| 28 Jul 2026
Climate Variability, Grain Supply, and Food Security in Ancient Rome
Abstract. The Roman world has become a central case in debates on past climate–society interactions, particularly regarding the empire's expansion, transformation, and decline. However, despite growing recognition that these interactions were mediated through complex socio-economic and ecological processes, the mechanisms through which climatic variability translates into historical outcomes are rarely investigated explicitly. In this paper, we investigate whether climatic variability was systematically associated with historically attested food shortages in ancient Rome, explicitly accounting for changes in the geographical configuration of the city's provisioning system. Combining annual palaeoclimatic reconstructions from climate field reconstructions (PHYDA and GEDA) with a dataset of 55 attested shortage events for the period 1–650 CE, we analyse temperature and hydroclimatic variability across Rome’s principal grain-supplying regions. The analysis combines probit regression models with Monte Carlo simulations to test associations with both short-term climatic shocks and sustained climatic anomalies. The results provide no evidence for a stable, system-wide climate–shortage relationship. Throughout the Imperial and Late Antique periods, statistical associations are weak and regionally variable, suggesting a provisioning system that buffered regional climatic variability. A clearer, albeit modest, climatic signal emerges only after the contraction of Rome’s supply network in the late fifth century CE. Overall, the findings indicate that, based on historically attested food shortages, climatic variability influenced Rome's food security not as a direct or uniform trigger, but as a conditional stressor whose effects depended on the changing structure of Rome's provisioning system.
Review of the manuscript
Climate Variability, Grain Supply, and Food Security in Ancient Rome
by Devi Taelman and co-authors
submitted for publication in Climate of the Past
General:
The manuscript investigates the climatic impact on the food supply of Rome in three different periods of the first millennium AD and challenges the question how strong this impact was. An important result is that neither hydrological nor temperature anomalies are reliable predictors for the food shortages in ancient Rome. Instead authors argue that climate influenced food shortages not in a direct way, but mediated and modified through the respective temporal and spatial state of the provisioning system.
The authors use state-of-the-art methodological tools and data sets to setup their analysis. In addition, the robustness of results is evaluated based on uncertainties included in the underlying datasets. Results are critically assessed and also put into an historical context, taking into account different societal subsystems (e.g. agrarian, economical, fiscal and logistical).
In addition, the manuscript is well written and shows an understandable structure. Below I added some suggestions, specifically related to improve the introduction and taking into account serial auto-correlation for the assessment of the robustness of the results.
Specific:
1 Introduction:
The introduction is quite short and should elaborate in greater detail the major background settings (social and climatological) for the focus region, i.e. The Mediterranean area. For instance, authors could already use their separation into different sub-periods and present the most important socio-economic and large-scale climatological settings for their subsequent investigations.
2 Rome’s Grain Supply and Agroclimatic Context
In this chapter a geographical map would be very helpful to showcase different source areas of grain supply and potential changes in the different periods. This could also include the different climatological background settings with different harvesting times mentioned in the text section.
3 Data and Methods
3.2 Palaeoclimate Reconstruction Data
Authors mention a number of studies they use for the climatological analysis. For continental Europe there also exist another study based on summer (JJA) temperatures on a 5x5 deg, including uncertainty estimates (Luterbacher et al., 2016). This study might be an independent source for comparison and the background climatic conditions.
3.3 Conceptual Framework and Analytical Strategy
In none of the two chapters on the Probit and Monte Carlo Analysis basic references for the methodology are presented. The authors should include information (links/references to software) they used for the respective analysis in order to allow access to the respective information and resources.
l. 215: Can the authors state why they chose the respective dates for the separation of the three different periods? Is there any objective basis supporting the respective dates?
3.4 Sensitivity Analyses
For those variants of the Analyses using temporally filtered data for the models, authors should keep in mind to evaluate the amount of serial correlation introduced by filtering that is changing the nominal level of significance. Filtering using only a very low number of years I would not expect large impacts concerning the effective number of degrees of freedom, but decadal-to-multidecadal filtering of data prior to the analyses might substantially impact on the statistical significance taking into account the effective number of degrees of freedom, changed by the strength in serial autocorrelation (cf. Ebisuzaki, 1997).
4 Statistical Assessment of Climate–Shortage Associations
4.1 Main Findings
l. 315: The spatial difference between Sardinia and Sicily is only a few 100s of km. Still, substantial differences are found based on the respective statistical analysis. How do the authors explain the large discrepancy in this comparatively small common area?
l. 349: what could be the reasons (economically/socially) rendering the period 451–650 CE being more susceptible to climatic changes compared to the other ones ? In these any reason why stands Sicily out that only for this region the two methods produce consistent results ?
4.2 Robustness of the Main Findings
To test the robustness of results I suggest again to the effect of filtering on the time series in the two models where temporal filtering is applied, i.e. taking into account serial auto-correlation and its effect on effective number of degrees of freedom and hence potential changes in the statistical significance.
5 Climate and Resilience of Rome’s Food Supply System
Can the authors also make statements on the impacts of changes in natural external forcings, e.g. Volcanism. Although compared to the 2nd half of the last millennium the volcanic activity was considerably muted, it might still worth to look into the volcanic time series if there might be some indications for any volcanic impact on the food supply in single years, e.g. Vesuvius 79 AD or some other Tropical and/or Northern Hemispheric eruptions in the first millennium AD. Here authors could consult volcanic time series based on the studies of Toohey and Sigl (2017) and a more recent dataset of Sigl and Toohey (2024).
6 Conclusions
Although the main investigations focus on the situation in ancient Rome, authors could provide a few words linking past conditions the present-day societies in terms of similar or changed susceptibility caused climatic disruptions on food supply in the study region.
Figures and Tables:
In general, all labels for axes and headings in the Figures should be increased in font-size.
Fig. 2: The y-axis could be better tailored to the min/max values of the filtered TA and scPDSI information).
Fig 3: The size of the text boxes size and font size should be increased for better readability.
Fig. 4: The general scope of the Figure is fine, but also here the font size of numbers and axes labels. could be optimized together with an increased colorbar (cf. also Fig. B1, B3, C1 and C3).
Table 1: The regions could also be represented in a geographical map indicating the respective regions in a spatial context. Then the table with the exact locations might be moved to the Appendix.
Additional References:
Ebisuzaki, W.: A Method to Estimate the Statistical Significance of a Correlation When the Data Are Serially Correlated. Journal of Climate, 10, 2147–2153.
Luterbacher, J., et al.: (2016): European summer temperatures since Roman times, Environmental Research Letters, 11, doi:10.1088/1748-9326/11/2/024001.
Sigl, M. and M. Toohey (2024): Volcanic stratospheric sulfur injections from 500 BCE to 1900 CE, eVolv2k_version4 [dataset]. PANGAEA, https://doi.org/10.1594/PANGAEA.971968
Toohey, M. and M. Sigl: (2017) Volcanic stratospheric sulfur injections and aerosol optical depth from 500 BCE to 1900 CE, Earth System Science Data, 9, 809–831, https://doi.org/10.5194/essd-9-809-2017.