationships are complex, and often far from clear-cur, but we do not agree that the presentation of the literature about climate–harvest relationships have the character of “cherry-picking”. We are indeed very familiar with the thesis by Martin Skoglund from 2023 and the articles published in it and subsequently by him (the corresponding author was his co-supervisor). His results, based on different climate data (mainly from Lund), showing a relative lack of a winter and spring temperature signal but a considerable spring and summer drought signal, apply to the province of Skåne in southern-most present-day Sweden. Ongoing, still unpublished results, also reveal that the same applies to the province of Halland just northwest of Skåne. However, other research – including assessments conducted by agrarian historian Lotta Leijonhufvud – has shown that the late winter and early spring temperatures indeed plays a major role for harvests in south-central Sweden (around Lake Mälaren) as opposed to in southern-most Sweden (Skåne). Indeed, we think that one very likely reason for the lagged mortality response to late winter–early spring temperature in south-central Sweden, as opposed to the lack of such a lagged response in southern-most Sweden, actual is a direct result of harvest not being sensitive to late winter–early spring temperature in southern-most Sweden while they are more sensitive to it in south-central Sweden (with a shorter growing season making the timing of its onset more critical). We will in the revision of the article further clarify this geographical difference as it appears from the reviewer’s comment that this was not written clearly enough.
Comment: As for the increased mortality in the same year as a year with a cold winter–spring, the lack of monthly mortality data makes analyses more difficult to make. We cannot know whether the increased mortality occurred at the same time as the cold winter–spring or in the autumn, after the harvest, when mortality could hypothetically be related to the nutritional situation (due to a poor harvest). The authors themselves seem to see this mortality (in the same year as a cold winter–spring) mainly as an effect of the cold weather itself. They mention increased risk of spreading infectious diseases as people spend more time indoors due to the cold and reduced resistance to respiratory infections due to cold weather. I think there should be given more room for an account of the scientific debate and an explanatory model showing that cold weather really leads to disease to support such an interpretation. Some space should also be given to address the geographical pattern in this context. According to the study, populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter–spring. Perhaps an interpretation of the correlation could be more clearly based on this seemingly paradoxical relationship?
Reply: The lack of monthly mortality data indeed poses, as we have outlined in our response to reviewer #1, a challenge for our study. Currently there are no monthly national series for Sweden for the 18th century and if, for example, the impact of harvests was larger in earlier periods, deaths may have spread more during the whole year. We will in the revised article added an entire new paragraph in the Introduction section to provide more information on the annual cycle of mortality. We have also decided, in the revision, to add a short commentary on this in the presentation of the data:
“Annual mortality data may not capture the seasonal nuances of mortality rates in temperate climates, where peak mortality typically occurs during the winter months, roughly from November to March. This limitation makes it challenging to attribute anomalously high mortality rates during a specific year solely to the effects of exceptionally cold conditions during certain months (e.g., January–February or November–December). Despite this limitation, our study provides valuable insights into the potential connection between climate and mortality. If adverse climatic conditions have an immediate effect on mortality, we would expect correlations without time lags in our annual data. However, even in annual data, correlations with time lags may emerge if there is a delayed effect between climate conditions and mortality. Future research using more granular monthly mortality data could offer a more detailed understanding of these dynamics.”
Regarding the fact that populations in the areas with the mildest winters were most at risk of dying due to colder weather in winter–spring we indeed think that the higher vulnerability to colder winter and early springs temperatures in regions with generally warmer winters and early springs can be related to poorer housing and heating facilities in these regions (e.g., in Skåne). We also now mention this in the Introduction that such factors play into mortality in contemporary times: “In regions with warmer winters and cooler homes, a given rate of change in winter temperature will have a larger impact on death rates (Eurowinter Group, 1997).” Regions with generally mild winters paradoxically also today tend to be those with the highest vulnerability to prolonged cold snaps. We will add an extra new sentence about this in the Discussion section in the revised manuscript.
Comment: One reflection is that in the study of how climatic variations affect mortality, the authors find relatively weak evidence that poor harvests increased mortality. According to the study, the mortality of the adult population seems to be more correlated with cold winter–springs the same year than the year before. It is, if I understand correctly, the mortality of children (in east-central Sweden) that clearly correlates with cold winter–spring in the previous year. One problem here, as mentioned, is that the authors do not measure harvests. This leads to somewhat vague analysis. The relationship between weather and harvest (and mortality) is likely to be more complex than can be captured in a computational model like the one used here. It is not that the results suggest that nutrition was not important, but that this study cannot capture this relationship to more than a small extent.
Reply: See our reply above with regard to harvests.
Comment: Furthermore, a limitation of the study is that it works only with correlations and not with regressions, which would have given us an opportunity to assess more clearly the strength of the impact of the reported relationships on mortality. To assess more clearly how much mortality is affected by the factors studied, such as temperature in winter–spring, an appropriate measure could have been to include a scatterplot. By showing the mortality rate against the temperature variation, such a plot would allow a visual assessment of the strength of the relationship.
Reply: This study primarily examines the correlations between mortality and climate (mainly temperature), rather than their direct impacts, which are more challenging to ascertain. The correlations identified are not particularly strong (only about 10% explained variance) – suggesting that relying solely on regression coefficients could be misleading. About 90% of the mortality variations are not related to climate (at least not to the climate variables under investigation here) and regressions could lead to uncertain or biased results we they majority of the mortality variations are related to factors not under investigation. While correlations can suggest certain trends, any regression analysis must also take into account relevant factors explaining more of the mortality variations to provide a more comprehensive understanding. This is not feasible in this case.
We are unsure whether it is really necessary to provide scatter-plots in the article and would like to leave it up to the editor to decide that considering that our article already is rather extensive and contains numerous figures. However, we have now created two scatter-plots as the reviewer wishes: one for spring temperature and mortality the same year and one for spring temperature and mortality the following year.
Comment: I also have a comment on Figure 6. One curve shows the correlation over time between cold springs and mortality in the following year (31-year moving window). This curve shows a strong correlation that starts to weaken towards the end of the study period. This is logical. As the livelihood situation in the country improves based on new cultivation and other factors, this category of mortality declines. The second curve shows the relationship between a cold winter–spring and mortality in the same year. This curve shows a very weak correlation until around 1790. After that, the correlation is strong until around 1825 and then weakens. This curve is difficult to interpret. Readers may be suspicious that some strong observations in the early 19th century are behind this stronger correlation. The pattern is also difficult to explain. The authors suggest that this correlation was not visible during the first period studied (1750–1790) due to a generally higher mortality level then. In other words, we would be dealing with a pattern that we can detect statistically only when mortality starts to fall. This could indeed be true. But shouldn’t the correlation continuously strengthen in such a case? Why does the correlation become weaker again after about 1825?
Reply: We agree with the reviewer that the interpretation of Figure 6 is far from clear-cut and already discusses this in the article. The most likely explanation, in our opinion, behind the weakening relationship between late winter–early spring temperature and same-year mortality after about 1825 is generally improved living condition, including nutrition, during the course of the 19th century. The general mortality level (mainly caused by infectious diseases) decreases during the 19th century along with reduced inter-annual fluctuations in death rates (as well as in birth rates). For example, smallpox becomes virtually extricated (with mandatory smallpox vaccination from 1816). We will add a sentence or two about this in the revised article. Furthermore, we will add the following clarification in the Conclusion section of the revised manuscript: “This suggests that in central and northern Sweden, the influence of temperature on mortality was primarily attributed to the adverse effects of cold weather on harvests, while this was not the case in southern Sweden. To establish causation definitively, future studies should incorporate data on both regional harvests and monthly mortality.”
Comment: All in all, this is a solid and very interesting study, as evidenced by the fact that it gives rise to many thoughts and questions. It is a shortcoming to omit the intermediate factor of harvest, but I still think the authors get away with it based on the tight structure and a certain awareness of the problem. However, there is a lot of noise in a study that skips such a step. But what comes through the noise is still clear. A cold winter–spring is likely to result in high mortality in the following year and implies a poorer nutritional status. The geographical pattern of mortality in the same and following year, respectively, after a cold winter–spring is remarkable and I believe still awaits explanation. The temporal curve of the relationship between mortality and a cold winter–spring in the same year is also still somewhat puzzling to me. In any case, I hope to see the article published. I believe it will generate new studies that will help researchers refine their formulations, approaches and methods.
Reply: Thank you again for an accurate summery of our work and for the encouraging words. We appreciate not only that the reviewer wants to see our article published, but also that he or she agree with us that additional research is needed to better constrain and understand the causal links between changes in climate variables and changes in mortality.
Citation: https://doi.org/10.5194/cp-2023-92-AC3
-
AC3: 'Reply on RC2', Fredrik Charpentier Ljungqvist, 21 May 2024
-
AC4: 'Comment on cp-2023-92', Fredrik Charpentier Ljungqvist, 21 Jul 2024
We have now provided detailed responses to all the reviewer’s comments on our manuscript individually since some time ago. As far as we can deem, we have answered each of the reviewers’ suggestions in detail. We have already almost finished the working with revising the article according to the reviewer’s suggestions.
Citation: https://doi.org/10.5194/cp-2023-92-AC4
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
574 | 113 | 52 | 739 | 36 | 32 |
- HTML: 574
- PDF: 113
- XML: 52
- Total: 739
- BibTeX: 36
- EndNote: 32
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1