The landscape, environmental and climate context of the southern Pannonian Plain 2000–1000 BC

The Carpathian Basin is defined as a bowl with a mountainous periphery and an exceptionally flat interior (the Pannonian Plain) punctuated by very rare and isolated hills or mountains (Fig 1). Features such as marshes appear to have played a role in shaping settlement locations in prehistory, but the massive rivers of the plain were even more formative. They are broad all year round but when in flood they become vast watery zones creating seasonal no-man’s land spaces. These rivers possess a ‘“yin/yang” character, in that they function both to hinder and to facilitate movement’, enabling societies to shape, interact and sustain large multi-local social aggregates or perhaps polities [20].

To better understand the changing environmental context of settlement between the MBA and LBA, and also the eventual abandonment of LBA settlements, data for paleoclimatic conditions and landscape management are revealing (Table 1). Studies from areas adjacent to the TSG present a consistent picture. Demény et al.’s [21] study of a variety of climate proxies provide the highest resolution records and are supported by other speleothem data from Ascunsă, Poleva and Urşilor caves, an ice core from Scărișoara cave, and sediment cores from Lake Brazi and Late Ighiel in western Romania [22–26]. Most LBA 1 and 2 datapoints indicating wetness levels are below those indicating the wettest MBA conditions, suggesting consistently higher rainfall. The warmest peaks of LBA 1 and 2 are just above the lowest points for the MBA in Demény et al.’s model, suggesting relatively cooler LBA conditions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. General trends in climate proxy data for the southern Pannonian Plain.

This table is based on data from Demény et al. 2019 and supported by data from Drăguşin et al. 2014. Demeny et al. is based primarily on data from Trio Cave and Dragusin et al. focus on Ascunsă, Poleva, and Ursilor caves. These are located ca. 200 km to the west-north-west and east-north-east of the centre of the TSG network respectively. All conditions in the figure are relative to those of 2000–1900 BC when this timeline begins. As the dating of changes is broadly construed within century blocks, when peaks occur within a block, the following trajectory is opposite to the preceding one, i.e. if preceded by cooling then following the peak, prevailing conditions shifted to warming.

https://doi.org/10.1371/journal.pone.0288750.t001

When the number and complexity of Middle Bronze Age (MBA) tells were increasing (broadly 1950–1750 BC), palaeoclimate data indicate comparatively warm and dry–but quite vacillating–conditions. Data suggest a steady increase in wetness and falling average temperatures from the 17^th to mid-16^th century BC [21]. The same data indicate the onset of less frequently fluctuating conditions in the 16^th century BC. This 16^th century BC change horizon was sufficiently pronounced for Kern et al. [27] to posit a 3.5ka climate event occurred across east-central and south-eastern Europe. During the LBA 1 and 2 phases the climate appears relatively cooler and wetter than the growth period of MBA tells. During the 14^th to 13^th centuries BC rainfall decreased alongside rising temperatures but with relatively stable hydroclimatic conditions overall. This trend accelerates ca. 1250 BC, with an increasingly arid climate and rising temperatures for ca. two centuries, bringing conditions close to those of the growth period of MBA tells [21, 23]. French’s [28] geoarchaeological study of the nearby Benta Valley suggests this 13^th-12^th century aridity played a role in the slowing of rivers in parts of the plain.

Hydroclimate and temperature observations are primarily derived from speleothem analyses, with the Hungarian data possibly biased towards reflecting wintertime conditions. There is less clarity on the seasonality of change (greater summer or winter rainfall patterns) in the Romanian data. In many ecosystems, winter rainfall provides an opportunity for soils to recharge and remain productive in annual cycles, particularly when comparatively dry summers are experienced [21, 23]. This may have had benefits for agricultural productivity.

Regarding vegetation history of the Pannonian Plain, there are few data relevant to the Bronze Age, largely a consequence of poor pollen preservation. Gumnior et al.’s [18] study of the site and hinterlands of Corneşti Iarcuri, a LBA mega-fort of the lower Pannonian network in Western Romania, shows diverse species and relatively high tree-coverage, with arboreal pollen accounting for 38–46% of pollen from the settlement and 60–65% of pollen from nearby lake sediments. Notably, anthropogenic pollen markers are reduced after 1200 BC, suggesting a shift in landscape exploitation [18].

Working at a multi-decadal scale, we can identify a general correlation between the MBA / LBA transition and changing climatic conditions. This provides a context, rather than a cause, for social transformations given that societies existed at the interface between environmental conditions and social expectations of landscape use [22, 29–36]. The exceptionally flat landscape of the southern Pannonian Plain with its frequent and massive waterways means that ecosystems were finely balanced relative to hydroclimatic conditions. For this reason, socially specific modes of land use would have been vulnerable to shifts in the predictability and levels of rainfall and their impact on the water table.

Material culture and the relative chronology

The following brief overview of ceramic relative chronologies introduces relevant terminology (Fig 2). The MBA in the south Pannonian Plain was a complex patchwork of material culture styles. At the peak of tell habitation, sites located south of the river Timiş primarily consumed variants of Vatin style pottery, for example at the tells of Židovar and Feudvar and the settlement at Vatin. Along the Mureş river, tells such as Pecica, Periam and Kiszombor were dominated by the Maros style of pottery alongside a minority of the Otomani-Füzesabonystyle that was dominant farther north [19, 37, 38]. Recent work favours the term Cornești-Crvenka style for pottery of the latest phase of the MBA, though Vatin style can also be found in literature of recent decades [38–40]. On many tells a combination of styles can be found in various proportions, though minority styles tend to be labelled as imports. There remains debate about boundaries and relevance of these groups, but relevant literature uses these conventions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Relative ceramic chronologies commonly used in the study area (Drawn by Caroline Bruyere).

https://doi.org/10.1371/journal.pone.0288750.g002

The long-standing Tumulus Culture invasion model was introduced to explain the adoption of new material culture styles from the 16^th century BC, prompting arguments for a causal relationship between inward migration and the fall of the tells, an idea now out of vogue [5, 41–46]. This model was based in part on the development of distinctive metalwork of the so-called “Koszider horizon” (ca. 1600–1450 BC) along with the “Tumulus culture” or HGK (Hügelgräber-Kultur) pottery style of the same general date range [44, 47]. Recent work demonstrates that the introduction of these was asynchronous, gradual and embedded within local craft and societal dynamics [48]. The term Koszider is common in the literature but problematic, even unhelpful today, as it has been alternatively linked to hoarding practices, metalwork styles or a period of time, with its relevance for understanding past phenomena shifting according to how the term itself is used [49]. We prefer tumulus culture (with a small c) as a typo-chronological term when dealing with the ceramic styles. Kapuran argues that the impetus to use tumulus culture ceramics was introduced into the area of the lower Pannonian network via two streams from the north across the Mureș and from the south along the Danube [47]. This is based on a bias in distribution of pottery of this style in the north and south, and less-so the middle, of Banat. This is seen for example in the published assemblage from Gradište Iđoš and from its cemetery at Budžak Livade in the north and the cemetery of Karaburma in the south [50–53].

In the Hungarian Danube region, Transdanubian Encrusted Ware develops during the MBA and the theory remains that migration brought this style into the Serbian Danube area, having developed into the related encrusted Szeremle style and influenced the Dubovac-Žuto Brdo style. The Szeremle style developed along the north-south stretch of the Danube corridor and Dubovac-Žuto Brdo style along the same river after its turn to an easterly flow. There are close similarities to both styles during the LBA with regard to motifs, decorative schema and vessel shapes, but the earliest phases of each style are traced to the MBA when they initially draw on related but distinct local traditions. Sidestepping discussion of minor differences for our purposes here during the LBA phase we consider them as a broad stylistic family under the acronym SDŽB. The consumption of SDŽB was focussed on the Danube but it was consumed commonly in the southern reaches of the lower Pannonian network [47, 54, 55].

In general, the southward spread of the consumption of tumulus culture and encrusted ware styles signal mobility. While we may only speculate on whether this was people, things or ideas on the move, it correlates with a disruptive phase at the end of the MBA. Their introduction alongside other ceramic styles is relevant for understanding the late 17^th to 16^th centuries BC as a pivotal time of reorganisation across the southern plain [54].

In the LBA, Belegiš I ceramics emerge as a stylistic tradition ca. 1600 BC and were related to the Vatin and Maros traditions [55–57]. The second phase, Belegiš II emerged in the late 15^th century and is closely related to pre-Gava styles in Hungary [38, 56, 58–61]. These ceramics were consumed throughout the lower Pannonian network, but also as far east as the foothills of the Transylvanian plateau and past the Sava-Drava confluence in the west. They are found less commonly south of the Danube prior to 1200 BC [62]. Notably, the co-occurrence of Belegiš I and SDŽB at TSG settlements, and the common use of incised and / or encrusted decoration, suggests our division between the two, based primarily on cemetery assemblages, is perhaps sharper than would have been viewed in prehistory. When the Belegiš II pottery style emerged, many features such as hanging garland decoration, cylindrical necks with everted rims and four equispaced protomes derived from the Belegiš I style. These features were shared also in a modified format on later SDŽB pottery, demonstrating entanglement of concepts even when broader decorative techniques were adhered to.

Beyond serving as typo-chronological markers, these trends show that the TSG communities were emerging when distinctive pottery styles of Belegiš, SDŽB and tumulus cultural traditions were consumed together. This demonstrated incorporation of the preceding Vatin and Maros traditions, of influences from the Danube and Mureș hinterlands, as well as of styles from non-adjacent areas from farther to the north. The image gained from earlier LBA craftwork / domestic wares is that communities were permeable enough to accommodate influences from varied networks. By the onset of the LBA 2 period ca. 1400 BC, there was increasingly widespread use of Belegiš II style, suggesting that material culture had become a vehicle for cultural integration, bucking a long-standing trend for it to be used to embody difference. Research in the Carpathian Basin and Balkans has often been dominated by culture history / typo-chronological frameworks. It is important to avoid slippage between speaking of ceramic groups as material culture categories with spatial and temporal relevance and the people who made the pots [63, 64]. Ceramic groups cannot viewed as direct proxies for ethnic or cultural identity and political boundaries, but represent expressions of choice and knowledge [65]. Nonetheless, ceramic traditions are linked to geographic spaces and between shape, decoration and function they structured the experiences / habitus of people using them. They were one among many features of daily life that embodied difference between groups. It is important because societies of the Carpathian Basin during the MBA and LBA produced a rich array of ceramic traditions. In aiming to explain, as well as describe, why distribution patterns change in temporal and geographic space we must account for choices and their implications for the materiality of communities. Pots may not equal people, but in this area they could and did express difference.

The middle Bronze age social and settlement background

Many EBA-MBA tell sites in the Pannonian Plain have been known since the earliest days of archaeology in the 19^th century, yet few have been systematically excavated and published, particularly in the east and south Pannonian Plain [37, 48, 66–70]. A sufficient number has been investigated to reveal there was no single foundation horizon. The origins of Bronze Age settlements are staggered at existing (some had been intermittently occupied since the Neolithic) or newly founded tells [66, 71, 72]. This site-type became increasingly common from the later third millennium BC to peak in the middle of the first half of the second millennium BC with respect to site numbers and density of occupation [73]. In the Carpathian Basin 188 MBA “proper tell sites”–defined by Gogâltan [66] as sites with over 1m of continuous occupation stratigraphy–are known. As definitions for what constitutes a tell may be debated, this can be seen as a minimum number.

Tells were a feature common to most, not all, MBA communities in the Pannonian Plain. There remained diversity in material culture and associated elements of death- and lifeways in different parts of the plain [12, 54, 57, 74]. While tells constituted a common settlement form, there was regional diversity in how they were integrated into settlement networks and organised internally [66, 71, 75]. Kienlin emphasises that Bronze Age tell sites represent a conceptual continuity with the Neolithic with regard to use of space and managing this landscape [76]. In essence, the commonality of tell societies lies in their sustained physical and ideological attachment to place. As we discuss below, this frames a sharp discontinuity in LBA ideologies of settlement and landscape.

It has been argued that a three-tier hierarchy existed between settlements in the Pecica polity, with Pecica-Şanţul Mare serving as a primary centre, other tells as secondary centres and smaller, commonly flat, settlements being third in the hierarchical order. In this model, primary centres were locales of craft production, limited centralisation of surplus produce and feasting in the east Pannonian Plain [37, 77]. It is evident that this model of hierarchical relations between tell and flat settlements is not tenable in all areas of the plain [12, 57, 71, 72, 75, 76]. A dearth of systematic excavation at flat settlements, and even the flat components of settlements surrounding tells, creates a bias in the data available to model hierarchies and networks [66, 75]. Current evidence indicates that not all MBA sites were equal in status, longevity, size or density and so a hierarchy of places was evidently recognised in the past [37].

A terminal phase between ca. 1600–1500 calBC is identified at almost all tells and many flat settlements with absolute dates throughout the Pannonian Plain, after which a widespread abandonment horizon has been identified [73]. This was for a long time considered to have been a regional scale collapse, though recent research demonstrates that this transformation was complex, staggered over a century, and that a transition rather than schism can be observed in some areas [19, 66, 71, 76, 78]. As we argue below, new evidence from LBA settlements indicates much of this change was proactive adaptation and innovation in the context of changing political conditions that undermined the systemic logic and power dynamics of southern MBA regimes.

In the northern Pannonian Plain Fischl et al. [12] detect a shift from dense and complex settlement patterns centred around MBA tells to less complex, open LBA settlements, whereas in the Benta Valley, despite a reduction of site numbers and estimated population level, hierarchical ordering of settlements survives into the LBA [69, 79]. In the southeast Pannonian Plain, decline can be detected between 1700 and 1550 BC, with some settlements being abandoned and the density of activity at previously prosperous sites diminishing [77, 80]. Changes in the 16^th century were not sudden, but were socially far-reaching, irreversible, spatially extensive and included virtually all central sites / primary centres / core network nodes and their cemeteries. This unfolded in less than a century–a timeframe consistent with anthropological models of collapse [36, 81, 82]. In our view, what collapsed was the political / ideological regimes, and widespread participation in these. Despite these fundamental shifts in how people organised themselves in settlements and communities, features of daily life such as craft, rituals and diet reveal resilience in macrocultural traits of the general populace. As Johnson [81] puts it “ordinary citizens tend to muddle through transitions and adapt to a new normal,” while those exploiting or directly profiting from higher-order organisational infrastructures tend to be destabilised or disenfranchised.

Around the northern frontier of the area to be occupied by the lower Pannonian network, abandonment of central sites and cemeteries of the Pecica polity occurred between 1600 and 1500 BC [19, 48]. The tells at Mošorin-Feudvar and Belegiš (on the western TSG limit), Židovar (southern limit) and Foeni along with the tell-like settlements of Vatin (east and south-east limit) and Pančevo-Najeva Ciglana and Zlatica–Omoljica east and south-west limit) will be discussed (Fig 1). It is immediately obvious from Fig 3 that 1) these important MBA sites are located on the margin of the area to become densely occupied in the LBA and 2) few MBA sites have been identified within that zone. On current evidence, the dividing line between settlements clustering around the perimeter and clustering within the core of the TSG area is set at ca. 1600–1500 BC. An overview of the abandonment of the above listed sites provides a context to the emergence of LBA sites and helps identify the nature of resilience, transformation and innovation as part of this societal reorganisation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Tumuli (Left) MBA (Centre) and LBA (Right) settlements and cemeteries in topographic perspective (Drawn by Marta Estanqueiro and Caroline Bruyére; basemap hillshade reprinted from https://www.eorcJaxaJp/ALOS/en/dataset/aw3d30/aw3d30_e.htm under a CC BY license, with permission from JAXA -Japan Aerospace Exploration Agency, original copyright 2023).

https://doi.org/10.1371/journal.pone.0288750.g003

Late Bronze age settlement–current knowledge

The LBA settlement archaeology of the wider Carpathian Basin is less well-researched than the MBA and so our knowledge of LBA society is commensurately less well-developed. Much focus has been on typo-chronologies of ceramics and metalwork, commonly derived from cemeteries and hoards, with little evaluation or theorisation of the LBA social and political landscape. In terms of contextualising our survey, therefore, we rely primarily on recent excavation projects that have focussed on higher-order, or simply larger, monumental settlements.

A key reason that so little is known about LBA settlements is their low visibility on the ground–most are simple flat settlements. Indeed, those LBA settlements with visible ramparts were the first to attract attention through excavation in recent years. This monumentality was a key design feature for many LBA settlements, as testified by the extensive use of ditches for enclosures [106]. Nonetheless, when no / small ramparts accompanied them, the infilling of diches over time means they have very low surface visibility today. Surviving upstanding ramparts are rare and are associated with the larger and most monumental sites, the extreme case being Corneşti Iarcuri at ca. 1765 ha [107]. This was the largest construction built in any part of Europe up to that time, dwarfing the citadels of the Aegean world and their surrounding towns. Other impressive sites, while smaller, were nonetheless massive relative to earlier and contemporary sites in Europe, with sites such as Sântana Cetatea Veche, Gradište-Iđoš, Sakule, Crepaja, Bašaid, Csanádpalota, Orosháza Nagytatársánc and Újkígyós all warranting the term megasite or megafort, ranging from 75 to 460 Ha of enclosed space (Fig 1) [50, 89, 108].

Recent excavations within some of these sites have provided new knowledge on LBA building episodes, design of space and domestic economies. At Cornești Iarcuri, evidence for LBA domestic structures and activity more generally is very limited despite its massive size. Food was clearly produced and consumed at the site based on finds of querns, archaeobotanical and zooarchaeological assemblages, but there are no indications as to whether this was permanent or occasional occupancy [109]. Similarly at Csanádpalota, despite its massive size of 460 ha, the evidence for permanent occupation within is very limited [108]. It should be stressed that their enormous size means a tiny fraction of these sites has been subject to excavation and that surface prospection is impacted by soil formation and cultivation activity. At both sites, there remains a possibility that the architectural traditions in a landscape devoid of stone may have resulted in poor archaeological visibility, particularly if they were not burned. The argument for poor architectural visibility is reinforced by the extensive development-led excavations at Șagu, where masses of domestic and industrial refuse were recovered in over 300 pits, but no structures were defined in the open-area excavations.

At the 130ha site of Sântana, located north of the Mureș, exceptionally wide ramparts were constructed that are unique in the area and the excavators point to similarities with Terramare ramparts in northern Italy [19]. Geophysics revealed three large buildings of ca. 50m length on their long axis and between nine and 20 other buildings in the inner enclosure (I), demonstrating the possible presence of monumental buildings, but also a notably different use of space than at TSG sites. The outer enclosures (II and III), however, have very low density buildings (of unknown date) indicating a similar concern with maintaining enclosed open spaces [89, 110]. In southern Hungary north of the Mureș river, the LBA sites of Orosháza Nagytatársánc (two enclosures ca. 105ha + 85ha, overall 190ha) and Újkígyós (ca. 70ha) are very large enclosed settlements that have many visible pale activity areas and multiple enclosures. Although their spatial organisation is slightly different, they incorporate a small enclosure found at many larger TSG sites.

At Gradište Iđoš, excavation has focussed on the enclosure ditches, yet the activity areas discussed below are indicative of domestic structures clustered together, leaving large parts of the enclosed space apparently unoccupied. Two habitation areas are defined by ditches along with an embanked citadel area and these are entirely enclosed by further ditches, defining a space of up to 200 ha. Gaydarska and Chapman argue that the absence of evidence for buildings at Gradište Iđoš should be read literally as evidence of absence and they correlate this with the dearth of visible structures at Cornești Iarcuri and Csanádpalota to suggest these sites were not settlements [111]. They contend that the primary function of the largest enclosed sites was as a place of assembly. However, viewing Gradište Iđoš as one of 100+ settlements in Banat with similar occupation traces defined by pale ‘activity areas’, and with reference to Orosháza Nagytatársánc and Újkígyós in Hungary, their theory cannot serve as a rule for all large enclosed sites / megaforts of this area.

Turning to cemeteries, of those in use during the MBA in the broad environs of the TSG, most were abandoned by 1600 BC [41, 48]. Unfortunately along with the dearth of MBA settlements, there is a dearth of known cemeteries of MBA date in the immediate proximity of TSG sites. Nonetheless, the establishment of new cemeteries after 1600 BC is a robust statement of new death and lifeways characteristic of the LBA and we can identify many that were established then and used until ca. 1200 BC (Fig 3). Only a small few continued to be used at a reduced level after that time [38].

The relationship of LBA settlements and cemeteries to the tumuli situated within and around them remains unclear because no tumuli have been excavated and published in the hinterlands of the TSG [112]. Despite the lack of chronological determination, if we accept they could be prehistoric and pre-LBA, their spatial relationship with LBA settlements and cemeteries suggests that tumuli played an important ideological role in LBA social landscapes (e.g. Figs 3, 4 and S3 File). For example, the flat cemetery of Budžak Livade beside Gradište Iđoš lies along the path of a tumulus alignment, which in turn follows a paleochannel. Tumuli throughout the area commonly align with water channels, with the greatest density associated with the largest East-West flowing water courses, with important alignments terminating near to the megaforts of Gradište Iđoš and Sakule (Fig 3). Considering excavated tumuli from the wider region, it is probable that large diameter and tall examples date to the 4^th to early 3^rd millennium BC and some of those with 10-20m diameter and low mounds (1–1.5m in height) may even be LBA in date. This possibility is based on the same metalwork styles found in tumuli with those proportions dated to the 14^th to 13^th century calBC in areas south of the TSG along the Western Morava and tributaries of the Sava to the south and Susani-Grămurada at the edge of the Pannonian Plain to the east of Cornești Iarcuri [58, 113–116].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Central area in TSG site of Baranda showing 1) Tumulus, 2) watercourse and 3–5) activity areas aligned with tumulus (Photograph by Barry Molloy).

https://doi.org/10.1371/journal.pone.0288750.g004

All size ranges of tumuli were built in close proximity to each other, occurring both within and immediately adjacent to many TSG sites. This defines a complex palimpsest landscape of settlement and mortuary sites. Whether tumuli are all older or if some are near-contemporary to the settlements, we see co-location as stating a claim–recent, invented, partly-constructed or entirely appropriated–to an ancestral landscape that had not been related to tell communities. LBA communities could thus “borrow” legitimacy by using places of the dead. Establishing flat cemeteries similarly created new ancestral relationships in time and space. These moves allowed new narratives of landscape, continuity and legitimacy to emerge based on strategic locational choices. This was conceptually in opposition to the continuous occupation of domestic spaces that had been a defining characteristic of MBA tells.

Characterizing the tisza site group

The 100+ TSG sites identified thus far in Serbia align broadly on a north-south axis on the east side of the Tisza-Danube north-south corridor. Sites with identical design occur immediately to the east between the Bega and Timiş rivers in Romania (Fig 3). They are located in similarly low-lying parts of the landscape, and whilee not visited on the ground as part of this project, their chronology has been confirmed by Dorogostaisky and colleagues [119, 120]. A handful of sites of this form have been identified thus far on the west side of the Tisza in Serbia, in Bačka county, including the settlement at Turija which had features dated to LBA 2 based on ceramics and absolute dates (Fig 1).

On the basis of aerial imagery, TSG sites range in size from ca. 6 ha to as much as 200 ha. Based on the interquartile range from all known TSG sites, the visible footprint of activity derived from remote prospection at most sites is in the range of 10–50 ha (Fig 8). In the case of unenclosed sites at the lower end of this spectrum size is calculated according to the minimum visible activity zone (extent covered by activity areas) and excludes the equivalent ‘empty’ space that lies between activity areas and boundary ditches at enclosed sites. The smallest defined site had only 2 (possibly 3) activity areas at Kumane 3 and the largest is Gradište Iđoš at nearly 200 ha which has multiple enclosures and two distinct contemporary enclosures surrounding activity areas.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Size and visible features of TSG and TSG-like sites in Banat and east Backa.

All definitions refer to features visible in aerial imagery. Single-enclosure sites have one complete or partially visible ditch outside of activity areas. Multiple-enclosure sites have two or more complete or partially visible ditches. Sites with no visible enclosures refers to detection using remote prospection visibility only and has not been ground truthed. (Map by Marta Estanqueiro and Caroline Bruyere; basemap hillshade derived from ALOS DSM AW3D30, reprinted from https://www.eorcJaxaJp/ALOS/en/dataset/aw3d30/aw3d30_e.htm under a CC BY license, with permission from JAXA -Japan Aerospace Exploration Agency, original copyright 2023).

https://doi.org/10.1371/journal.pone.0288750.g008

Enclosed settlements are most commonly sub-circular to sub-oval in plan (Fig 9), though they are occasionally sub-rectangular with one or more straight boundaries (e.g. Dobrica and Bavanište). Ditches are commonly V-shaped in profile, in some cases shifting to U-shaped near to entrances, and are usually between 2 and 3 metres deep (from the modern field surface) and 3 to 6 metres wide at the top. Some enclosures can be irregular in shape when utilising landscape features, including watercourses and slopes (e.g. Melenci, Torda 2, Mramorak). Sites can also include a combination of these landscape features (e.g. Gradište Iđoš, Jankov Most, Novo Miloševo). Around 50% of sites, usual smaller ones, have no enclosure visible in aerial images, 30% of sites have one enclosure, while larger and more complex sites can have up to four concentric or sequentially accessed enclosures, as seen at Crepaja, Sakule, Bašaid, Mokrin and Gradište Iđoš for example (Figs 3 and 10). An estimated 30–40% of the megafort at Crepaja has been eroded in the northeast through the migration of a now-extinct minor watercourse. XXXX MOVE FIGURE REF HERE XXXXXXXXXXXXXX? Fig 9: Outline of a selection of TSG sites showing range of organisational features including ditches and natural watercourses. (Drawn by Caroline Bruyere)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Outline of a selection of TSG sites showing range of organisational features including ditches and natural watercourses.

(Drawn by Caroline Bruyere).

https://doi.org/10.1371/journal.pone.0288750.g009

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10.

1) 5km radius range around TSG sites. 2) Distance-based cumulative distribution functions. For F, G and K functions—black lines: observed values; red lines: theoretical values; gray values range: Monte Carlo envelopes considering 99 simulations. For J function—blue line: theoretical values; black line: Kaplan-Meyer estimate; red line: Hanisch estimate; green line: reduced-sample estimate. (Map and graphs by Marta Estanqueiro; basemap hillshade derived from ALOS DSM AW3D30, reprinted from https://www.eorcJaxaJp/ALOS/en/dataset/aw3d30/aw3d30_e.htm under a CC BY license, with permission from JAXA -Japan Aerospace Exploration Agency, original copyright 2023. Graphs plotted using R 4.2.0).

https://doi.org/10.1371/journal.pone.0288750.g010

A functional hierarchy can be recognised in site design, whereby the larger sites are invariably more complex than smaller ones and they invariably have spatially more extensive and intensive surface traces of activity (apart, perhaps, from Bašaid). Internal divisions are defined by enclosures visible either in aerial imagery or through geophysical survey [50]. The variety of site locations, from gentle slopes to isolated islands on the edges of wetlands, may also suggest that apart from size and complexity, different sites had different roles or special purposes within this network.

Activity areas commonly lie 20–40 metres apart and are rarely more densely spaced or overlap. They occur from the geometric centre of a site in a cluster and their distribution stops at least 40, and up to 100, metres away from enclosing ditches (when present). This indicates that a specific zone with no activity areas was usually retained around habitation loci but within the boundaries of a settlement, creating intentionally “empty” spaces. Fernández-Götz has recently discussed the relevance of such divisions and empty spaces for understanding complexity at non-urban settlements [128, 129]. The density of activity areas is relatively consistent across sites and so the number usually, but not as a rule, correlates with the size of the enclosed area if present. This trend can be broken when sites are spread along a watercourse rather than clustering around a spot beside it, as seen at the Jankov Most, Novo Miloševo or the Melenci clusters of sites. Intensive systematic surveys of a sample of settlements conducted by C. Bruyere and geophysical survey data by J. Pendić and G. Topić will shed further light on spatial layout and density of activity within some enclosures in publications in preparation.

All sites are set within active agricultural lands today that are subject to ploughing, thereby damaging them annually and impacting on preservation and visibility of features. Despite this, at some sites activity areas were slightly raised above the surrounding surface indicating that many, and perhaps all, had once been positive features standing proud of the surrounding ground level. Test excavations at Sakule confirm that activity areas were a locus for sequential phases of activity, similar in concept but different in execution to intergenerational modes of tell occupation.

TSG sites are densely spaced in the landscape, and the vast majority lie within 5km of at least one other known site (Fig 10.1). Their close proximity, and the fact that the catchment areas (5km radius) overlaps in most cases, suggests that the communities that lived there were closely related, most probably through kinship but certainly politically. This clustering created a very dense societal network that controlled not only the surrounding territory at the local scale but this entire landscape at the larger regional scale, including resources and communication routes. This in turn demonstrates the presence of a complex political system that integrated, managed and served the needs of multi-local communities living within and around hierarchically, and possibly functionally, differentiated settlements.

At a visual level, Fig 10.1 suggests that, similar to the distribution of activity areas within a settlement, the distribution of settlements in the landscape was not random. In order to evaluate if this was systematic clustering, several spatial functions were estimated and plotted. From a statistical perspective, a completely random spatial pattern can be described by a homogeneous Poisson process with intensity λ, so we computed several distance-based cumulative distribution functions to explore the spatial distribution of settlements and their departure from this hypothesis.

For this assessment, G (nearest neighbour function), F (empty space function), J and Ripley’s K (detection of spatial patterns at multiple distances) functions were chosen (Fig 10.2). The first three are based on nearest neighbour distances, while the K function is based on distances between all pairs of points (settlements). Point-wise Monte Carlo envelopes were also computed. As we can observe in Fig 10.2, the empirical lines for settlement distribution indicates a trend towards an aggregated/clustered pattern. The F function assumes that if the settlements are randomly spaced, so is the empty space between them. Therefore, if a clustered pattern is present, the empirical line would be below the theoretical one, as seen in Fig 10.2A. The opposite can be said of the G function, where an empirical line above the theoretical one indicates clustering (Fig 10.2B). The J function is a combination of the previous two, and values of J(r) < 1 would suggest clustering (Fig 10.2C).

In order to assess the spatial pattern at different scales, Ripley’s K function was also used. This can help detect more complex patterns at different distances, not only at nearest-neighbour scale. In this case, the Fig 10.2D suggests a clustering trend at multiple scales. For our data we observed that until approximately 3km the empirical lines followed the theoretical ones, after that, there are indications of clustering. We can also see that most settlements have a nearest neighbour within a 6km (approximate) distance. This tells us that site distribution was unlikely to be random and that close spacing / proximity was a function of a social strategy more than the unplanned emergence of settlements within each other’s hinterlands. An element of horizontal movement of communities is plausible within the duration of our quite broad chronological blocks of 100–200 years, though this would not explain the relative uniformity of the clustering across this social landscape.

The gap between a northern and southern cluster of sites in the land between the Bega and Timiş rivers appears to be an ancient phenomenon. Whether that was socially and / or topographically dictated is unclear, but the clustering of tumuli along the southern bank of the Timiş suggests this line was meaningful in prehistory (Fig 3). Following the chain of TSG-type sites between the Bega and Zlatica rivers to the east in Romania, the easternmost sites are situated in the immediate vicinity of the megafort of Cornești Iarcuri. The settlements of Carani, Biled and possibly Satchinez Râtul Popii are of TSG type and lie 8, 24 and 19 km respectively to the west of Cornești Iarcuri. Despite the uniqueness of this latter site, its location at the end of this chain of sites with similar designs indicates a social relationship.

Climate change and the new domestic landscapes of the Tisza Site Group

Nicodemus and O’Shea [77] observed that MBA tells around the Mureș and Tisza rivers commonly targeted landscape features that were raised, and Ljuština [57, 101] argues the same for Židovar, thereby occupying natural islands within a marshy and wet hinterland. This created spatial discontinuity by land, emphasising the importance of riverine transport. We might assume–on the basis of settlement distribution alone–that certain areas of the landscape in the MBA were undesirable due to seasonally or perpetually marshy conditions. For example, this avoidance may have arisen from (perceived) poor suitability for their farming conventions or weaker connections to major rivers, which were important elements in MBA lifeways and communication networks.

Locational choices, including use of landscape features, changed substantially in the LBA as areas with previously low-density occupation became very high-density settled landscapes. Many TSG sites incorporated water channels running through them. In cases where water channels did not pass through sites, they commonly ran immediately adjacent (Table 2). Particularly in the southern cluster, where a terrace is present with flat areas above and below, sites are commonly located at the base of the slope separating the two levels, for example at Bavanište, Dobrica, Pančevo, Pančevo 2, Dolovo, Mramorak, and Mramorak 2. When a more gradual slope is present, some sites have elements at the base and along the slope, for example Kačarevo 2 (Fig 7) and Sakule. At major sites such as Crepaja and Sakule, as well as smaller sites, substantially higher ground on sandy terraces lies within ca. 5km to the east, but these were evidently rejected as settlement loci, with few exceptions e.g. Bavanište 2. The elevation variance between the base and top of terraces close to sites commonly exceeds 8 metres height differential. The characteristic ditches that surround sites may have been intended to aid drainage, although this was unlikely to be their primary function in light of water channels passing through some sites and others ditch circuits being built on sloping ground (Fig 11).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Aerial image of Mramorak showing water courses, activity areas and site location at base of higher terrace (Photograph and markup by Barry Molloy with thanks to Darja Grosman).

https://doi.org/10.1371/journal.pone.0288750.g011

The topographic location of most sites is either within or beside some of the lowest points in the landscape and this landscape is in turn one of the lowest lying and flat expanses of the Pannonian Plain, which is the lowest lying part of the Carpathian Basin (S2 File). Remote imagery reveals myriad palaeo-channels alongside natural and artificial waterways active today. While geomorphological investigation into which waterways were active in the Bronze Age has not been undertaken, the location of prehistoric tumuli along smaller channels indicates many were flowing in the Bronze Age. Historic maps and records for the area demonstrate that much of the province of Banat was subject to flooding and standing bodies of water prior to large-scale drainage works in the 19^th century AD and subsequent upgrades [130, 131]. The levees on the River Tisza were constructed to a height of 5 metres to prevent flooding and lateral movement of the now culverted water course [132]. Prior to this, over 35–50% of the Serbian province of Vojvodina, covering much of the southern Pannonian Plain, was under water, marshland or regular threat of flooding, whereas today 92% is under cultivation [130, 133]. Even so, flooding remains a regular seasonal threat in the south Pannonian Plain, particularly in low lying areas including former marshes, cut-off channels and oxbow lakes of old river courses. The risk of flooding is, and must have been in prehistory, unpredictable with some years being considerably more difficult to navigate than others. The development of soils in the Banat region supports high groundwater levels [134]. Retention of surface water due to a high water table after a rainy period was a further potential risk for the inhabitants of the TSG sites, perhaps greater than flooding of major nearby rivers. Historical records also show a contrary pattern whereby a substantial reduction in rainfall can have substantial short-term impacts on soil hydration and productivity due to the interconnected nature of waterways and their relevance to the water table across a wide area [133]. Most TSG sites are located on deposits formed in the Late Pleistocene, indicating alluvial deposition from flooding was not a major factor shaping Holocene topography at those locations [135]. Alluvial deposits of Holocene age have been documented near to the sites along present-day river courses.

With the low-lying topography and relationship to a network of water channels and possible seasonal or perennial open bodies of water in mind, the builders of these sites established a wetland-oriented settlement strategy which took advantage of the particular environmental conditions this offered. De Marinis [136] and Dalla Longa [137, 138] both observe for the contemporary Po Valley a choice to occupy low-lying areas associated with water courses (Terramare sites) or to build houses on stilts within areas of standing water (Palfitte sites). This reflects what Dalla Longa calls [138] “a clear intention to set up settlements along main river courses or on their banks … sites are located near or within wet basins, following the “wetland-oriented” settlement strategy well-known in literature”. This similarity in settlement strategy may be a further reflection of systematic cultural, ideological and probably economic exchanges linking these two exceptionally large and flat plains [139–141]. Furthermore, this switch from the avoidance of low-lying areas in the MBA to embracing them in the LBA can be read as changes to worldviews and cosmologies, potentially including the integration of previously liminal spaces as inhabited spaces. The highly similar locational choices for almost all TSG sites, however, also reveals a dearth of risk spreading, particularly in light of their apparent co-dependence.

In the case of the TSG, these settlement strategies suggest that communities exploited a finely balanced ecosystem that was enabled by the prevailing climatic conditions between 1500 and 1200 BC. Though requiring systematic environmental research, paleoclimate markers for increased rainfall indicate that areas that had been marginal marshy lands could have become more predictable bodies of water, thereby becoming a resource rather than a liminal or hazardous space. This hypothesis may also be supported by the relative stability of hydroclimate markers of the LBA relative to the MBA, which indicate conditions that would have reduced the risks associated with settling in close proximity to water courses and wetlands. As a cautionary note, it should be emphasised that the stable isotope data from the cited speleothem records primarily relates to wintertime rainfall levels and we lack proxies specific to summertime rainfall or temperature levels [21, 142]. If higher water levels were sustained annually this would ensure hydration of soil during the comparatively hot summer months, a phenomenon identified in other studies [35]. Future analyses of chironomid head capsules from wetlands, for example, may provide insights into summertime conditions in the Middle and Late Bronze Age to further explore this [143–145]. We argue that the prevailing climate in the 16^th century BC encouraged decisions to engage in novel land use regimes. Emergent social strategies no-doubt balanced myriad societal factors, so that climate and environment should be read as enablers operating at a multi-decadal scale that facilitated, but did not determine, adaptive human choices [30, 146, 147].

There are insufficient environmental data to model the conditions in the TSG landscape at this time, but the settlement location choices themselves tell us two things about the environmental conditions. First, they were stable enough to allow sites to be located at low-lying elevations (relative to water courses) without substantial risk of flooding, indicating reasonably predictable conditions. Second, the relationship of people to their environment was different to the Middle Bronze Age, during which time these same topographic locations were actively avoided.

In the highly similar environment of the Po Valley, it has been observed that the cooler (not cold) but humid conditions of 1500–1200 BC created an environmental niche in which particular forms of society developed. The practices and ideologies of landscape use were tailored to supporting specific needs of communities within a specific hydroclimate and landscape hydrology niche. In the Po Valley, climate conditions with predictable winter rainfall were paired with a comparatively high water table in the LBA (Recent Bronze Age) relative to the MBA and Final Bronze Age [148]. Looking from the beginning of the TSG phenomenon, pioneering groups must have demonstrated an attractiveness to the new lifeway in this environmental niche that became attractive to others over time, leading to this unprecedented density of settlements.

There are convincing grounds to argue that climate change played a role in how change from MBA to LBA social systems unfolded and again during the time of the decline and abandonment of TSG sites from the 13^th to 12^th centuries BC. The decline of MBA communities was protracted and socially driven, but the changing climatic conditions were arguably less favourable to their modes of using the landscape. This was not a catastrophic change and indeed it may even exemplify resilience and adaptability, whereby societal scale change was needed to successfully adjust to a new normal [34, 149]. The MBA system had long-since entered a crisis state by the 16^th century BC and so rather than narrowing options, climate change appears to have broadened them providing alternative pathways to prosperity. These were readily exploited but appeared to correlate with a fundamental reorganisation of settlement and society. LBA settlement locations led to infilling and linking up of areas between the distinct tell-networks (or perhaps polities) characterised by the use of Maros pottery in the north and northwest and Vatin pottery in the south and east. There was evidently sufficient agricultural land located around LBA settlements to sustain this very dense network of communities.

A second factor influencing site location appears to have been the desire for colocation with pre-existing mortuary landscapes, tumuli in particular (Fig 4).This may have enabled LBA communities to differentiate their ideologies from those of MBA entities and create a neutral linchpin to aid integration of a heterogenous population base. An association between tumuli and watercourses is also evident, for example in the hinterland of Glogonj, Mramorak, Novo Miloševo 4, Idvor, Melenci 2 and Mokrin. This suggests a three-point relationship for settlement choice in many cases, requiring watercourses and tumuli. Concentrations of tumuli along water courses served a dual purpose as navigation points marking river-land cross-roads and as boundary markers signalling control over rights to disembark boats or to cross between politically differentiated areas.

Relative chronology

The chronology of the pottery we collected during site visits confirmed those exhibiting all TSG features were all LBA 1–2 in date. With very few exceptions, those that had clustered activity areas (with no ditches) were also LBA 1–2 in date. It is unlikely that all TSG sites were occupied at the same time, and there must have been an element of horizontal movement of communities building sites. However, if this occurred within a short space of time–even decades–the resolution of neither the absolute nor relative chronologies is precise enough to document this. The regular spacing observed between ‘activity areas’ within the confines of each site suggests that all, or most, domestic spaces were occupied simultaneously or within living memory.

Our dating of sites relies on diagnostic surface finds (Fig 5, Table 2 and S4 File). Belegiš I and II pottery are found throughout the TSG area. As noted by Szentmiklosi [55] and Falkenstein [94], SDŽB pottery occurs in the areas surrounding the TSG in the south, and our survey demonstrates it was consumed quite commonly within the southern cluster of sites. Distribution of ceramics of this broad family of encrusted pottery can be mapped now in an area extending from the Danube to south of the Bega river. It is present, but less frequent, in the northern cluster of TSG sites and is exceptionally rare in the northernmost TSG sites.

Considering our survey finds, there is a greater frequency of sites that have only Belegiš I / SDŽB pottery along with relatively fewer or no sherds of Belegiš II in the southern site cluster (Table 2). It is emphasised that this can be a bias of the recovery of diagnostic sherds, as only a handful of diagnostics were recovered from sites that only had Belegiš I pottery. Nonetheless, the difference in assemblages between sites north and south of the Bega appears to be a real phenomenon. This can be explained in two ways. The first is that the making of Belegiš II pottery and displacement of Belegiš I occurred gradually during the 14^th century BC, with the stylistic trend spreading from the north while SDŽB was still being made in the southern cluster, thereby blurring or offsetting the LBA 1–2 ceramic transition more there than it does in the northern cluster. The other possibility is that a higher proportion of sites in the south were abandoned before LBA 2.

Pottery indicates that the earliest TSG sites in both clusters were occupied early in LBA 1 and that there was an increase in activity, and possibly site construction, at the beginning of LBA 2. Based on both relative and absolute chronologies, all TSG sites appear to have been abandoned in the decades around 1200 BC. Of these, only the largest sites at Gradište Iđoš, Sakule and possibly Bašaid have indicators of reoccupation in small areas during the 11^th to 10^th centuries BC. The site of Jaša Tomić is an exception as this exhibits TSG features but was (probably) a new foundation of 12^th to 10^th centuries BC based on pottery recovered.

A common feature at TSG sites was reoccupation in either the Sarmatian (final centuries BC) and / or Medieval periods (broadly 1100–1400 AD). The Sarmatian and Kumani (11^th to 13^th centuries AD) groups are known to have extensively practiced mobile pastoralism [150, 151]. Re-use of prehistoric tumuli and construction of new ones in this area is a phenomenon of the 11^th-13^th centuries AD, potentially related to the documented arrival of non-Christian Kumani groups from the Steppes [50]. Other periods were very rarely represented at TSG settlements–notably the Neolithic–potentially indicating related concepts of landscape and land use occurred only at specific points in the history of the region.

Absolute chronology

Defining an absolute chronology for the TSG sites is difficult given the number of sites, their complexity, and their massive size. This is compounded by the dearth of scientific excavation. Absolute dates from well-stratified contexts at Gradište Iđoš and the nearby cemetery of Budžak Livade provide a general framework for the occupation history of one such community. Ceramics from graves with absolute dates include vessels of Belegiš I and II and of tumulus culture forms, typical to the LBA assemblages of the northern TSG area. This LBA 1 and 2 chronology is contrary to Daróczi et al.’s [152] recent unsubstantiated re-attribution of tumulus culture ceramics from this cemetery to the MBA. The cemetery was established by 1600 calBC and it continued in use into the 13^th century calBC (Fig 12 and S1 Table). The earliest occupation thus far excavated at the settlement is 15^th century calBC and there are no dates later than the decades around 1200 calBC, until a limited 10^th-9^th century calBC reoccupation (Fig 13). In his regional survey of absolute dates, Sava defines a 16^th century calBC date for the first sites using Belegiš pottery, with increased numbers in the 15^th century calBC and a rapid drop off of sites in the 13^th to 12^th centuries calBC [38]. This drop off in site numbers after 1200 BC is also suggested in survey and some excavation data for nearby parts of southern Hungary [41, 108, 153]. Absolute dating at other sites in the TSG network, while the product of small windows of excavation, reveal a complementary picture. Details of sample type, context and uncalibrated date are provided in S1 Table. All necessary permits were obtained for excavation and sampling from the Ministry of Culture and Information of the Republic of Serbia, complying with relevant regulations for archaeological research (Permit numbers: 631-02-154/2021-02, date 08.11.2021.; 631-02-94/2019-02, date 28.10.2019.; 631-02-146/2022-02 date 27.10.2022).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 12. Pottery from dated contexts at Budžak Livade.

Grave 2: 3, 6; Grave 9: 4, 16; Grave 12: 1; Grave 22: 10, 15; Grave 23; 18; Grave 24: 2, 5; Grave 26: 12; Grave 29: 17; Grave 31: 8,9; Grave 32: 14; Grave 33: 13, Grave 36: 7; Grave 37: 11 (Drawn by Dragan Jovanović).

https://doi.org/10.1371/journal.pone.0288750.g012

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 13. Absolute chronology of TSG settlements and cemeteries.

(Figure by Christopher Bronk Ramsey).

https://doi.org/10.1371/journal.pone.0288750.g013

At the settlement of Sakule, test excavations were focussed on activity areas. These returned dates covering the full duration of the use of Budžak Livade and Gradište Iđoš (Fig 14). The site appears to have been first occupied at the MBA to LBA transition and dated contexts cover most centuries until ca. 1200 calBC. After that, the site appears to have been abandoned until the reoccupation of a small area from ca. 1000 to 800 calBC. A similar hiatus is observed at Cornești Iarcuri [16]. Dates from an activity area in the TSG site of Stari Tamiš reveal occupation at the turn of the 16^th to 15^th century calBC and the settlement of Bavanište was occupied since at least the 15^th century calBC. The inhumation burial from Mali Akač has an uncertain specific context, but is probably from beside a tumulus in close proximity to Novo Miloševo 4 settlement and this skeleton is dated to the 14^th century calBC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 14. 14C phase model for Gradište Iđoš and Budžak Livade (Figure by Christopher Bronk Ramsey).

https://doi.org/10.1371/journal.pone.0288750.g014

Looking at absolute dates from the wider LBA settled landscape of the south Pannonian Plain, the increased intensity of activity from ca. 1400 calBC seen in dates from Gradište Iđoš is reflected in the chronology of other well-dated enclosed sites at Sântana Cetatea Veche, Corneşti Iarcuri, Csanádpalota and Turija Gradište as well as smaller unenclosed sites in Romania [38, 154]. There appears to be evidence for burning / and or infilling of ditches at several excavated sites in the earlier 13^th century calBC. Activity continues at all other well-dated sites in the 13^th century and at some possibly into the early 12^th century calBC. At the site of Iđoš 2, 14C data show it was occupied into the 12^th century calBC [155].

Absolute and relative dates suggest that TSG sites and cemeteries–and indeed most of the lower Pannonian network–were abandoned en masse around 1200 BC. It is noteworthy that this was precisely the time of the contraction of settlement on the Titel plateau and reoccupation of the tell at Mošorin-Feudvar. A similar pattern, based on surviving strata, is probably detected at Židovar, which was resettled. No re-occupation horizon is observed at tells in the area of the Maros group in the 12^th century BC. These changes indicate that knowledge of tell sites was sustained in societies during LBA 1–2 but they were marginalised as loci for settlement until another major societal shift took place after 1200 BC through which loci for LBA settlements were rejected.

TSG collapse

The low-lying location of TSG sites within a very flat landscape afforded them low diversity and exposed them to highly similar risks, for example flooding. It was argued above that they emerged within a specific and relatively stable climatic niche and exploited this landscape in an extensive manner never practiced before then. The relationship is not directly causal; we argue that change in land use was opportunistic rather than forced and ultimately supported rather than drove changes in settlement organisation. Apparent climatic shifts ca. 3.6/3.5 and 3.2 kya indicate that the LBA was bookended by warmer and drier conditions [191–193]. These shifts were not extreme and nothing indicates they caused catastrophic social change. Taking a decadal perspective, however, it is plausible that the increasing aridification recorded in the 13^th to 11^th centuries BC incrementally compounded social stressors, becoming a force multiplier for other causes of social conflict. The exceptionally dense and co-dependent settlement regime of the TSG may itself have laid the foundation for competition and conflict to arise if resources became scarcer, even marginally so. The finely balanced TSG landscape is a good candidate for the classic overshoot collapse model where increasing productivity and population co-occur in a wet cycle only to be undermined by decreased productivity and capacity to sustain a population in a dry cycle [34, 194]. Cautious of the balance between documenting correlation and causation, it is salient that both settlements and cemeteries were abandoned en masse in the 13^th to early 12^th century BC in the lower Pannonian network (Figs 12 and 13). Social change occurred, and it was devastating for these communities. We may find hints of the responses to this crisis in the material record. The deposition of hoards of broken bronze objects is commonly considered to have been a response to threats to individuals (e.g. violent conflict) or whole communities (e.g. environmental changes) or as votive offerings (e.g. propitiating divine forces) across time and space in European prehistory [195–198]. A massive increase in hoarding in the south Pannonian Plain is documented precisely from the 13^th century to 12^th centuries BC [199–203].

Settlement location choices left the social system of the TSG in an inherently fragile state arising from a dependency on particular hydrological conditions [29, 31, 34, 147]. This is not to say people were not adaptable, but rather deviation from a given system in order to be adaptable undermined the organisational logic of the system. The level of change to tip such a fine balance and push TSG communities beyond their stability threshold appears to have been relatively small [204]. One hypothesis would be that a need to seek out resources beyond the cramped hinterlands of the TSG network led to a doubling down on mobile pastoralism as a lifeway. Whole communities could have taken advantage of mobility to maximise their productive potential in a changing physical landscape. This would have led to gradual disaggregation between lifeways and political systems, potentially over decades, but constituting a tumultuous transitional period, what Raffield has recently termed a “shatter zone” in social trajectories [205]. The pieces remained, but their order was broken.