2.1 Natural settings

The Pergamon micro-region is characterized by a Mediterranean climate (Csa) with hot, dry summers and mild, wet winters (Köppen–Geiger classification [43]). The annual precipitation cycle exhibits an unimodal rainfall distribution, with its monthly minimum occurring in August (5 mm on average) and its maximum in December (103 mm on average). The rainy season lasts from October to March (Fig 2b). The annual precipitation sums up to 620 mm. The mean annual temperature is 16.7 °C (World Meteorological Organization Station 17742 in Bergama for the period of 1991–2020 [44]). Coastal areas around Dikili in the vicinity of the ancient Atarneus and Zeytindağ, close to ancient Elaia, experience maritime influences with higher average annual temperatures. In contrast, the hinterland around Paşaköy, located near ancient Allianoi, exhibits more continental temperature patterns (Fig 7). The regional wind patterns are mainly influenced by the Etesian winds (tur meltem); the strong, dry, northerly winds typically occur from mid-May to mid-September and play a dominant role in shaping the weather system around the Aegean Sea [45].

Download:

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

Fig 3. Overview of the agroclimatic potential modeling framework. The model incorporates climate, topography, and soil data at a resolution of approx. 1 x 1 km. The EcoCrop parameters comprised resampled soil, topographical, and climatic inputs derived from TanDEM-X, SoilGrids, WorldClim 2.1, and the MPI-ESM 1.2 (paleo-climate model). The soil data was validated using field data from [67]. Climate anomalies are interpolated over five snapshots, with the outputs validated by pollen-based climate reconstructions derived from LegacyPollen 2.0 records [57].

https://doi.org/10.1371/journal.pone.0325779.g003

Download:

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

Fig 4. Comparison of pollen-based climate reconstructions and MPI-ESM 1.2 simulations.

a) Comparison between z-scores of modeled climate variables from the MPI-ESM 1.2 simulation and the pollen-based climate reconstruction between 400 BCE and 400 CE. Both models are compared using the dynamic time warping (dtw) algorithm. b) Alignment paths between both reconstructions for temperature and precipitation values.

https://doi.org/10.1371/journal.pone.0325779.g004

Download:

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

Fig 5. Overview of the six most conventional crops in the environs of Pergamon.

Changes in cultivation suitability over the selected five time periods (310-290 BCE, 10 BCE-10 CE, 100-120 CE, 240-260 CE, 340-360 CE), highlighting periods with different climate conditions. The hybrid model incorporates the modeled suitability of the six conventional crop types, including four cereal and two legume types. The crop-specific outputs were trimmed to prevent overrepresenting nearly unsuitable areas in the hybrid model, considering only suitability estimations >20%.

https://doi.org/10.1371/journal.pone.0325779.g005

Download:

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

Fig 6. Change of the areal share of different suitability classes.

Suitability classes (stacked lines) are in relation to the total area from crops combined in the hybrid model. Bar plots depict the class-specific changes in suitability between periods, showing fluctuations in cultivation suitability over time.

https://doi.org/10.1371/journal.pone.0325779.g006

The topography of the western Anatolian coast is characterized by an east-west oriented graben system that built up since the Early Miocene [46]. The environs of Pergamon are shaped by the Bergama graben; the Bakırçay River, the ancient Kaikos, is the main receiving stream and flows through this graben. The micro-region exhibits a mix of sedimentary, metamorphic, and igneous rock formations, contributing to a heterogeneous landscape. The Karadağ, Kozak, Madradağ, and Yüntdağ mountain ranges surround the Bergama graben. Elevations range from the sea level in the west to about 1800 m a.s.l. in the Madradağ mountains. The Bergama graben is alluviated by the multiple tributaries of the Bakırçay, forming shallow alluvial fans as they enter the graben and an extensive alluvial plain. Morphologically, the alluvial areas of the Bakırçay River are categorized into two distinct regions: the upper and lower Bakırçay plains. The upper Bakırçay plain is located upstream of the river’s entry into the Bergama graben, southeast of Apollonia (Fig 1). The lower Bakırçay plain refers to the alluvial area within the Bergama graben. Pergamon is situated in the center of the Bergama graben, marking the division of the lower Bakırçay plain into its western and eastern parts, which is commonly used for its morphological description.

The mid- to late Holocene landscape of the region is characterized by a polygenetic nature, primarily influenced by climatic shifts and human activities [8,47]. Sediment profiles across the Pergamon micro-region reveal several sedimentation phases with substantial changes occurring around 4 ka BP, attributed to supra-regional aridization and rapid climate events, compounded by the onset of local human activities [14,47]. Unlike other prominent graben systems of the western Anatolian coast [48], there is no evidence of a Holocene estuary in the Bergama graben that silted up after the Holocene transgression. Nevertheless, the harbor of Elaia east of the Bakırçay delta (Fig 1), which is regarded as a maritime satellite of Pergamon [49–51], bears traces of human-accelerated siltation. Here, phases of sedimentation are associated with peak periods of local land use and are particularly characterized by deforestation, increased agriculture, and the cultivation of olive trees during the Roman Imperial period [51–53].

The current geomorphological and hydrographic context of the Bakırçay catchment area is characterized by hydraulic engineering interventions [39]. In addition to the channelization and diversion of the Bakırçay in the 20th century, its eastern plain, in particular, was subjected to extensive consolidation. To create a plain that could be agriculturally used efficiently, large-scale drainage and irrigation canals were constructed, Bakırçay tributaries were channelized, small hills in the plain were removed, and hundreds of wells were built [54]. From the mid-20th century onward, several rivers were dammed to facilitate irrigation of the agricultural land [54]. Compared to ancient times, modern engineering measures have a strong influence on the present water balance, particularly dam constructions in the tributary’s headwater areas. In antiquity, the city of Pergamon was supplied with water by an extensive system of aqueducts from springs all over its hinterland [55]. The extent to which these systems influenced the agricultural use of the plain is currently unknown.

2.2 Social setting

Pergamon was a major superregional Hellenistic and Roman political and cultural center that continued to exist as a small town in the Byzantine and Ottoman periods before becoming a regional center in the Turkish Republic. Its ancient history is evident in the architecture, traditions, and cultural practices that are still evident in modern-day Bergama and its surrounding villages. Although the site shows traces of settlement dating back to the Middle Bronze Age, Pergamon emerged as a politically and economically independent polis by the mid-4th century BCE as part of the reign of Alexander the Great [42,56]. With the death of Alexander and the subsequent division of the empire, his retainer, Lysimachos, entrusted Philetairos with the safekeeping in Pergamon. Philetairos strategically utilized the treasury entrusted to him, which laid the foundation for establishing the Attalid dynasty, also known as the Pergamene Kingdom [42]. Under the Attalid rule, the city continued to grow and prosper, becoming one of the most important centers of the Hellenistic Period in the 2nd century BC [41,56]. Agriculture must have played an important role in the local economy of the city and its spheres of influence, testified by the local practices for olive oil and wine

production, as well as grain processing, including the use of a large-scale network of millstones and presses [30,60]. There is evidence of a large number of settlements on the Bakırçay plain and the surrounding hills [3,61], indicating extensive land use as well as a development of prosperity.

The transformation of Pergamon from a Hellenistic polis to a Roman Imperial metropolis occurred in approximately four major phases [42]. The first phase is characterized by the transformation of Pergamon into a Hellenistic city during the 3rd century BCE, with the city’s center remaining concentrated on the upper part of the city hill. The second phase paralleled the peak of Attalid power during the 2nd century BCE, marked by the integration of the largest part of the city hill into a fortification, accompanied by extensive building activities and infrastructure improvements. During the third phase (1st century BCE to 1st century CE), the city mainly developed within the fortifications of the early 2nd C BCE. At the same time, during the Pax Romana (after 27 BCE), the onset of military security led to a decline in fortified settlements across the Pergamon micro-region [3,61]. The fourth phase of urban development and prosperity of Pergamon captured the period of the late 1st and the 2nd century CE. This phase was characterized by urban reorganization paired with extensive urban expansion towards the Bakırçay plain, along with a large building program, including prestigious buildings such as the Sanctuary of the Egyptian Deities, also known as the Red Hall. Together, these interventions likely contributed to a considerable population increase. Following this last major expansion of the ancient city, a phase of stagnation ensued, as documented by the decline of residential areas on the city’s hill slopes in the 3rd century CE and a decline in building activities. In addition, the so-called “Gothic Wall” was built in the mid-3rd century CE as an extended fortification against the raiding Goths in Asia Minor [42,62].

Today, the environs of the ancient city of Pergamon support a mix of agricultural, industrial, and residential land uses (Fig 1). Agriculture, including both traditional and modern farming practices, plays an important role in the local economy, with crops like olives, wheat, cotton, and tobacco cultivated in the plains, their direct surroundings as well as intramountainous basins [63,64].

3.1 Ecological Crop Requirement (EcoCrop)

EcoCrop, developed by the Food and Agriculture Organization (FAO) of the United Nations [31], is a database designed to understand the ecological niches of specific plants and crops. It includes environmental requirements and limitations for 2,568 taxa, specifying the optimal and absolute ranges of environmental factors (e.g., climatic conditions, soil characteristics) required for successful growth. EcoCrop is based on the law of the minimum by providing parameters for each crop within both absolute and optimal ranges [71]. The database is grounded in the law of the minimum, meaning that the least favorable parameter determines the overall suitability for plant growth on a given plot of land [35]. These tolerance ranges are expressed as fuzzy data, encompassing both absolute and optimal values for each parameter (see S1 File). While this approach does not account for the full variability of ancient crop performance – especially the phenotypic plasticity of premodern landraces – it ensures transparency and reproducibility.

In our study, we considered climatic (monthly mean temperature, monthly precipitation sum), topographical (slope, topographic wetness index (TWI), wind exposure, solar radiation), and soil parameters (texture, pH) as variables (Fig 3). The EcoCrop model generates a crop-specific suitability index by combining these inputs. Input parameters were categorized as either static (unchanging across time and seasons) or dynamic (varying seasonally and across time frames) [32]. Solar radiation, while quasi-static, incorporates monthly sunshine hour differences but does not account for year-to-year variations. Topographic and pedological parameters, given the spatial resolution of ca. 1x1 km, were integrated as static inputs [72].

This study aimed to create a simple yet robust multivariate model that realistically reflects naturally complex processes. To ensure replicability, we used non-commercial datasets. Topographic parameters such as solar radiation, TWI, wind exposure, and slope were derived from the TanDEM-X DEM [40,73]. Soil data (texture, pH) were integrated using the multivariate global regression model SoilGrids [74]. While the original EcoCrop database does not account for topographic factors such as slope, wetness, and wind exposure, we have incorporated these to enhance the spatial differentiation of cultivation suitability. Slope and TWI parameters were adapted from Laabs and Knitter (2021) [29], while wind exposure tolerance ranges were developed based on present-day land use in the Pergamon micro-region. Soil texture classifications (light, moderate, heavy) from the original EcoCrop database were translated into specific grain-size compositions (clay, silt, sand) to align with SoilGrids data.

The EcoCrop model was implemented using the R package “Recocrop” [32,75]. Details of the model setup and the associated code are provided in the supplement (S1 File). The suitability index for each crop was then normalized and classified into five classes, ranging from very low to very high suitability for cultivation.

3.2 Archaeobotanical records

We ran the EcoCrop model for different crop types commonly used as food sources during the Hellenistic to Roman Imperial periods in the eastern Mediterranean. The selection of these crops was informed by an analysis of archaeobotanical records (Table 1; Fig 2A). Specifically, we utilized the work of Marston and Birney (2022) [59], who assessed archaeobotanical datasets from the Greek coast, central Anatolia, and the Levant to identify frequently cultivated crops. Additionally, we incorporated data from the Ademnes database [58,69,70], which harmonizes archaeobotanical records across the eastern Mediterranean and Near East. The selected crops are consistently attested across multiple regional sites, including both inland and coastal settlements, suggesting broad cultural and ecological relevance.

The list includes staple cereals such as common wheat, durum wheat, emmer, and barley, which appear in nearly all archaeobotanical contexts and formed the dietary base (Table 1). Lentil and bitter vetch were included for their frequent occurrence in the archaeobotanical records, their drought tolerance, and their functional role in diversified cropping systems. However, agricultural practices in the region likely included a broader range of crops to enhance the resilience of food supplies. Other crops (e.g., grass pea, chickpea, spelt, millet) were excluded from the final model due to sparse attestation, limited regional relevance, or uncertain cultivation status during the study period. However, the integration of additional crops during model development did not result in further analytical benefit and was therefore not pursued in the final model.

In the case of barley, the EcoCrop database provides generalized climatic thresholds for the species level and does not distinguish between specific subspecies or varieties, such as 2-row or 6-row, hulled or naked barley. While archaeobotanical evidence suggests that 6-row hulled barley was commonly cultivated as food in western Anatolia [76], our model reflects the broader species-level parameters.

The physician Galen mentions wheat, emmer, and einkorn and vegetables such as lettuce consumed in Pergamon [30] (cf. Gal. 6, 518). While not all food sources can be reconstructed from physical remains, archaeobotanical studies across the Mediterranean offer valuable insights into crops that were likely cultivated. Pollen records from the coastal city of Elaia indicate a consistent presence of cereal pollen between 4 ka and 1 ka BP [52], although they cannot distinguish specific crop types. Similarly, the Kara Göl, a mountain lake located in the Kara Dağı southwest of Pergamon (Fig 1), includes records of cereal pollen [77]. However, its low temporal resolution makes it unsuitable for detailed comparison.

While the selected crops (Table 1) are consistently attested across the broader eastern Mediterranean region during the Hellenistic and Roman periods, recent work by Marston and Castellano (2023) [78] highlights that the importance and distribution of individual crops likely shifted over time. Similarly, Douché et al. (2021) [79] provide a synthesis of archaeobotanical records from first-millennium BCE sites in Greece, confirming the continued presence of staple cereals and pulses – such as emmer, barley, and lentil – at both inland and coastal sites. However, in the absence of sufficient high-resolution archaeobotanical evidence for the Pergamon micro-region, this study adopts a temporally averaged perspective on crop use.

Additionally, perennial crops such as olive, grape, and fig – although central to the ancient Mediterranean economy – were not included in the model due to their multi-year life cycle, differing agroecological requirements, and dependence on more intensive or managed systems (e.g., terracing, pruning, irrigation). Since the present study focuses on the climatic potential for rainfed annual field crops, perennial species should be considered in future approaches.

3.3 Climate data

To assess the paleoclimatic history along the Turkish Aegean coast, we utilized a transient paleo simulation [80] conducted with the Max Planck Institute Earth System Model version 1.2 (MPI-ESM1.2) [37]. The simulation results were downscaled using WorldClim 2.1 [81], a modern dataset based on station and satellite observations with a spatial resolution of about 1 × 1 km. Validation was performed using an updated version of the LegacyPollen 2.0 dataset [57], which provides pollen-based climate reconstructions from a globally harmonized dataset.

Five snapshots were selected for downscaling (see section 3.3.3). Focusing on snapshots helped reduce processing time and storage requirements. These snapshots were selected to align with historically important periods, providing a contextualized understanding of modeled climatic changes within broader phases of climate variability (Fig 2). Each snapshot represents a 20-year interval, including monthly average temperature and precipitation data. While the standard climate normal period spans 30 years, we adopted a 20-year interval to reflect better the temporal scales relevant to agricultural adaptation and human perception of climate change. Modern decision-making for annual and perennial crop systems typically occurs over 5–10 years [82]. However, given the generational nature of ancient agricultural traditions [83], a 20-year period was considered an appropriate compromise for this study.

4.1 Climate variation in western Anatolia

Between 400 BCE and 400 CE, the climate in western Anatolia, as represented by four MPI-ESM 1.2 grid cells (Fig 2A), exhibited variability in temperature and precipitation (Figs 2B and 4A). The MPI-ESM simulation and the pollen-based reconstruction both highlighted alternating phases of warmer and drier conditions, as well as colder and wetter conditions. However, the exact timing and intensity of these phases varied between both. These climatic phases vary in length, with major phases lasting up to 60 years and minor phases persisting for 10–30 years.

The climatic states derived from the MPI-ESM simulation used as inputs for the agricultural model were based on the five snapshots (310–290 BCE, 10 BCE – 10 CE, 100–120 CE, 240–260 CE, and 340–360 CE; Fig 2b) representing different climatic states within the study period. The snapshots reflected the variability within the dataset rather than precise historical conditions. For example, wetter and colder conditions were more prominent in the earlier snapshots, while later snapshots tended to represent warmer and drier scenarios. Both models indicated substantial variability even within these generalized phases, with some snapshots capturing transitional states between broader trends (Figs 2 and 4A). These results emphasized the complexity of climatic variability in western Anatolia during the study period.

The first snapshot, corresponding to the period from 310 to 290 BCE, represents a time marked by colder and wetter conditions, as confirmed by the pollen-based reconstruction. Precipitation values were generally above average, while temperature values were below average, indicating a wetter and cooler phase relative to the overall variability of the study period.

The second snapshot, from 10 BCE to 10 CE, reflects a transitional phase. Temperature values in both datasets – the MPI-ESM simulation and the pollen-based reconstruction – were closer to or slightly above average, while precipitation displayed moderate variability. This snapshot captures conditions of relative balance between warmer and drier tendencies, as well as wetter and colder intervals.

The third snapshot, spanning 100–120 CE, represents a particularly warm phase, with temperature values consistently above average in both the MPI-ESM simulation and the pollen-based reconstruction. Precipitation, while variable, remained generally close to average, indicating a predominantly warm and stable drier climate state during this interval.

The fourth snapshot, corresponding to the period from 240 to 260 CE, captures a transitional phase characterized by above-average but slightly decreasing temperatures and precipitation. The snapshot indicates somewhat cooler and wetter conditions compared to the preceding snapshot, reflecting a shift toward a more moderate climate state.

The fifth snapshot, spanning 340–360 CE, represents a cooler setting than observed in the fourth snapshot. While temperature remained above average, precipitation values were close to or slightly below average in the MPI-ESM simulation. However, the pollen-based reconstruction reflected a wider difference between warmer temperatures and below-average precipitation.

These snapshots, incorporating 20-year averages, served as the climatic forcings for the agricultural model, providing a framework for assessing potential cultivation conditions under varying temperature and precipitation regimes.

4.2 Crop-Specific suitability for cultivation

The spatial statistics refer to the Pergamon micro-region, which is defined by the combined boundaries of the walking distance and the watersheds of its two main rivers (Fig 1). The crop-specific models indicate a moderate to high suitability of areas in the alluvial plains of the Bergama graben for agricultural cultivation across all analyzed snapshots.

The areas showing the highest modeled suitability for cultivation are in the coastal region surrounding the Gulf of Çandarlı and the Gulf of Dikili. The coastal regions exhibited only minor temporal variations in cultivation suitability (Fig 5). In contrast, some areas in the western lower plain (between Atarneus and Elaia) and the eastern lower Bakırçay plain (approximately 5 km east of Pergamon) were classified as unsuitable for cultivation across all time slices. High topographic wetness primarily characterizes these areas compared to other parts of the Bakırçay plain; in our model, topographic wetness is a temporally static parameter (see S1 File). In contrast to some of the crop-specific models, the hybrid model does not indicate any cultivation suitability on the Kozak plateau (approximately 20 km north of Pergamon) or in any other highlands (Fig 5).

4.2.6 Variability of cultivation suitability across snapshots.

The coefficient of variation across the five snapshots provides insights into which crops and areas are most susceptible to climatic fluctuations in terms of cultivation suitability. Barley and lentils exhibited the lowest variability in suitability across the analyzed snapshots (Fig 6). In particular, the Kozak plateau and the eastern parts of the Pergamon micro-region showed minimal variability in suitability for lentil cultivation. In contrast, these same areas experienced extensive fluctuations in the suitability for common wheat, emmer, and bitter vetch. Notably, the suitability bitter vetch varied on the Kozak plateau (Fig 5), while common wheat and emmer displayed pronounced variability both on the plateau and in the hills between Pergamon and Allianoi. The durum wheat suitability, by contrast, showed fluctuations in the piedmont zone of the Bergama graben and the eastern lower Bakırçay plain. However, the lower Bakırçay plain as a whole exhibited relatively low variability. These spatial and crop-specific differences reflect the varying climatic sensitivity of individual taxa, as determined by their environmental tolerance parameters (see Supplement S1 File).

The hybrid cultivation suitability model mirrored the pattern observed for durum wheat, with changes in suitability concentrated along the piedmont zone of the Bergama graben. Regarding the climate variables, the coefficient of change for both the annual sum of precipitation and the mean annual temperature (Figs 8A, B) indicates that coastal areas – particularly in the vicinity of Elaia and Atarneus – exhibited the greatest variability in precipitation (up to 79%) across the five snapshots, compared to the more stable conditions in continental hinterland. In contrast, the eastern and northern mountainous areas of the Pergamon micro-region showed the highest variation in temperature, reaching up to 65%.

Download:

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

Fig 7. Change in cultivation suitability across the seven common crops and the integrated hybrid model.

Change is expressed as a coefficient of variation across all five snapshots (Fig 5). The change in the binary hybrid model is described as the sum of changes across all snapshots.

https://doi.org/10.1371/journal.pone.0325779.g007

Download:

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

Fig 8. Spatial distribution of downscaled average climate variables across the five snapshots (310-290 BCE, 10 BCE-10 CE, 100-120 CE, 240-260 CE, 340-360 CE) a) average annual sum of precipitation b) average annual temperature c) coefficient of variation of the average annual sum of precipitation across the five time periods d) coefficient of variation of the average annual temperature across the five time periods.

https://doi.org/10.1371/journal.pone.0325779.g008

5.1 Spatio-temporal dynamics and their implications

The hybrid cultivation suitability model revealed that the areas with the highest suitability for cultivation between 400 BCE and 400 CE were concentrated in the coastal plains around Atarneus and Elaia, as well as large parts of the western lower Bakırçay plain (Fig 5). In contrast, the eastern lower plain exhibited more variable suitability, being cultivable only during certain snapshots. This variability reflects climatic and environmental gradients, as the western plain was characterized by higher average temperatures and lower annual precipitation. In contrast, despite similarly warm conditions, the eastern plain experienced greater humidity (Fig 8).

Our findings diverged from previous studies of cultivation suitability in the Pergamon micro-region, such as those by Laabs and Knitter (2021) [29], which suggested high arability across the entire lower Bakırçay plain. A key reason for this discrepancy is their reliance on static topographic parameters – such as slope, aspect, and topographic wetness – as primary indicators of agricultural suitability. Our results, incorporating both climatic and topographic data, demonstrated a west-to-east gradient in temperature and precipitation, with the eastern plain less suitable for the continuous cultivation of multiple crops. Today’s land use patterns and suitability classifications, as noted by Laabs and Knitter (2021) [29], equally characterize the western and the eastern lower Bakırçay plain as agriculturally suitable areas. However, this does not necessarily reflect historical conditions, as the Bakırçay alluvial plain and the coastal areas of the Gulf of Çandarlı and Gulf of Dikili have been extensively modified by irrigation and drainage systems. Historical travel reports, such as those by Cockerell (1810–1817) [96], describe extensive waterlogging in the plains during the rainy winter months, necessitating the construction of elevated causeways. Today, a network of drainage ditches mitigates this risk of waterlogging. In our model, areas identified as unsuitable for cultivation corresponded to zones of high topographic wetness, suggesting they were historically prone to waterlogging (see S1 File).

Despite these environmental challenges, ancient accounts by scholars such as Strabo and Galen described the Kaikos (Bakırçay) floodplain as a fertile landscape [30,56,97]. Seasonal flooding brought fine, mineral-rich sediments into the plain, creating vertical accumulations that supported agricultural activities [30]. Early 20th-century mapping by Philippson (scale 1:300.000 [97]) indicated that the Bakırçay meandered with low sinuosity, traversing the plain centrally. These hydrological dynamics likely required adaptation to periodic flooding, reinforcing the view that the western Bakırçay plain was more fertile and intensively cultivated compared to the eastern plain, where livestock farming may have been more prominent [30].

The high cultivation suitability of the western lower plain aligns with settlement findings from archaeological surveys across the western Bakırçay plain and its adjacent foothills [3]. Numerous Hellenistic and Roman settlements, identified as farmsteads or hamlets, were located near major road networks that crossed the alluvial plain [3,98,99]. Spatial analysis revealed shifts in settlement patterns between the Hellenistic and Roman periods [61]. While Hellenistic settlements were often located on slopes, ridges, and hilltops, Roman settlements increasingly occupied piedmont areas and, to a lesser extent, the alluvial plain itself [3,61]. This shift suggests a closer connection between Roman settlements and agricultural activities on the plain. In contrast, the eastern lower Bakırçay plain has not been extensively analyzed archaeologically yet, precluding a direct comparison between ancient settlement density and modeled suitability for cultivation. However, farmsteads and hamlets similar to those on the western plain have been documented in the north-eastern hinterland of Pergamon, in the hilly intramontane areas between Pergamon and Allianoi [3]. These elevated areas adjoin the eastern plain and were identified as increasingly suitable for cultivation in the hybrid model. The eastern expansion of agricultural suitability reached its greatest extent in the fourth snapshot (240–260 CE), encompassing much of the eastern plain. The results suggest that while the lower eastern Bakırçay plain may have been agriculturally favorable, it likely supported a lower diversity of crops compared to the western plain. By contrast, the foothills between Pergamon and Allianoi exhibited well-balanced cultivation suitability (Fig 5), making them potentially important for diversified agricultural practices.

5.2 Agroclimatic potential and settlement activities

The hybrid cultivation suitability model highlights distinct spatial patterns of agricultural potential in the Pergamon micro-region across five modeled climatic states. These states, represented by snapshots derived from the MPI-ESM 1.2 simulation, reflect potential conditions rather than precise reconstructions of historical realities. However, they provide a useful framework for exploring how varying climatic conditions might have influenced agricultural practices and socio-economic developments in the region.

Across all snapshots, the western Bakırçay plain, including the alluvial fan of the Selinos River, consistently exhibited high agricultural suitability. These areas likely offered stable conditions for cultivation under most scenarios, aligning with the role of fertile plains as the foundation of agricultural systems. In contrast, the eastern Bakırçay plain and the Kozak plateau exhibited greater variability, suggesting that their agricultural potential was more sensitive to climatic shifts and required adaptive strategies to optimize land use.

These dynamics are also reflected in the modeled suitability of individual crops. While cereals such as barley and common wheat consistently perform well in the western plain, bitter vetch – a crop known for its resilience and long cultivation history [70,76] - appears only marginally suitable in some snapshots. The EcoCrop model emphasizes optimal yield conditions based on modern climatic thresholds [35], which may underestimate the crop’s tolerance to marginal environments historically exploited through low-input or flexible cultivation strategies. Additionally, the crop parameters used in the EcoCrop model are based on modern cultivars, which may differ from ancient crops in their tolerance to environmental stress, growth periods, and yield stability. As Heinrich and Hansen (2021) [15] have emphasized, locally adapted ancient crop varieties likely exhibited a broader phenotypic plasticity than modern commercial strains. While the use of modern performance data remains the best available option in the absence of ancient physiological datasets, it introduces a degree of uncertainty that must be considered when interpreting suitability scores. These outcomes highlight the heuristic role of the model in identifying interpretive tensions between ecological niche modeling and archaeobotanical evidence. Our goal was not to reconstruct precise historical cultivation practices, but to assess how even moderate climatic fluctuations within species-level tolerance ranges could have impacted agroclimatic suitability and, by extension, socio-ecological resilience.

5.3 Adaption and Risk Management

The higher modeled agricultural suitability of the western plain compared to the more humid eastern plain raises the question of whether increased precipitation, typically considered beneficial for agriculture within a Mediterranean Csa climate, enhances agricultural potential.

In climates classified as Csa, where dry summers and wet winters typify the agricultural cycle, the interaction between precipitation and temperature plays a critical role in determining agricultural suitability [68]. While precipitation is generally essential for crop growth, its effectiveness depends on timing and intensity. In this context, the modeled results suggest that moderate increases in average annual precipitation, when combined with higher average annual temperatures, may create optimal conditions for agriculture in the environs of Pergamon. These conditions likely reduce water stress during the growing season while simultaneously extending the range of suitable areas for cultivation.

During the wet winter months (typically October to May), the cooler temperatures in a Mediterranean Csa climate minimize water loss through evaporation, making precipitation during this period particularly effective for replenishing soil moisture. The high frequent winter rains [116] often promote, rather than constrain, agricultural productivity by providing much-needed moisture during the typical growing period for annual crops [68]. However, as observed in the more humid eastern plain, excessive humidity or precipitation may reduce suitability by increasing the risk of waterlogging and limiting the cultivable area, particularly for crops sensitive to overly wet conditions.

The results of the agricultural model highlight the importance of balanced precipitation and temperature conditions. A suitable combination of these factors, as seen in the western plain, appears to enhance agricultural potential by reducing the dependency on extreme climatic conditions. However, the Mediterranean climate’s reliance on the rainy season remains a challenge for long-term agricultural planning, as seasonal shifts or interruptions in precipitation patterns can have severe impacts on crop yields and sustainability [117].

It is nevertheless important to consider that higher annual temperatures, particularly in the western plain, may have a more pronounced impact on the cultivation suitability at the regional level. On the one hand, warmer conditions can extend growing seasons in higher elevations, enabling the cultivation of a wider variety of crops and even increasing yields for certain heat-tolerant species. On the other hand, higher temperatures can also lead to faster phenological development and, therefore, shortening of growing periods [118]. The benefits of longer growing periods must also be weighed against the risks of prolonged dry spells during the summer months, which could stress crops if water resources are not carefully managed and utilized.

Furthermore, winter precipitation in the Aegean region often occurs as extreme events [119], leading to surface runoff, soil erosion, and damage to agricultural fields. Only a small fraction infiltrates in the soil and becomes available for plants [120]. Land degradation poses a challenge to land management and long-term soil fertility [121]. While these extreme events can have a severe impact on the landscape and agricultural practices, their effects are difficult to capture in climate analyses that rely on monthly averages, which tend to smooth out these extremes.

Accordingly, while temperature and precipitation both substantially influence the region’s suitability for agriculture, in this particular Mediterranean context, higher annual temperatures appear to have a more positive impact on the region’s agricultural systems. However, the unpredictability and erosive impact of extreme precipitation events remains a critical challenge, particularly in areas prone to soil degradation and runoff [121].

While the favorable cultivation conditions – and, therefore, a presumable higher species equitability – offered advantages for agriculture in the western lower Bakırçay plain, the exploitation of these favorable conditions likely increased pressure on the landscape, particularly on the adjacent slopes and the plain soils [12,51,66]. Intensive agricultural practices, typical of antiquity [76], may have contributed to soil degradation and erosion, creating negative feedback loops that increased the landscape’s sensitivity to environmental changes [10,11,66]. This degradation would have been particularly pronounced in areas associated with monocropping and limited crop rotation – practices attributed to some Roman agriculture systems, especially in zones of intensified production [51,76]. However, these practices cannot be generalized across the Roman Empire. The diversity of land tenure systems, production scales, and ecological settings across the Roman Empire suggests a spectrum of cultivation strategies – including mixed cropping and fallow-based systems – that complicates simplistic assumptions [15,104,122]. In the context of urban centers like Pergamon, agricultural organization may have been shaped as much by demand and logistical networks as by environmental conditions.

While such socio-economic diversity complicates attempts to generalize cultivation strategies, it also underscores the value of analytically separating ecological potential from actual land use. Therefore, the presented models do not attempt to reconstruct agricultural systems or specific cropping decisions but instead focus on the environmental suitability for cultivation under different climatic scenarios. This distinction is particularly relevant when contrasting the relatively stable conditions of the lower plain with the more heterogeneous landscapes of the surrounding mountain ranges.

Mountainous areas, located at a distance from the coastal zones in the eastern part of the Pergamon micro-region, present a contrasting scenario to the high suitability in the western, lower Bakırçay plain. These regions, characterized by frequent changes in cultivation suitability (Fig 7), would have required more adaptive strategies to cultivate crops and maintain food security. Crops like wheat, barley, and lentils appear more resilient to these fluctuations, allowing for continued cultivation despite climatic shifts (Figs 5 and 6). In contrast, other crops, such as emmer, durum wheat, and bitter vetch, showed high variability in their suitability over time (Fig 7), particularly in the foothills and intermountain basins surrounding the plain (Fig 5). The variability of these crops, especially durum wheat, was most pronounced in the foothills northeast of Elaia and Pergamon, where agroclimatic conditions fluctuated markedly. Therefore, for the mountainous parts of the Pergamon micro-region, adaptive strategies such as intercropping may have played a crucial role in creating a more resilient system and reducing the risk of crop failures [15,68]. The extent to which trade influenced food security and the prosperity of the Pergamon micro-region is unknown. This shifts the focus from local water availability to broader networked dimensions, including concepts like virtual water [123]. Here, not only would actual agricultural products be considered for trade across the nodes of superregional networks, but also regional input parameters, such as soil nutrients, available water, and labor, that facilitate the growth and enhance the productivity of a specific crop.

Historical examples of adaptive processes related to climatic fluctuations are found in the south and west of Smyrna (approximately 100 km south of Pergamon). Here, the archaeobotanical study of Maltas et al. (2023) indicates a decline in the cultivation of glume wheat (einkorn and emmer) and a notable increase in barley between the Early Bronze Age (ca. 3100/3000–2600 BCE) and the Middle Bronze Age (~2140–1890 BCE) [2]. The authors suggest that this shift reflects an adaptive strategy in response to climatic variability, as barley is a drought-resistant cereal alternative to glume wheat. Compared to our modeling results – which cover a time frame approximately 1500 years later – barley is identified as the most suitable and stable crop among all the modeled taxa (Figs 5 and 6). Although our models indicate that emmer had limited suitability for cultivation in this region, the comparison between the various climatic forcings demonstrates only a marginal influence on the natural potential of emmer, suggesting a rather resilient natural suitability. It should be noted, however, that our models do not account for soil enrichment and other forms of melioration, which may contribute to the declining resistance of such marginal crops in actual practice. In addition, the records around Smyrna indicate that legumes, such as bitter vetch, were cultivated in greater quantities, while the amount of lentils decreased over the periods [2]. Although areas occupy a large portion of moderately to highly suitable areas in our study (Fig 5) barley shows fewer fluctuations in suitability across different climatic forcing scenarios per snapshot. Lentil and bitter vetch demonstrate a notable shift in suitability, particularly in the second snapshot (10 BCE – 10 CE).