Palaeoenvironmental Data from a Historian’s Perspective

The natural sciences are highly diverse and pluralistic. There is no single palaeoscience (a discipline or branch of the natural sciences that would deal with the past in toto), nor is there a single interdisciplinary environmental history. Most historians who interact with the earth sciences either specialize in working with a single branch, gradually acquiring the expertise to critically engage with the evidence it produces (these are “interdisciplinary historians” in a strict sense), or they draw on specific studies relevant to a particular historical inquiry, which they use to place their research questions in a broader context (we might call these scholars “open-minded” historians, and say they have the habit of “reading widely”). Few members of either group are aware of just how diverse, rich, or heterogeneous the world of the earth sciences—and, by extension, the palaeosciences—actually is.

Of the methodologies at the interface of history and the natural sciences, the study of past climate change has probably the longest research tradition.Footnote ⁸ In the context of modern historiography, climate change is a general notion that translates into a wide range of processes and phenomena occurring at different spatial and temporal scales. These can involve extreme rainfall events taking place on short timescales such as hours or days, droughts that persist for weeks or even years, or abnormal weather conditions over a full season or series of consecutive years (such as a decade of weaker monsoons in tropical zones following a strong volcanic eruption). They also include longer-term trends in climate conditions over a given area, occurring at multi-decadal or even centennial scales (for instance, drying or cooling trends caused by external forcing or internal variations of the climate system). The variety of weather and climate phenomena makes clear that no easy generalization can be made about climate’s role in history. Depending on the type of climate change we are concerned with, its interaction with any particular society (and related ecosystems, understood together as a holistic social-ecological system) will differ according to what weather or climate phenomena actually occurred—or, to be more precise, what our interpretation of the palaeoclimatic data suggests occurred. And this, of course, comes on top of the many intricacies of the specific cultural, economic, and political systems which conditioned how that society would or could respond to said climate change.Footnote ⁹

Past climate variability has left traces in both human and natural records, which can take the form of lake and marine sediments, peat deposits, speleothems in caves, living and dead trees, ice layers, and so on. When properly interrogated, each of these archives can return a wide range of proxies, that is, measured physical or chemical indicators that approximate past environmental conditions, though they do not reflect any parameters in a direct, one-to-one relationship. As the field of palaeoclimatology has developed and matured, the proxies used have become increasingly numerous and diverse. Since these archives were created through a variety of (natural) environmental processes, they require different methods of extraction and analysis. To make sense of an archive and establish a proxy, one may need to recognize the morphology of plant parts or animals, measure inorganic compounds and element concentrations, or undertake isotopic measurements of organic and inorganic materials. Many different scientific fields are involved, each requiring a high degree of specialization.

Climate reconstructions based on documentary and natural proxies contain several uncertainties, though this does not undermine their reliability. An important aspect of working with these uncertainties is proxy calibration: as we cannot measure past climate directly, we only measure extant organic and inorganic matter that result from complex processes influenced by past climate variables or other meteorological phenomena. We calibrate the proxy data—collected in the field and measured and analyzed in the laboratory—by comparing them with modern observations from meteorological instruments, usually encompassing the nineteenth and twentieth centuries. This enables us to make temperature or precipitation estimates for specific parts of a given year. In some cases, the proxies reflect a combination of temperature and precipitation and can provide more generalized information about winter conditions or droughts. Even more significant is chronological uncertainty, which will be discussed below. The overall combination of intrinsic dating uncertainty and the margins of error involved in the process of palaeoclimatic reconstruction (proxy calibration) ultimately contributes to a kind of composite or multifarious uncertainty, which a historian using this data must navigate.

Natural scientific studies of past landscapes—palaeoecology—share many challenges with palaeoclimatology, especially when it comes to chronology. Palaeoecology or palaeoenvironmental science also employs highly heterogeneous approaches to reconstruct the past. In the majority of cases, palaeoenvironmental researchers work on sediment cores, that is, long columns of organic and inorganic matter (or “mud”) extracted from locations such as lake beds or peat bogs that favor its accumulation layer after layer, and which ensure the preservation of organic material over long periods. These cores can be analyzed using various methods, beginning with sediment geochemistry to establish local erosion patterns and the physical-biological processes behind the production and accumulation of the sediments. Major shifts in such parameters may reflect important transformations of a local ecosystem or even an entire landscape at particular moments in the past, most often linked to human interventions. To understand these interventions in more detail, we could, for instance, analyze the pollen grains preserved in the sediments, produced by plants that lived around our sedimentary basins (study locations), and thus reconstruct changes in cultivation and agriculture. Even though this method, known as palynology, allows for a high degree of precision in identifying specific plant species or families and establishing changes in their presence over time, it likewise remains only an approximation. We are able to reconstruct relative changes, but we are often unsure of the exact spatial signature of vegetation dynamics: while the pollen of certain plants such as vines or cereals is dispersed in a relatively local radius, that of others such as olive and pine trees travels over long distances, often confounding local and regional signals in a single dataset.Footnote ¹⁰ A variety of other techniques allow for the evaluation of human impact on landscapes, notably the analysis of an array of biomarkers—specific organic compounds reflecting the presence of plants or animals in the past, such as beta-stanols from the excrement of omnivores, usually pigs or humans.Footnote ¹¹ None of these markers allows for a direct measurement or reconstruction of past ecosystems or human activities in the landscape, but together they permit reliable approximations that historians and archaeologists can further contextualize with the help of written and material sources.

The stories we tell about the past are located in specific places and relate to specific moments in time. Historians use time and space to structure narratives—we study a specific problem in a particular when and where. To a large extent, it is also time and space that define specializations and subdisciplines within academic history: most historians are specialists in a particular time period in certain geographical areas. This is far less often the case with natural scientists, whose specialization is frequently method-based. Not surprisingly, historians exploring the potential for integrating natural scientific data and insights into their research usually begin by surveying the data available within a specific spatiotemporal framework. Often, we look for overlaps between these data and a given research topic and start building new histories once these overlaps are found. Reliable approximations of time and space thus become crucial for integrating diverse lines of evidence from often radically different disciplines. Before we consider concrete examples of different study designs, methodological contexts in which humanistic and natural scientific communities interact and in which the datasets they produce become entangled, we must first focus on how spatial and temporal scales are conceived and approximated in the natural scientific disciplines that most frequently interact with Late Holocene history.

The Epistemological Challenges of Using Proxy Data: Climate Phenomena and Climatic Proxies

Climate-related phenomena can be local, regional, hemispheric, or global, and thus define the spatial relevance of proxy records. As a consequence, when we examine the spatial scales of historical climate change we are often faced with a perplexing diversity. To begin with, we have relatively rare global or hemisphere-wide phenomena, often associated with specific types of forcing such as strong tropical volcanic eruptions or El Niño–Southern Oscillation (ENSO). These, in turn, translate differently into regional climates, prompting for instance shifts in seasonal rainfall patterns on a semi-continental scale, related to monsoons or surface temperature changes in the oceans. Finally, due to microclimatic and microregional variation, the same hemispheric or even regional climatic phenomena can have different—even opposite—local effects, depending on a wide range of factors that might include land-sea interactions, vegetation, and land morphology. Trends in temperature variability are more likely to be homogeneous over larger areas (often on a continental or sub-continental scale), whereas hydroclimate (rain, snow, actual soil-level humidity, etc.) can vary significantly on regional and local scales.

In order to appreciate the spatial issues involved in proxy interpretation, let us look more closely at the proxy of (mainly) winter hydroclimatic conditions at Lake Nar in central Anatolia. As mentioned above, modern meteorological data (measured by instruments) are used to understand which climate phenomena are reflected in particular proxies. These modern data can also be used to understand the spatial relevance of the proxy, or which areas it actually “represents.” Figures 1 and 2 show statistically significant spatial correlations of autumn-winter precipitation conditions from Lake Nar and the surrounding area on two timescales: thirty and sixty years. As we can see, on a shorter timescale of roughly one human generation, or thirty years (fig. 1), the Lake Nar proxy record would be relevant mostly for central Anatolia (ca. 0.8 spatial Spearman correlationFootnote ¹²), less so for the rest of Anatolia (ca. 0.6 correlation), and even less for Bulgaria, Romania, and south-western Ukraine, where only half of autumn-winter precipitation would follow a similar pattern to the Lake Nar area. If we move to a longer timescale of sixty years (fig. 2), spatial correlation values over areas that are not in the immediate vicinity (that is, outside central Anatolia) decrease, but a larger area is covered by weaker but still statistically significant correlations: this now extends to southern Greece, as well as the northern parts of Iraq and the Levant. This example clearly demonstrates the need to consider spatial as well as temporal scales when using a proxy in historical analysis. Lake Nar can definitely be used to write a climate history of central Anatolia, but the trends it shows should not be used for other parts of the Mediterranean unless they are corroborated by other proxies; we should also remember that only some of the longer-term trends are shared between proxies (and this, too, is inconsistent over timeFootnote ¹³). There is thus no single Mediterranean or even eastern Mediterranean history of climate.Footnote ¹⁴ The temporal precision of the proxy, in turn, depends not on the specifics of the climate phenomenon the proxy represents, but first and foremost on the natural archive from which it was derived, whether a lake bed, peat bog, cave stalagmite, tree, or something else. In some cases, palaeoclimatologists can develop annually resolved climate reconstructions (for instance, based on tree rings), but usually they deal with multi-decadal or centennial resolution in their datasets. The distance between individual data points in a proxy can thus be a single year, decades, or even centuries—and these chronologies can be characterized by a degree of uncertainty that often ranges from a year or two to several decades.

Figure 1. Spatial Spearman correlation of autumn-winter precipitation between Lake Nar and surrounding areas, 1981–2010

Figure 2. Spatial Spearman correlation of autumn-winter precipitation between Lake Nar and surrounding areas, 1951–2010

Note: in both figures precipitation time series are detrended by removing the long-term linear trend.

This has to do with the radiometric dating methods, such as ¹⁴C (Carbon-14) or Uranium-series dating, that are widely employed for developing proxy chronologies. Scientists measure the radiometric properties of selected samples from different core or speleothem depths, which, due to the sediment accumulation process, represent different “dates,” or locations in time. They then extrapolate from those measurements to develop an age-depth model that can be used to provide all other samples with age estimates (which are not dates as such). Radiometric methods, like any measurement, have their associated margins of error. Even if continued technological advances mean that these can be relatively small, not all material is suitable for dating and the recovery of good samples is random: some parts of the core end up being better dated than others. The chronological uncertainty associated with radiometric methods is further compounded by the need to perform a calibration process. In the case of ¹⁴C dating, this calibration relies on a comparison of radiocarbon measurements with known calendar dates, such as those obtained through dendroclimatologicalFootnote ¹⁵ methods (fig. 3). The process of calibrating a radiocarbon measurement will result in a chronological uncertainty that can vary considerably with temporal segments within the calibration data (for instance, uncertainties can be larger for one half-millennium and smaller for another). All this means that, in the majority of cases, when using palaeoenvironmental data one operates at the level of decades, half-centuries, or even centuries, rather than the annual or monthly scales of histoire événementielle.

Figure 3 Example of radiocarbon calibration from Lake Engir in Anatolia (measured “raw” ¹⁴C age of 1320±30)

Note: the vertical axis shows the radiocarbon measurement (yellow line) with its theoretical probability distribution (red = one standard deviation, 67% probability; light red = two standard deviations, 95% probability). The blue line is the northern hemisphere radiocarbon calibration curve. The horizontal axis shows the calibrated probability distribution (black = one standard deviation, 67% probability; gray = two standard deviations, 95% probability).

Source: prepared with the OxCal v4.3.2, Christopher Bronk Ramsey, “Methods for Summarizing Radiocarbon Datasets,” Radiocarbon 59, no. 6 (2017): 1809–33; IntCal13 atmospheric calibration curve, Paula J. Reimer et al., “IntCal13 and Marine13 Radiocarbon Age Calibration Curves 0–50,000 Years cal BP,” Radiocarbon 55, no. 4 (2013): 1869–87.

Another example can help explain these complex problems. Let us look again at central Anatolia, this time at Lake Engir, not far from Lake Nar. Sediments from both lakes have been subjected to pollen analysis, and both show that mixed-farming agriculture (combining crops and livestock) collapsed in Cappadocia at the end of antiquity. The problem is the precise date of this change. While Lake Nar’s varved sediments mean that it can be pinpointed to the 660s CE,Footnote ¹⁶ the chronology of Lake Engir is based on two radiocarbon dates: 1320 ± 30 years Before Present (BP) at 114.5 cm and 1540 ± 30 years BP at 175 cm.Footnote ¹⁷ The change in the sediments occurs at the depth horizon of 137 centimeters; it is therefore crucial to correctly estimate the age of the sample taken from the 137th centimeter of the core. With two radiocarbon dates, this can be done using two age-depth models: linear regression and linear interpolation. Technically, linear interpolation is a stepwise linear regression that involves interpolating estimated ages between ¹⁴C dates and the surface (i.e., the date in the early twenty-first century at which the core was taken), with different linear functions for each of the two sections of the core—that is, between the two dates and between the second date and the surface (fig. 4). Estimating sample age based on an age-depth model is a matter of probability: we do not obtain precise data but rather a distribution of probability over time, which shows where the actual (unknown) date most probably falls. Thus, we see that for the sample from the 137th centimeter the linear regression model (red) “spreads” this probability over three hundred years, while the linear interpolation (blue) restricts the distribution more or less to the seventh century, pointing to its earlier part. The choice of model, in the end, depends on mathematical and geological arguments (and more complex models are usually better), though here both models show it is highly probable that the agricultural collapse at Engir was roughly synchronous with the similar change at Nar. It should be emphasized that the Lake Engir model is a simple one—such models can contain dozens of radiocarbon dates and might involve major breaks or irregularities in the sedimentation process, significantly complicating the age-estimation process.

Figure 4. Age-depth model of Lake Engir sediment core and the estimated age probability distribution for the depth of 137 cm

Note: the model represented in blue shows linear interpolation between data points; that in red shows linear regression and the estimated age probability distribution for the depth of 137 cm (the sample showing the transition between late antique agriculture and early medieval land abandonment). In the top panel, the green lines represent the two radiocarbon measurements with their probability distribution, plus the surface of the core (early twenty-first century). The blue and red lines show the best single age estimate for each core depth in each age-depth model; the light blue and red areas show the 95% probability distribution of the age estimate in each age-depth model. The lower panels show what the probability distributions for a specific section of the core depth (137 cm) actually look like in detail, depending on the model (blue = linear interpolation; red = linear regression).

Source: graphics executed in the age-modeling code CLAM: Maarten Blaauw, “Methods and Code for ‘Classical’ Age-Modeling of Radiocarbon Sequences,” Quaternary Geochronology 5 (2010): 512–18. The radiocarbon dates come from Çetin Şenkul et al., “Late Holocene Environmental Changes in the Vicinity of Kültepe (Kayseri), Central Anatolia, Turkey,” Quaternary International 486 (2018): 107–15.

If we return to the historiography, we can place this problem in the context of how recent generations of historians have thought about time. It is evident that the change the historical discipline is currently undergoing, together with recent advances in the palaeosciences, can also be understood as a redefinition of the Braudelian structure of historical time. The (sub)centennial and regional scale of much of the palaeoenvironmental data is exactly the medium-term level of change at which Fernand Braudel located the social and the economic.Footnote ¹⁸ With the increasing availability of palaeoenvironmental data, it has also become possible to translate environmental phenomena from Braudel’s deep level of the longue durée, where no humanly perceptible change occurs, to the level of the medium-term, where “common destinies” and “trends”—including climatic trends, or conjonctures climatiques, as Emmanuel Le Roy Ladurie would call themFootnote ¹⁹—occur, with changes taking place at a relatively fast pace from one generation to the next or even quicker. This does not mean that there are no long-term processes in the Braudelian sense, nor that “environmental” histories occur only at medium-term timescales. Rather, the processes are defined by their pace and not by their specificity: the environmental accompanies the social and the economic, as well as the political and the cultural, at every timescale on which the phenomena we study actually occur.

Along these lines, some participants in more recent debates on the longue durée have suggested that to work properly with longer-term perspectives in historical studies it is necessary to deconstruct or break down the long-term into finer temporal scales. This should, they argue, make it possible to understand the social processes that form the basis of centennial trends.Footnote ²⁰ Unsurprisingly, the French historians of the Annales school proposed similar strategies three decades ago as they attempted to resolve the deadlocks of quantitative history: their most common solution was to understand the specific events of microhistory as models for larger-scale phenomena,Footnote ²¹ so that a successful study of quantitative trends would necessarily combine the longer-term approach with a more microhistorical, or at least higher-resolution and qualitative, historical method.Footnote ²² With the current explosion of palaeoenvironmental data, historians are not limited to combining quantitative and narrative approaches from within the historical discipline itself, but can also integrate evidence from outside the humanities and social sciences into their work. Combining different types of environmental proxies with archaeological and textual evidence makes it possible to narrow the range of hermeneutic possibilities and limit the extent of potential interpretations, not least when it comes to the temporal and spatial scale of the phenomena we study.

Historians working at the crossroads of history and archaeology or geology were among the first to confront this challenge. Twenty years ago, for instance, the landscape historian Gérard Chouquer argued that instead of seeing different disciplines as supplying information on different “layers” of temporality—from the deep(er) past of geologists to the contemporary human-natural space of geographers—scholars should work together, with the resources at their disposal, to address phenomena of interest along spatiotemporal scales relevant for a particular study.Footnote ²³ Indeed, the progress that palaeoscientific approaches have achieved in the last decade, both in the range of phenomena they can reconstruct and in the temporal accuracy of their methods, means that many problems faced by earlier generations of historians who pioneered collaboration with the earth sciences are gradually disappearing. As the earth sciences increase their focus on the origins of the Anthropocene, there is justified hope that even better synergy will emerge between scholars from the humanities and those from the earth sciences.

What follows will present examples of interdisciplinarity in practice. We have decided to focus on three groups of studies that represent what we consider to be the most common models currently linking historical research with natural scientific data in the field of environmental history: broad-stroke survey-type studies, more focused quantitative studies, and interdisciplinary studies supporting new readings of textual sources.

Survey-Type Studies: Bringing Together Disparate Datasets

Survey-type studies were the earliest examples of interdisciplinary environmental history. They can be described as attempts to lend historical context to natural scientific data, sometimes created much earlier or over a long period, often independently from traditional historical questions such as the fall of the Roman Empire or the Thirty Years’ War. In other words, historians work with the scientists who produced the data and seek to provide them a posteriori with historical meaning and significance. Over the last two decades, most such studies have focused on past climate.Footnote ²⁴ This early model of survey-type studies often included a delicate yet important marketing aspect, with the historical dimension being used mostly to broaden the application of an otherwise typical natural science project and attract publicity. These studies should therefore be understood not simply as historians using scientific “products,” but also as the sciences (or rather, natural scientists) using history to validate their claims and infuse their findings with additional significance.Footnote ²⁵ It is vital to mention this question of instrumentalization early on in our discussion, because considerable tension and even power struggles can be threaded through interdisciplinary projects and the use and interpretation of scientific data. When engaging with the palaeosciences, historians can often be drawn into debates with which they are only partially familiar, if at all. This may lead to the overemphasizing of a specific dataset’s historical significance or the disregarding of source criticism and the limits of plausible historical interpretation. History can be misused and the work of historians misinterpreted.Footnote ²⁶ At the same time, the opposite can also occur: scientists can unknowingly intervene and even take a side in a subject of considerable historiographical debate, giving one interpretation the advantage of being supported by the “objective” natural scientific data.

A recent example of this approach reveals both its weaknesses and its strengths.Footnote ²⁷ Importantly, the fact that many of the present article’s authors were also involved in this 2016 publication means that we can analyze it from a critical angle. Sicily, where the study was focused, is a particularly fertile terrain for exploring the relationships between climate and society in the preindustrial era, on the one hand, and the historical and environmental data, on the other. The island has an abundance of human records, both written and material, and is situated in a region with a heightened vulnerability to contemporary climatic change. There is thus a wealth of data—produced over a number of years by different natural scientific, historical, and archaeological projects—that can be pulled together to create a new narrative of its environmental history over the longue durée.

Throughout the first millennium CE, Sicily was a unique place. Part of the Roman Empire from the late third century BCE, at the end of Mediterranean antiquity it became a province of the Western Roman Empire. When that collapsed in the fifth century, after an Ostrogothic interlude the island was integrated into the Eastern Roman Empire. Sicily thus remained under imperial Roman control until the Arab conquest of the ninth century. During that time, it experienced significant long-term changes in climate, seemingly extensive enough to have altered the agricultural productivity of the island. The study discussed here was carried out by a team of natural and social scientists, focusing on the interplay of climatic shifts and social-ecological processes in late antique Sicily.Footnote ²⁸ By considering the methodological aspects of their work, we hope to show in more detail how historians and natural scientists combine different types of data to develop potential scenarios of social-ecological change and how they typically evaluate their plausibility in a survey-type study.

The point of departure for this interdisciplinary research was a multi-proxy study of a sediment core from Lake Pergusa, located in the central east of the island.Footnote ²⁹ A single sediment core, or “mud column,” was retrieved from the bottom of the lake and sampled according to several analytical techniques. Applying different scientific approaches to the same core ensured that there would be no uncertainty regarding the chronological relationship between the results produced. The sediments were analyzed in particular for pollen—used to reconstruct past vegetation cover and agricultural history—and oxygen isotopes (δ¹⁸O)—used to reconstruct winter rainfall, as in the context of Lake Pergusa the proportion of different oxygen isotopes is linked to precipitation and evaporation. The core was radiocarbon dated and provided with a relatively robust age-depth model, which resulted in chronological uncertainty (a confidence interval) of around fifty years for both pollen and isotopes. Existing pollen and palaeoclimatic data from other Sicilian and southern Italian sites were also taken into account, thus expanding the spatial coverage of the analysis. In terms of the quantitative evidence for socioeconomic processes, the study drew on numismatic data (the presence of Eastern Roman coinage on Sicily) as well as archaeological survey datasets (rural settlement counts); it also referred to other archaeological material, mainly pottery, in its final interpretations (fig. 5).

Figure 5. Palaeoenvironmental data from Lake Pergusa, together with numismatic and rural settlement data from Sicily

Source: based on Laura Sadori et al., “Climate, Environment and Society in Southern Italy during the Last 2000 Years: A Review of the Environmental, Historical and Archaeological Evidence,” Quaternary Science Reviews 136 (2016): 173–88; figure modified after Adam Izdebski et al., “Realising Consilience: How Better Communication between Archaeologists, Historians and Natural Scientists Can Transform the Study of Past Climate Change in the Mediterranean,” Quaternary Science Reviews 136 (2016): 5–22.

The study argued for a strong four-sided correlation over a period of three hundred years (from the mid-fourth to the mid-seventh century) between (1) an increase in winter precipitation, as evidenced in oxygen isotope measurements; (2) a dearth of evidence for any arboreal expansion which would, in the region, typically occur in periods of rising humidity (without, that is, the human interference that was a feature of this period); (3) an expansion of cereal and olive cultivation, visible in the pollen data; and (4) a discernible increase in rural site density from the mid-fourth to at least the mid-seventh century. The authors suggested that the rise in winter rainfall (1) should have facilitated both an increase in farming and visible human impact on the landscape (2 and 3), as well as rural settlement expansion (4), particularly on the marginal, less-fertile lands of the usually dry island. It is important to note, however, that while there is no doubt about the temporal relationships between phenomena (1), (2), and (3), as they are reconstructed from the same sediment core, the correlation with phenomenon (4) remains problematic. The dating of rural settlements depends on ceramic typo-chronologies, the resolution of which can range from decades to centuries. Moreover, these typo-chronologies are subject to change as research on ancient ceramics progresses.Footnote ³⁰ The association between phenomena (1) and (2) and phenomenon (4) thus remains a picture painted in broad strokes, based on approximate chronological association and the assumed causal link between an increase in rural site numbers and an expansion of agricultural activities.

The interdisciplinary study on Sicily placed this potentially climate-related rural economic expansion in the context of a relatively well-known phenomenon: the contemporary growth of grain markets in the eastern Mediterranean, Italy, and North Africa following the foundation of Constantinople in 330 CE.Footnote ³¹ Importantly, the authors refrained from arguing for a strong causal link between climate, environment, and societal developments despite the relatively high temporal correlation between the climatic, environmental, and human phenomena observed, recognizing that human factors such as the foundation of Constantinople and its consequences were in themselves sufficient to explain the visible socio-ecological changes, or phenomena (2), (3), and (4). While climate may have played a part in these developments, its role was not necessarily causal. Rather, it may have offered a context in which certain socioeconomic processes could take place more efficiently. In this sense, it could be seen as an amplifier, facilitating the social-ecological change identified in the analysis.

The authors observed another interesting fact, namely a sequence of (a) declining winter precipitation starting in the late seventh century; (b) contracting cereal cultivation, which led to land abandonment and woodland expansion and was preceded by a short-term rise in rye, a drought-adapted cereal species; (c) a demonetization of the island; and (d) the waning of Sicily’s political significance within the Eastern Roman Empire and its final conquest by the Arab rulers of North Africa. As in the previous example, the temporal relationship between phenomena (a) and (b) is certain, as the data come from the same sediment core. The numismatic data are likewise characterized by a high degree of chronological precision.Footnote ³² Moreover, as the article notes, the decrease in coin deposition in Sicily was possibly related to broader economic and political processes, not least the general demonetization of the Byzantine economy and the imperial government’s fading interest in the island.Footnote ³³

What remains elusive, however, is the temporal connection between phenomena (a), (b), and (c) and phenomenon (d), the Arab conquest of the island. A chronological sequence of events is not enough to prove a causal connection between them (nor is temporal correlation), and the authors of the study were unable to provide further evidence to support the hypothesis that climatic deterioration weakened the island’s Byzantine social-economic system and led to its failure in the face of the Arab conquest. Notably, they remark that the Arab conquest of Sicily was very gradual—taking place over nearly a century—and that Byzantine political-military circumstances and the local (Italian) political situation suffice to explain both the Arab victory and the snail’s pace of the conquest. There was thus no collapse as such, but rather a slow process of conquest and socioeconomic transformation, which may have been in part triggered by a change in the regional winter precipitation regime that set in during the late seventh and early eighth centuries.

This study showcases both the potential and the shortcomings of the purely quantitative approach usually favored by palaeoenvironmental scientists, who often encourage historians and archaeologists to provide “numbers” that enable inherently quantitative palaeoclimatic and palaeoenvironmental reconstructions to be fitted with what are often inherently qualitative written sources. Studies like the Sicilian one can reveal intriguing co-occurrences and sequences, or indeed a lack of temporal connection, between climatic, environmental, and societal phenomena, but they are limited when it comes to establishing plausible causal links. Such approaches can serve as a useful first step in exploring the connections between climate and society, but the fragmentary and heterogeneous nature of the various types of data means that they cannot provide us with the type of answers that historians or environmental scientists often seek. In other words, the extent to which datasets created in different projects, with different agendas, can be used for other purposes, remains an open question. Nevertheless, these datasets are regularly incorporated into new projects, often with little consideration of their pros and cons. As long as this happens within the context of a similar line of inquiry or within the same field, it is usually not that problematic. However, when data originating in one field are used to answer questions in another, and as part of that transfer are detached from their original research context, important interpretational issues, with the potential to compromise final conclusions, may emerge.

To improve this common approach of combining palaeoclimatic reconstructions with the fruits of historical research, we need to ensure that all types of evidence mobilized—historical, archaeological, environmental, and so on—are operating at the same spatial, societal, and temporal scales. This is not always easy, as different datasets are frequently the result of disparate analyses: the final products were never truly designed to overlap with each other. The same is true in the palaeosciences. Most scientists collect only part of the data they use to arrive at their final conclusions themselves, instead incorporating substantial amounts of existing information. Bringing more data to the table has obvious advantages, facilitating larger-scale analyses and even interdisciplinarity, but it also means that authors risk losing control over the data employed and weakening the precision of their study. One solution to this problem is to invite the relevant specialists to join the study, which of course can heighten the complexity of the research but may also raise logistical problems and slow the project down. Nor does an interdisciplinary approach offer a magic bullet: researchers tend to be constrained by the nature of their evidence, so that comparison and correlation testing of differently calibrated time series potentially amounts to comparing incomparable variables. For this approach to be really effective, such studies should ideally result from workshops focused on specific research questions embedded in a clear spatiotemporal framework. Even better would be to establish more permanent interdisciplinary teams geared toward the restructuring of already existing series and, whenever necessary, producing new, ready-calibrated data.

The Pitfalls and Potential of Quantification and Statistical Inference

Sources that contain deliberate, continuous, and systematic hydrometeorological, atmospheric, or astronomical observations (for instance, recording potentially weather-relevant sunspot, auroral, or dust-veil phenomena) are highly prized, particularly for pre- or early instrumental periods—for Europe, this means roughly before the eighteenth century, when meteorological measurements began to be systematically taken and recorded across the continent. Such sources include weather diaries, sometimes made by trained professionals but often the virtuoso productions of keen amateur observers. These texts employ set terminologies to describe climate phenomena and their intensity or duration, reflecting their authors’ attempts to limit the subjectivity of their descriptions. In Europe, they are principally available from the later medieval period on,Footnote ³⁴ but systematic environmental observations were certainly maintained elsewhere in other—often institutional—contexts, with standout examples being the Babylonian astronomical diaries that survive from the seventh to first centuries BCE and Egyptian Nilometer records.Footnote ³⁵

Being readily quantifiable, such sources have advanced our understanding of past climatic changes and their societal consequences.Footnote ³⁶ But so too have more qualitative sources, with, for example, medieval annals and chronicles furnishing records of extreme weather, political events, and social behaviors that researchers have quantified in frequency and intensity.Footnote ³⁷ Annalistic sources have often been sidelined by scientific studies, in part because their focus on singular extreme events prevents a full annual reconstruction of average conditions. Le Roy Ladurie nevertheless argued that even given their limitations, rejecting annalistic sources out of hand would be “hypercritical … flying in the face of texts and wantonly refusing to admit witnesses who have a right to be heard.”Footnote ³⁸ Indeed, more recently event-focused texts have been increasingly used in tandem with natural archives to test specific hypotheses concerning past climatic behavior and societal vulnerability and response, particularly in medieval periods for which other sources are comparatively scarce.Footnote ³⁹

Figure 6. Page from the Annals of Ulster, showing entries from 852–858 CE

Note: in the inset, the encircled text reads “K1. Ianair.” This is an abbreviation of the “kalends of January,” that is, the first day of January, and marks each annual list of entries. The dashed underlined text gives the Anno Domini date in Roman numerals, and the solid underlined text shows the first historical entry for the year 856 (incorrectly given as 855 in the manuscript). This entry reports that “There was much ice and frost so that the principal lakes and rivers of Ireland could be crossed on foot and horseback from the ninth of the kalends of December [November 23, 855] to the seventh of the ides of January [January 7, 856]. A tempestuous and harsh year.”

Source: Trinity College Dublin MS 1282, the Annals of Ulster, fol. 42r. Reproduced by permission of the Board of Trinity College Dublin.

Take, for instance, the surviving corpus of medieval Irish annals, which provide reliable reporting of major historical events starting in the sixth century CE or earlier and ceasing only in the seventeenth century with the disruption caused by English colonial rule.Footnote ⁴⁰ These texts credibly identify some sixty-five severe “cold events” across this period (fig. 6). The information they provide can be combined with evidence for the timing and magnitude of historic explosive volcanic eruptions, which caused elevated fallout of atmospheric sulfate on the Greenland ice sheet and are thus identifiable by sulfate measurements in ice cores extracted from the region. Considering the two together reveals the major influence of explosive volcanism on the climate of the northeast Atlantic region represented by Ireland across more than a millennium.Footnote ⁴¹ Evidence of severe cold relative to Ireland’s mild maritime climatic norm was identified in the written sources (including frost, snow, or general descriptions of cold) and quantified by type, frequency, and seasonality. To allow a count of meaningful correspondences in the dates of cold events and eruptions—that is, those occurring close enough in time to suggest a plausible causal association—it was also necessary to consider the accuracy and precision of the respective ice-core and annalistic chronologies, as well as the timescales of relevant atmospheric and climatic processes, in particular those governing sulfate transport times to Greenland and lag times between eruptions and their climatic impacts. Some thirty-seven cold events (53.6% of those identified) were thus seen to occur within a few years after eruption dates, implying a sizable role for volcanism in the episodes of extreme cold that heavily impacted Irish society.Footnote ⁴² The Irish climate can, of course, experience severe cold even without volcanic forcing. But the correlation is a strong one: the probability of such a high level of co-occurrence arising purely by chance is vanishingly small at just 0.03%, reinforcing arguments in favor of the apparent connection.

Most cold events recorded in the annals were, moreover, reported in winter, regardless of whether they were associated with volcanism. The written evidence thus complements our wider understanding of historic volcanic impacts based on natural (biological) archives such as tree rings, which can usually only provide indications about spring-summer climate during the period of active growth. Some scholars have expressed concern that natural archives are accorded automatic primacy over human archives in studies of our interactions with the environment,Footnote ⁴³ but the Irish example shows that the two should be considered complementary. This understanding nonetheless requires an unbiased approach to these archives and the possibilities they present, an effort to learn their respective strengths and limitations, and confidence in what each discipline has to offer. Credible quantification of the Irish evidence required, for example, an understanding of the genealogy of the annalistic texts, which due to their shared ancestry often duplicate reports, and because of scribal errors over time can locate these under different years. This can effectively disguise their duplicate status and mean they are counted twice or more. Understanding the motivations for recording weather was also crucial to identifying reports of dubious veracity: we must consider how extreme events were perceived by both the scribes and their intended audience, including as portents or vehicles of divine retribution, and whether exaggeration or fabrication could serve a political, rhetorical, or theoretical agenda.Footnote ⁴⁴ While earlier efforts in historical climatology were criticized for paying insufficient attention to such concerns,Footnote ⁴⁵ historical expertise is now deemed fundamental to the use of written evidence in climatic reconstruction.Footnote ⁴⁶ Its importance is clear in the Irish example. If, in a hypothetical worst case scenario, all dubious reports were included as genuine and all duplicated reports went undetected, the total number of cold events would increase to 181, an inflation of 178.5%.Footnote ⁴⁷

Studies such as the Irish example place historians, their sources, and their methods in service to natural scientific questions. Another example, centered on a study of ancient Egypt led jointly by a historian and a historical climatologist,Footnote ⁴⁸ highlights the use of natural archives, quantification, and statistical modeling in pursuit of historians’ questions concerning causality. Here, quantification of socioeconomic activity alongside major historical events for one of ancient Egypt’s most richly documented eras, the period of Ptolemaic rule from 305 to 30 BCE, made it possible to test new and long-standing hypotheses about the historical influence of hydroclimatic variability.

Egypt provides an excellent laboratory for the study of human-environmental relations, with its general lack of rainfall driving dependence on the Nile to compensate via irrigation and flood recession agriculture. This makes Egyptian society historically vulnerable to the hydroclimatic impacts of volcanic forcing. Loading of the stratosphere with sulfate aerosols from large explosive eruptions imposes short-term energy imbalances on the climate system due to modified reflection of the sun’s radiation and can result not only in strong near-surface cooling on a hemispheric or global scale (the same mechanism at least partially responsible for severe Irish winters), but also related changes in the hydrological cycle. Instrumental data and climate modeling suggest that strong tropical eruptions tend to reduce global average precipitation (mainly via cooling that reduces evaporation), while eruptions in the higher latitudes of the northern hemisphere can reduce the north-south temperature contrast that drives the northward migration of moisture-bearing summer monsoonal winds. This is key for the Nile summer flood, as it is mainly fed by African monsoon rainfall over the Ethiopian Highlands. Before the construction of dams in the twentieth century, this monsoon rainfall would result in the Nile beginning to rise at Aswan on Egypt’s southern border from early June. The highest levels were reached in August and September, with the water usually receding by the end of October, when sowing began.Footnote ⁴⁹

The Ptolemaic era was a watershed in Egyptian history that witnessed the establishment of a new political and economic framework, including a new capital at Alexandria—then one of the largest urban settlements in the Mediterranean world. Free-threshing wheat, comparatively vulnerable to drought, was one of the principal crops throughout Egypt during this period, serving the preferences of its Greek elite and a wider Mediterranean market.Footnote ⁵⁰ New fiscal institutions, the issuing of coinage, and the development of banks and tax farming placed Egypt’s already famous agricultural productivity under tighter control, allowing the state to extract more surplus from its territory.Footnote ⁵¹ Major wars with its main rival, the Seleukid kingdom in western Asia, dominated interstate politics during the third and second centuries BCE.Footnote ⁵² Generally regarded as the richest and most successful of the Hellenistic successor states to Alexander the Great’s empire (and indeed comprising the longest single dynasty in Egyptian history), Ptolemaic Egypt was increasingly troubled from the end of the third century BCE, wracked with intermittent but increasing social unrest and facing pressure from an expanding Rome, continued animosity from its Seleukid rivals, and internal tensions associated at least partly with demographic growth.

Figure 7. The effect of volcanic eruptions on the Nile flood and the Ptolematic kingdom

Note: panel a): ensemble mean precipitation minus evaporation (P-E) response to five twentieth-century volcanic eruptions in CMIP5 (coupled model intercomparison project 5) model output. The response is the average P-E over the first summer season (May to October) that contained or followed the eruption, relative to the five summers preceding it. Only anomalies statistically significant at the 5% level are shown. The area within the green line is the Nile watershed. Panel b): annual Nile summer flood heights from the Islamic Nilometer composited relative to the ice core-estimated dates of eruption years for sixty large eruptions between 622 and 1902 CE (eruption years are represented at point 0 on the horizontal axis; years 1 to 8 then represent the eight years after these eruptions, and years -1 to -8 the eight years before). Shading indicates the 2-tailed 90% confidence interval, estimated using the t-distribution. Nile summer flood heights are on average twenty-two centimeters lower in eruption years (P < 0.01).Panel c): onset dates of internal revolt against Ptolemaic rule composited relative to the dates of sixteen volcanic eruptions (represented at year 0 on the horizontal axes; years 1 to 8 then represent the eight years after these eruptions, and years −1 to −8 the eight years before), excluding an eight-year buffer at the start and end of the period considered (305–30 BCE). Dots indicate statistically significant values estimated using Barnard’s exact test. The green dashed line gives the 95% confidence threshold, also estimated using Barnard’s exact test.

Source: figures and caption text adapted from Joseph G. Manning et al., “Volcanic Suppression of Nile Summer Flooding Triggers Revolt and Constrains Interstate Conflict in Ancient Egypt,” Nature Communications 8 (2017): https://doi.org/10.1038/s41467-017-00957-y.

In the study presented in figure 7, the link between Ptolemaic political and economic history and explosive volcanism was identified thanks to the integration of evidence from ice core-based volcanic forcing data, climate modeling, and ancient papyri and inscriptions. The authors began by comparing summer flood heights, recorded primarily at the famous Nilometer on Roda island near Cairo, with dates of explosive volcanism between 622 and 1902 CE, established through the study of ice cores. This revealed a persistent reduction of the summer flood following large tropical and extratropical eruptions, a result supported by analysis of climate model output showing drier conditions across the Nile basin after major twentieth-century eruptions (figs. 7a and 7b). To confirm that this volcanic flood suppression effect also occurred during the Ptolemaic period, qualitative indicators based on papyri and inscriptions were used to create annual flood-quality ranks on an ordinal scale and then analyzed.Footnote ⁵³ Flood quality was indeed shown to tend toward lower ranks during volcanic years. This association between explosive volcanism and suppression of the agriculturally critical Nile summer flood thus makes it possible to hypothesize a plausible link between eruptions and major historical events in Ptolemaic Egypt.

The ability to compare the timing of repeated volcanic events with that of repeated historical events was also foundational to the study’s approach, allowing the identification of statistically significant temporal associations across the Ptolemaic period. Due partly to the quality of the Egyptian documentation, the important uncertainties in dating that can hamper such analyses were minimized, while the history of volcanic forcing was derived from the recent synchronization of data from multiple ice cores.Footnote ⁵⁴ Several kinds of recurring and datable historical events could thus be considered, including ten internal revolts against the Ptolemaic kings.Footnote ⁵⁵ The onset of three of these coincided with eruption years, with another five occurring within the space of two years after an eruption (fig. 7c). The same holds for other kinds of events. Of the nine occasions on which Ptolemaic interstate warfare with the Seleukid kingdom ceased, three fell in an eruption year, with two others occurring within two years and one more within three years; two out of nine priestly decrees were issued in eruption years, with one more in the first year following an eruption. Finally, the sale of hereditary lands, long considered symptomatic of economic distress, with families usually averse to such transfers,Footnote ⁵⁶ was also observed to spike upward, on average, for several years following eruptions.Footnote ⁵⁷

What is striking is thus the repeated coincidence between a small number of well-dated events: explosive eruptions, internal revolts and the likely related cessation of external military campaigning, attempts at reestablishing or maintaining political order through priestly decrees issued on behalf of the Ptolemaic kings, and an increased frequency of land sales.Footnote ⁵⁸ It must be recognized that this suggestively close conjunction of events and eruptions could have occurred randomly. Nevertheless, the probability that the association between volcanic eruption dates and these events was entirely down to chance was examined, and the likelihood found to be low (at or considerably below 5% in all cases).Footnote ⁵⁹

To illustrate how statistical models can usefully inform historical research on past climate-societal interactions, it is worth briefly elaborating on the study’s methodology. Consider the observed association between the onset of revolts against Ptolemaic rule, documented in papyri, and the dates of volcanic eruptions known from the polar ice cores. Even if these events were in reality unrelated, with no causal link between them, it would still be expected that an eruption would sometimes occur in the same year as a revolt, purely by chance. The relevant statistical question is, then, how unusual is the number of correspondences actually observed? Writing a probabilistic model for the number of correspondences to be expected if there were no real association between these events is straightforward. The authors conducted simulations that randomly reassigned each eruption to a new year between 305 and 30 BCE, then counted the number of correspondences between revolt onset and the reassigned eruption dates, repeating this process a million times to build a random reference distribution.Footnote ⁶⁰ Since the number of correspondences observed in reality exceeded the number of correspondences in 98% of these simulations, they concluded that there is a greater than 98% chance that a real association exists between revolt onset and explosive volcanic eruptions; in other words, that the association has not occurred randomly, and the observed correspondence is thus statistically significant at 98% confidence.

This testing of statistical significance supported the authors’ argument for a causal link between explosive volcanism and revolt onset. The main causal pathway they propose hinges on the demonstrable hydroclimatic impact of volcanic eruptions on the Nile and the known heavy (if not complete) dependence of Egyptian agricultural productivity on its flood. The reduction in crop yields would then promote food scarcity, price increases, fear of starvation, difficulties in paying state-imposed taxes, land abandonment, and migration to urban areas in search of relief. Ultimately, this upheaval could result in revolts, particularly when certain background conditions—not least ongoing ethnic tensions between native Egyptians and Greek elites, or costly military mobilizations—added to the tension.Footnote ⁶¹ A finer understanding of the role of such conditions, which were certainly in flux over the nearly three centuries of Ptolemaic rule, remains to be achieved.

Alternative eruption-to-revolt pathways are also possible, either as primary or auxiliary causes underlying any given revolt. One hypothetical example would have eruptions disrupt Egyptian society through abrupt cooling that would reduce crop output and promote animal mortality; another through a dust-veil phenomenon favoring unrest associated with local religious interpretations—for instance if the dimming or discoloration of the sun’s light were taken as signaling divine displeasure. Depending on the prevailing historical context, each causal link could be mitigated (or aggravated) by human action. For example, tax reductions, the imposition of food price ceilings, export bans, state granaries, and emergency grain imports might conceivably weaken the connection between reduced crop yields and increased social stress. It is important in this respect to note that not every volcanic event in the Ptolemaic period is associated with a revolt, and that Ptolemaic Egypt was by implication resilient to some volcanic hydroclimatic shocks.

A notable feature of social unrest in the period is the degree to which it varied in duration, extent, and severity. Some known periods of revolt seem to have touched the whole of Egypt and lasted twenty years, while others appear to have been more short-lived and localized. A simple explanatory model of unrest as a response to climate shocks cannot fully explain this variability. Might there be a correlation between the severity of climatic events and the intensity of the revolts? The scale of Nile flood failure, on some occasions causing only short-term local distress and at other times provoking multiple years of drought impacting the entire Nile valley and potentially leading to major famine, would likely have resulted in different scales of response. But evidence from the sources complicates this hypothesis: more localized flood failures may also have produced panic and fear as inhabitants recalled earlier episodes of crisis, as documented in the 238 BCE Canopus Decree.Footnote ⁶² In both scenarios, stress resulting from poor flooding can be hypothesized as an amplifier of existing ethnic tensions in certain parts of Egypt.Footnote ⁶³ There is thus substantial scope for primary work on historical sources to complement historical-statistical approaches and enrich their interpretation. By returning to the documents, historians can further refine their understanding of the nature and scale of periods of social unrest in Ptolemaic Egypt, independently of the palaeoclimatic data. They can also consider their geographical resolution—topography may play a role in how vulnerable a particular area was to a reduction of the Nile flood by a given amount—and divide them into meaningful typologies in order to distinguish more fine-grained associations between historical events and hydroclimatic shocks.

Approaches such as those illustrated in the Irish and Egyptian examples remain comparatively rare in the broader field of environmental history.Footnote ⁶⁴ We would argue, however, that these kinds of studies have much to offer, given the increasing focus on generating quantitative data and specialized methods to analyze them. In particular, they can open the field up to modes of reasoning characteristically reserved to the social sciences and applied to data-rich modern societies. Certainly, there are also limitations and concerns. As Aryn Martin and Michael Lynch have remarked, while “it is a simple practice when considered abstractly, in specific cases counting can be quite complicated, contentious, and socially consequential. Categorical judgments determine what counts as an eligible case, instance, or datum, and these judgments can be difficult and controversial.”Footnote ⁶⁵ In historical climatology, the categorization and quantification of historical information (especially qualitative information) is essentially and inherently reductive.Footnote ⁶⁶ Some degree of nuance and context is inevitably lost when coding events or phenomena according to set categories that can never perfectly encompass all dimensions of an event, be it a drought, a famine, or anything else.Footnote ⁶⁷

There is no way to perfectly capture the relevant content of complex written evidence, but not everything must be quantifiable to legitimately employ this approach to “reading” a text, or, inversely, to allow nuance and contextual knowledge about a source’s creation to inform quantitative analyses. Ultimately, the benefits of this approach, as illustrated in the Irish and Egyptian examples, are real. But for it to contribute to the fullest extent possible, it must be received in an open-minded way, with a balanced appreciation of its potential and its limitations and the understanding that it complements rather than competes with more established approaches in history.Footnote ⁶⁸

Reinterpreting Textual Sources in Light of the Scientific Data

The two types of interdisciplinary practice described above focused on situations where the natural scientific data could in one way or another be complemented or enriched by additional information from the textual and material evidence that is usually the preserve of historians. The main difference between these approaches concerns the extent to which they apply quantitative methods and how they organize their reasoning as a result. In this section, we will look at two cases in which the direction of the intellectual inquiry goes the opposite way, that is, where the scientific data are used to interpret the textual evidence at the heart of the study. The first example deals with literary sources and focuses on late antique Italy; the second looks at documentary records and focuses on medieval Poland. In both cases, the availability of new palaeoclimatic or palaeoecological data has improved our understanding of written sources that historians have used for generations.

First, let us consider the example drawn from late antique Italy. Beginning in the 1980s, scientists connected natural archives suggestive of a powerful volcanic eruption in the mid-sixth century CE to contemporary written sources that appeared to record the subsequent atmospheric clouding.Footnote ⁶⁹ Some argued that the climatic downturn associated with the eruption, known as the “536 event” for the year in which the aftereffects became noticeable, had a huge impact on societies across the entire northern hemisphere.Footnote ⁷⁰ The insights made available by natural proxies, principally polar ice-core records and long-running dendroclimatological series, have encouraged a steady stream of studies revisiting and reinterpreting textual and archaeological sources in light of new environmental information.Footnote ⁷¹ Seemingly disparate historical texts have thus been collated together as witnesses to a climatic event, for which the palaeosciences provide additional context.

Debates concerning the scope and magnitude of the eruption’s historical impact have been adjudicated almost entirely based on the palaeoscientific evidence,Footnote ⁷² with the textual sources receiving less attention.Footnote ⁷³ However, a close reading of one famous source significantly alters the narrative framework associated with the event, all while remaining consistent with the climatic realities established by other disciplines. The longest known description of the 536 event is contained in the Variae of the official Cassiodorus (ca. 485–ca. 580), a collection of his correspondence while in the service of the Ostrogothic monarchy that ruled Italy between 493 and 553.Footnote ⁷⁴ At first reading, the letter supports a catastrophist interpretation. We are told that the sun has lost its usual brightness and that temperatures have been cool for nearly an entire year.Footnote ⁷⁵ The seasons themselves have been thrown into disarray: “we have had winter without storms, spring without mingled weather, summer without heatwaves.”Footnote ⁷⁶ Realizing that crops could not grow under such conditions, Cassiodorus orders a subordinate to open state granaries to stave off imminent famine.Footnote ⁷⁷

Though this appears to be a powerful image of the unrelenting power of climate over premodern society, it should not be taken at face value. The many descriptions of environmental disasters found in Cassiodorus’s writings are sometimes pulled directly from earlier accounts of similar anomalies. His observation that the strange clouding lasted “almost a full year” echoes word-for-word Pliny the Elder’s account of the sole other recorded example of prolonged solar dimming,Footnote ⁷⁸ in the aftermath of Julius Caesar’s assassination in 43 BCE.Footnote ⁷⁹ Another element of Cassiodorus’s striking description—how “air laden with snow … blocks the heat of the sun and deflects the view of human frailty”Footnote ⁸⁰—has a parallel in a second well-known account of Caesar’s comet found in Plutarch’s Life of Caesar: “heat came to earth only faintly and ineffectually, so that the air hung dark and thick on the earth because of the lack of radiance to penetrate it.”Footnote ⁸¹ That such intertextual allusions were common stylistic hallmarks of Roman literature serves as a warning against treating these statements uncritically.

Identifying how the 536 eruption impacted Italy’s climate and society with any precision then becomes an impossible task, and we are left with the generalizing observation that environmental oddities of some sort were noticeable at this date. Reconstructions of past climatic events and processes should be assessed using the scientific evidence. The use of historical texts, especially those rich in rhetoric and artifice, for this purpose is a flawed enterprise because they rarely specify the duration, scope, or magnitude of the phenomena they describe. However, our interpretation of these texts is enriched through the addition of information that can only be derived from the palaeosciences. The intertextual references in Cassiodorus’s description, for instance, are in fact an effort to transform the unprecedented into the legible using a framework familiar to Roman literary culture. To date, most of the new interdisciplinary research has focused on understanding climate’s impact on human populations. We should also ask how the authors of our sources engaged with climate, and the study of the literary sources may be one of the best ways to do that.

The second case that we want to present shows that focusing on documentary rather than literary sources—that is, on texts a priori devoid of the (potentially misleading) intertextuality apparent in Cassiodorus—leads to a similar conclusion. This example concerns Greater Poland (Wielkopolska), the cradle of the Polish state, established in the second half of the tenth century CE and for a long time one of the most economically developed, populous, and richest regions in the country. The economic history of Greater Poland (and Poland as a whole) has been the subject of many studies, especially concerning the early modern period. Nevertheless, one of the primary obstacles remains the limited number of written sources. Administrative records such as tax registers or estate inventories only appear in the sixteenth century. The availability of documentation for the early Middle Ages is particularly limited: though an inventory of archaeological excavations of early medieval material in Greater Poland does exist, it tends to focus on proto-urban centers, despite the fact that the dominant forms of economic activity in the region were agriculture and livestock farming.Footnote ⁸²

To reconstruct the network of medieval rural settlements and the state of the rural economy, historians have thus relied on one particular kind of source, the “location privileges.” These charters recorded the rules governing relationships between landlords and peasants and were issued from the thirteenth century, when German colonization began in Greater Poland.Footnote ⁸³ Recent research has established the fourteenth century as the key period in this process, and charters from that time record a significantly larger number of existing and newly established villages.Footnote ⁸⁴ As a result, it has become commonplace to argue that the network of rural settlements in Greater Poland—which, with its regular layout of fields and placement of houses, has in some areas survived until today—was a consequence of changes that took place during the thirteenth and fourteenth centuries.Footnote ⁸⁵ Villages were transferred to German law, and the farming methods used on the surrounding lands modernized at the same time.

However, although these location privileges created a kind of legal reality, we must recognize that they actually represent no more than the plan or intention to establish a new village or to modernize an existing one. They should not be taken as proof that such plans were immediately or indeed ever implemented.Footnote ⁸⁶ Moreover, primary sources of this kind do not record the chronology of a village’s demographic and economic development, nor even moments of crisis; all they do is document relevant legal actions. The inherent limitations of these sources have long been an obstacle to understanding the dynamics of Greater Poland’s medieval economic transformation. Yet this impediment can be overcome by using environmental archives and conducting joint historical-environmental research.

A recent historical-palaeoecological project carried out in Greater Poland has taken the form of micro-studies, with high-resolution multi-proxy palaeoecological research conducted on the basis of material extracted from the region’s peat bogs.Footnote ⁸⁷ Focusing on breakthrough moments in its political and economic history, integrating historical and environmental archives, and mobilizing both humanistic and scientific research methods, this study has altered our understanding of the course of late medieval colonization in the region. Notably, the frequency of the radiocarbon dates obtained (more than one per century) made it possible to better grasp the dynamics of the colonization processes. The analysis of environmental data revealed that the shift to German law evoked in the written sources was less dynamic than historians once argued, and probably followed a different, far longer chronology than was previously thought. Palynological data show not only a gradual deforestation, but also a similarly gradual increase in the proportion of pollen from cultivated plants, notably cereals (fig. 8).

Figure 8 Development of rural settlement in Greater Poland

Note: the upper panel is based on pollen analysis data; the center and lower panels on data from location privilege charters. The middle panel reflects the number of new settlements in the immediate vicinity of the site shown on the upper panel, namely the Kazanie peat bog, near Pobiedziska in Greater Poland; the lower panel shows data for the entire Greater Poland region.

Source: based on Konstanty Jan Hładyłowicz, Zmiany krajobrazu i rozwój osadnictwa w Wielkopolsce od xiv do xix wieku (Lwów: Kasa im. Rektora J. Mianowskiego, 1932), 109–227; Sambor Czerwiński et al., “Znaczenie wspólnych badań historycznych i paleoekologicznych nad wpływem człowieka na środowisko. Przykład ze stanowiska Kazanie we wschodniej Wielkopolsce,” Studia Geohistorica 7 (2019): 59 and 63; Czerwiński et al., “Environmental Implications of Past Socioeconomic Events in Greater Poland during the Last 1200 Years: Synthesis of Paleoecological and Historical Data,” Quaternary Science Reviews 259 (2021): https://doi.org/10.1016/j.quascirev.2021.106902, p. 8.

The Kazanie peat bog in central Greater Poland is a good example of such slow, progressive development. Although most of the settlements around the site were established or reorganized according to German law in the thirteenth and fourteenth centuries, a significant increase in human economic activity resulting from the development of agriculture can be observed only from the fifteenth century. It was at this point that the population increased noticeably and farming methods were probably modernized via the introduction of the three-field system, more advanced tools such as the plow and iron harrow, and a new pattern of field organization. Only at the end of the fourteenth century does the share of cereal pollen in the samples collected for the study permanently exceed 1%, before growing over the fifteenth century to reach 3%. This means that the colonization which began in the thirteenth and fourteenth centuries did not entail large groups of settlers. The newly established or modernized villages were relatively small, with a demographic and economic potential that increased slowly over a long period of time to reach a peak in the sixteenth century.Footnote ⁸⁸

Supplementing the existing base of written documentation with environmental sources thus results in a substantial revision of the traditional historiographical vision of German colonization: the palaeoenvironmental data show a more gradual pace of rural expansion in Greater Poland, based to a large extent on natural demographic growth. As with the Latin literary texts discussed above, incorporating natural scientific data can either lead to a completely new interpretation of the written sources or substantiate one existing argument at the expense of others—some of which have been developed over decades and long persisted in historical scholarship. There is no doubt that as the precision and versatility of the palaeoscientific data grows, we will gain a more comprehensive and solid understanding of the past. It is to be hoped that the methodological and practical perspectives offered in this article will precipitate and facilitate this process.Footnote ⁸⁹

Constructing Historical Narratives in Heterodisciplinary Environments

In his 1967 monograph, Histoire du climat depuis l’an mil, Le Roy Ladurie concluded that over the long term the human consequences of climate seemed to be slight and difficult to detect.Footnote ⁹⁰ Now, after more than fifty years of research, we know that this statement was overly skeptical or at least reductionist, tending toward anthropocentric solipsism. In fact, Le Roy Ladurie was wary of the simplistic interpretations that might be attributed to his work. Over thirty years later, when the global climate crisis directed public and scholarly attention back to climate’s role in human affairs, the same author formulated a stronger case for the role of climatic variability in human history in his three-volume Histoire humaine et comparée du climat, published in 2004.Footnote ⁹¹ In a way, this was a return to the positions held by the generation of historians active in the first decades of the twentieth century, such as Ellsworth Huntington in the United States or Franciszek Bujak in Poland.Footnote ⁹² More generally, however, this shift shows the tensions faced by historians who decide to write about the “natural” aspects of history, not least the ready ideological master narratives into which one can easily be subsumed, even against one’s will.

In most studies relating to climate, the master narrative that comes to the fore depends on which humanistic or natural scientific discipline is predominant in the analysis.Footnote ⁹³ To understand what is at stake in this struggle to define the master narrative, it is worth discussing briefly the main rhetorical tropes that dominate the emerging field of interdisciplinary environmental history, applying Hayden White’s schema of metaphor, metonymy, synecdoche, and irony to the environmental context.Footnote ⁹⁴ Until recently, the catastrophism and determinism that served as the main rhetorical trope in climate change history was something of a metonymy. Its reductionist representation of reality focused on a single cause to promote a simple mechanistic argument, ultimately deploying a story dominated by a tragic plot. Such stories usually have the radical intention of shaking their readers’ consciousness and encouraging them to take action.Footnote ⁹⁵ This is why, despite their ability to mislead by streamlining the past and their fixation on a tragic ending, catastrophism and determinism are not without merit. They can be deployed in historical narratives to motivate human action, and it is no surprise that such master narratives tend to prevail in stories authored by natural scientists. In the end, it is they who are most acutely aware of environmental change and of the severity of the environmental crisis we are approaching.Footnote ⁹⁶

Of course, metonymy-catastrophism is just one of the multiple tropes available to historians (White described four, but emphasized that there are probably more). A mature field of research such as environmental history, therefore, should not restrict itself to a single trope, a unique narrative arc for recounting the past. At present, the most favored alternative master narrative is “resilience,” something of an irony, if not a synecdoche, in White’s terms. Synecdoche is characterized by an integrative mode of representation that uses the part to consider the whole (resilience accounts aim to present holistic perspectives based on specific examples). It combines an organicist argument (emphasizing the unity rather than conflict of society and nature) and a conservative ideology (resilience is often used to reinforce neoliberal policies by invoking humanity’s capacity to “bounce back”). Unsurprisingly for a synecdoche trope, whose basic “emplotment” is comedy (ending on a positive note), resilience narratives tend to be cautiously optimistic. This optimistic tone can make even their authors feel uncomfortable in the context of global climate change, and studies that foreground these narratives often include admonitory notes about not taking their optimism too far or underestimating the seriousness of the planetary crisis.

Resilience has emerged within scholarly discourse as one method for assessing the impacts of climate and climatic variability on human populations past and present, migrating from an original definition that focused first on engineering and then on ecological systems.Footnote ⁹⁷ At its heart, it attempts to measure a system’s durability and stability in the face of various stressors that threaten its equilibrium: the system’s resilience is its “resistance to change.”Footnote ⁹⁸ In environmental history, the concept has most often been used to study climate’s impact on the state and socioeconomic institutions of past societies at varying levels, from the empire to the village,Footnote ⁹⁹ notably in terms of the decline and formation of states, agricultural productivity, settlement patterns and density, and so forth. Resilient societies were able to weather the storms, as it were, and adapt to new realities while maintaining their fundamental social attributes (as defined by the modern observer-researcher). Despite the place catastrophism has held in natural scientific publications on climate history and in popular environmental history books, it has never completely dominated: structuralist interpretations of the relationship between the environment and human history, akin to resilience theory, are some of the most enduring within the field, stretching back to Le Roy Ladurie’s first publications.Footnote ¹⁰⁰ Though they avoid the declensionist tendencies common to metonymy-catastrophist narratives, synecdoche-resilience narratives nonetheless perpetuate the idea that climate acts and humans respond to it. Even successful adaptations and ameliorations are still reactions to outside stimuli. To date, most interdisciplinary research has thus focused on understanding climate’s impact on human populations.

The humanities and social sciences have nevertheless put forward a strong philosophical critique of socio-ecological resilience, in particular how it serves to defend the neoliberal status quo. In a resilience narrative, action seeks to strengthen existing systems so that they can better survive environmental crises, rather than to change the systems that provoked those crises in the first place. Past precedents of resilient societies (based on historical analysis) should not be used to legitimize an enduring socioeconomic regime that endangers life on earth. As soon as one recognizes that neoliberalism is the dominant ideology of the contemporary globalized world, metanarratives of resilience and their optimistic use of synecdoche (in which the different parts of a system can survive environmental crisis) can be seen for what they are: a conservative position (maintaining the existing order) rather than a liberal one (unconcerned by people affected by climatic and environmental changes). This is yet another reason why resilience narratives should not be privileged unquestioningly in the field of interdisciplinary environmental history, despite the repeated calls to move away from catastrophist narratives.Footnote ¹⁰¹

Ultimately, we should reject neither metonymy-catastrophism nor resilience-synecdoche narratives, but rather work to expose the strengths and weaknesses of each and continue experimenting with their application to produce a truly interdisciplinary environmental history. The more aware the field becomes of its theoretical and rhetorical entanglements, the easier it will become to blend tropes, invent new ones, and carry across a socially powerful and intellectually insightful message that is needed in our times.

To conclude, we wish to emphasize that the interdisciplinary environmental history proposed in the preceding pages does not displace the role or function of the traditional historical narrative. Instead, we would simply argue for the incorporation of different types of evidence—and their interpretation—when they are relevant. As demonstrated in this article, the relevance of the palaeosciences is surprisingly wide and encompasses nearly every type of historical question.

The main challenges to this incorporation remain practical and theoretical rather than financial. While many socio-environmental research questions do indeed require larger-scale projects with funding from national or European research agencies, several groundbreaking studies have emerged through grassroots networking and collaboration. Most of the case studies presented in this article, for example, were prompted by a geologist producing interesting data and then seeking out a historian to help contextualize it. The real obstacles to this interdisciplinarity more often lie within our own field. The traditional academic structures of history as a discipline are not accustomed to researchers moving across periods and regions, trying to harness the expertise of entire teams in order to tackle socio-environmental research questions. Although the discipline as a whole is proud of such undertakings, which prove its vigor and relevance to a wider public, often its practitioners prefer to stay within their “tribal” structures and tacitly reject the new academic habits and values, the openness that is the prerequisite of interdisciplinary history. Those scholars firmly entrenched in history, used to being in control of both the narrative and the data, often do not value the teamwork that leads to joint publications in which it is the synergy of several authors that creates intellectual value, rather than the well-demarcated contribution of an individual. If the discipline of history, and particularly the field of preindustrial environmental history, is to avoid becoming dated or slipping into a kind of antiquarianism, it must openly confront the issues of co-authorship and collaboration in teams with mixed expertise—already current in the natural sciences or even history’s sibling, archaeology. For that, perhaps the surest way forward is to reorient how we educate and train those historians interested in the environment who need exposure to the relevant natural sciences. They must learn how to handle both the written sources and the natural scientific data with care. With the flourishing of different approaches to the past, history as a discipline—and as an intellectual adventure—has a great future ahead of it.