Long-Term Settlement Dynamics in Ancient Macedonia: A New Multi-Disciplinary Survey from Grevena (NW Greece)

Abstract

This paper discusses the evolution of human settlement in ancient Macedonia from the Neolithic to the Late Roman periods, based on the results of a new multi-disciplinary and multi-scale archaeological survey in northern Grevena (NW Greece). Building upon an unpublished (legacy) survey, we developed a GIS-structured workflow that integrates site-revisiting and surveying strategies (material collection and test pits) with multi-temporal remote-sensing analyses, offering analytical information about site distribution, characterisation, dating, and taphonomy. Notably, the new study led to a 64% increase in the number of known sites. The combined results indicate that prehistory is less represented in the surface record than historical periods, likely due to the impact of soil erosion episodes. The Late Bronze Age and Early Iron Age saw increased site numbers and the emergence of a settlement structure that characterised the area until the Hellenistic period. During the Roman period, the pattern shifted from a seemingly limited use of the landscape towards a model of more extensive habitation. This was driven by the appearance of new rural sites that introduced a land-use regime designed to support agricultural intensification by implementing anti-erosion measures, such as field terraces.

1. Introduction

The quantification and statistical appraisal of site data obtained through regional archaeological surveys (e.g., chronological phases of use, artefact distribution densities, spatial extent, and visibility parameters) have been a real breakthrough in the Mediterranean since the rise of systematic field-sampling techniques in the 1970s [1,2,3,4]. However, the volume and complexity of the collected datasets, combined with varying management styles or personal factors, has often complicated the publication process. A recent ‘survey of surveys’ suggested that only one in four research projects has achieved both analysis and extensive publication of results [4]. Consequently, a huge body of unpublished information, termed as legacy data [5,6,7,8], remains underexplored and inaccessible to the research community. This situation poses challenges not only to established research strategies, but also to the recognised need for paving a communication route among different survey projects to verify, discuss, and transfer datasets under a fair framework of common good practice within archaeological surveying [3].

In earlier works [9,10], we demonstrated the potential of extracting analytical information from bibliographically ‘missing’ data, taking the Grevena Project (1986–1994) as an example. The Grevena Project was one of the very few large-scale archaeological surveys conducted in the uplands of northern Greece [11,12,13,14,15,16,17,18,19,20]. This ambitious regional programme applied an all-period extensive, site-based strategy for the reconnaissance and study of the regional archaeological record. In addition to its material collections, it incorporated multi-disciplinary approaches, such as geomorphology [21], pollen analysis [22], and ethnoarchaeology [23]. Despite the publication of numerous articles and reports, the only available information is the names of the sites, their approximate coordinates, and bags of collected finds, with no further analysis or documentation. Essential methodological considerations are absent, such as details on the fieldwalking and artefact collection methods, as well as site characterisation, dating, and size estimates. For example, a statistical overview of all the collected finds revealed inconsistencies in the sampling methods—at times a type of ‘total-collection’ approach was adopted, whereas, in most cases, a similar percentage of all the ‘diagnostic’ categories was kept [10]. A justification for such variations was not provided, which, in combination with an a priori focus on certain cultural periods, i.e., the Early Neolithic, Late Bronze Age, and Roman [18,19,20], hinders the possibility of producing a comprehensive synthesis of human activity in the region. This also limits potential comparisons with other geographical zones, especially southern Greece, where quantitative survey data are more abundant [4].

To address these issues, we developed a new multi-disciplinary and multi-scale landscape approach using survey data to construct meaningful long-term narratives of settlement and land-use dynamics for this part of Macedonia, spanning from the Neolithic to the Late Roman periods (see

Table 1. Conventional chronologies of the main cultural/historical periods of western Macedonia.

Major Cultural Period	Chronology (BCE/CE)
Early Neolithic (EN)	6600/6500–5800 BCE
Middle Neolithic (MN)	5800–5300 BCE
Late Neolithic (LN)	5300–4600 BCE
Final Neolithic (FN)	4600–3300/3100 BCE
Early Bronze Age (EBA)	3300/3100–1900 BCE
Middle Bronze Age (MBA)	1900–1700/1500 BCE
Late Bronze Age (LBA)	1700/1500–1050 BCE
Early Iron Age (EIA)	1050–700 BCE
Archaic	700–480 BCE
Classical	480–323 BCE
Hellenistic	323–148 BCE
Early Roman (ER)	148 BCE–200 CE
Middle (MR) and	200–641 CE
Late Roman (LR)
Byzantine/Medieval	641–1389 CE
Ottoman	1389–1912 CE

for the chronological framework). By deploying a GIS-based workflow focused on multi-temporal remote-sensing analyses completed using site-revisiting strategies, we illustrate how our methodology is not only capable of reproducing but also expanding unpublished survey datasets with new discoveries. In particular, we provide statistical observations of site locations, their preservation, and taphonomy, while establishing dating and visibility biases based on the collected materials. An overall methodological and period-by-period discussion of the integrated survey results with other regional and also supraregional case studies follows. Finally, we assess the strengths and limitations of the proposed workflow alongside what we consider to be the most promising research directions for future archaeological studies in the region.

2. Study Area

This new survey was conducted in the mid-mountainous, northern part of the modern Grevena prefecture (region of Western Macedonia, Greece), extending over an area of 246 km² between the village of Ayios Georgios and the town of Grevena (Figure 1). This area was chosen based on environmental factors, land-use patterns, and previous fieldwork, i.e., the high density of archaeological sites reported by the Grevena Project [24] and earlier publications [25,26,27,28]. Geologically, this zone belongs to the so-called Mesohellenic Trough. The geomorphology of the area features an elongated ensemble of plateaus, heavily incised by abundant streams and small river catchments. These are characterised by irregular topographies of relatively large ridges, connected by steep slopes to the valley bottoms, and exhibit an elevation between c. 500 and 850 m asl [29,30,31].

Today, this environmental setting corresponds to a predominantly agricultural area focused on the cultivation of wheat, barley, and lavender, with its ploughsoils favouring the application of pedestrian survey techniques (

Table 2. The CORINE Land Cover (CLC) inventory (https://land.copernicus.eu/ [accessed on 22 August 2024]) for 1990 and 2018 compared to the location of the archaeological sites in the study area.

Land Cover Class	Ground Coverage (%)	Nº of Sites (1990)	Nº of Sites (2018)	Difference (%)
122—continuous urban fabric	0.87%	0	1	+100%
211—non-irrigated arable land	48.8%	52	49	−5.8%
242—complex cultivation pattern	2.08%	3	3	=
243—land principally occupied by agriculture with significant areas of natural vegetation	19.46%	4	5	+20%
311—broad-leaved forest	10.97%	3	3	=
323—sclerophyllous vegetation	2.13%	2	2	=
324—transitional woodland–shrubs	7.94%	0	1	+100%

) [3]. In large areas, mechanised cultivation has resulted in the exposure of the local bedrock, which largely comprises soft sandy marls, sandstones, and polygenic conglomerates (e.g., some lavender fields are planted directly in the ploughed marl) [32]. Thin colluvial soils constitute the dominant surficial formation, including local detrital materials, partly anthropogenic and partly from the parent bedrock, which have been transported to most parts of the slopes and the valley bottoms [24,33]. Such soils are considered very productive in the hilly landscapes of Greece, due to their ability to maintain moisture and supply nutrients to the vegetation [34]. However, they are also prone to runoff soil erosion from the upper relief to the footslopes, which is exacerbated by intense ploughing. Geomorphological evidence from the Xerolakkos stream, on the northern border of the surveyed area, indicates that this process is not exclusively modern, and started at least during antiquity [33].

Previous visits to the study area during the course of the 20th century mentioned a rich archaeological record of sites attributed mainly to the Hellenistic and Roman periods (Table 1), including stone architectural remains, inscriptions, potsherds, and small finds [16,25,26,27,28,35,36]. However, similarly to the Grevena Project, the site information concerning their exact location and spatial extent were never published. The absence of systematic excavations and, therefore, of stratified artefact assemblages (e.g., pottery) in the area also creates a significant limitation to establishing high-resolution chronological sequences of the local material culture, for which relative dating depends inevitably upon typological similarities with materials from neighbouring regions (e.g., Epirus and Kozani).

In classical sources, the region was part of what was known as Upper Macedonia, a term applied to denote the upland, mountainous environment, in contrast with Lower Macedonia, which referred to the Aegean coastline and coastal plains to the east, such as the area of Thessaloniki [37]. In studies related to the political geography of the area during the Classical period (c. 480–323 BCE), six territorial subdivisions have normally been recognised in Upper Macedonia, and Grevena has been suggested to have included parts of Elimiotis to the northeast (Mt. Vourinos) and Tymphaea to both the west (Pindus range) and south (Mt. Antihasia) [38,39,40]. The Aliakmon River has usually been proposed as the frontier between the two regions, taken per se as a natural topographic boundary. Such modern interpretations have largely been applied based not on archaeological findings but on conventional geographical observations, and so they must be considered inconclusive [41]. As a result, Ayios Georgios, located in the middle of the region, has been viewed at times as part of Elimiotis [42] and at other times as part of Tymphaea [39,43]. The presence of Hellenistic (3rd–2nd centuries BCE) and Roman (2nd century CE) dedicatory inscriptions at the hill site of Ayios Nikolaos were central for identifying the area as one of the regional capitals of Upper Macedonia [28,44], but assumptions about the toponym and its relation to the Macedonian hierarchy have been grounded so far on occasional evidence.

3. Materials and Methods

The methodology applied in this case study followed a multi-disciplinary landscape analysis supported by geospatial technologies (e.g., GIS). This collaborative approach aimed to generate data across multiple scales of resolution, from intra-site chronostratigraphies to artefact distribution maps and long-term settlement pattern hypotheses. The primary fieldwork method was structured around site surveying [3,45,46,47], complemented by multi-temporal remote-sensing and large-scale geomorphological investigations. While each technique has been successfully used elsewhere in regional projects [3], their integrated application here is innovative because it enables the comparable management and geospatial processing of multi-source data under a unified surveying strategy, from the recording stage to the final interpretation of the evidence.

A GIS geodatabase using QGIS Desktop 3.34.6^® was designed as the toolkit for conducting this new survey. A standardised geodatabase was created for all the entry categories (e.g., pottery recording, artefact sampling during fieldwalking, test pit descriptions, etc.) to enable consistent entity–relationship comparisons between the legacy and these new data. This database was connected to a mobile device application, QFieldCloud, which was customised as a multi-user surveying form based on the standard GPS capabilities of smartphones (with an approx. 2 m mean error) and using real-time cloud syncing to secure the data acquisition [48,49].

In the first stage, the database was used to map all the previously known sites in the study area, incorporating quantitative and qualitative analyses of the collected materials (e.g., classification and dating), photographs, personal notes, and bibliographical references. Most of the sites were drawn from the Grevena Project index and supplemented with additional locations published by other researchers [25,26,28], as well as reports from the Grevena Ephorate of Antiquities. This dataset served as the guide for relocating the archaeological sites of the region. Moreover, a compilation of multi-source and multi-temporal remote-sensing materials—including cartography, archival aerial photographs, and satellite imageries—were georeferenced and integrated as basemaps into the existing geodatabase to detect potential new sites, visualise the local topography, and also to trace recent landscape changes and their impact on the surface record of Grevena [10].

An extensive, site-targeted survey was carried out to investigate all the identified sites within this case study. Our workflow aimed to (a) locate the sites through pedestrian fieldwalking; (b) describe their environmental setting (topography and land-use evolution) and the relationship between the local geomorphological dynamics and site taphonomy (e.g., soil erosion vs. stability and pedogenesis); (c) define their size/extent by documenting the spatial distribution of the surface material culture; (d) assess their chronological span of use by collecting a sample of the visible finds and comparing it with the legacy collection; (e) determine the nature of human activity based on the artefact typology and surrounding topography (e.g., settlement, cemetery, or workshop); and, beyond the site boundaries, (f) identify potential off-site activity, that is, the cultural remains recovered outside the strict boundaries, but likely associated with the presence of the site [50,51,52,53,54], and use it as evidence for reconstructing past land-use practices. Thus, this survey was extended to a radius of 200–500 metres around the site limits, depending on ground visibility and accessibility conditions [48,55,56,57,58,59].

During three survey seasons (2021–2023), a team of two to four surveyors walked transects in parallel, divergent, or cross-sectional routes (Figure 2). When a material concentration was noted, GPS measurements were recorded every 10 to 30 metres, with each point assigned a descriptive entry for the visible materials (e.g., description of pottery densities, categories, shapes, and chronologies). The collected materials were stored in a bag tagged with a unique field code that corresponded to the GPS coordinates and included artefacts found within a defined radius around the GPS point, which was also mentioned in the GIS entry to provide spatial information (e.g., ‘materials come from a 20 m radius around the GPS point’) (Figure 2).

The artefact collection was targeted to acquire a representative sample of rooftiles and, in general, building materials (e.g., mudbricks and plasters), diagnostic sherds (rims, handles, bases, decorated sherds, and sherds with different fabrics or ‘chronotypes’), and all small finds (e.g., metal objects, stone tools, and clay figurines) [3,60,61]. Relative, artefact-based chronologies were divided into the main cultural periods and their sub-periods (Table 1) through a typological study of comparable materials from the region of Upper Macedonia and, if necessary, from the wider zone of northern Greece. Single burials, kilns, and architectural remains, whether found separately or in discarded heaps, were given separate GPS measurements, followed by a basic description and photographing. The ground conditions, including the current land-use status (cultivated, uncultivated, ploughed, or planted), cultivation types, and ground visibility (estimated subjectively at 0–25%, 25–50%, 50–75%, or 75–100%) were also recorded [61].

The GPS entries were used to determine the extent of the sites in terms of their surface concentrations and densities [61,62,63]. The results were visualised with vector polygons alongside a synthetic description of all the gathered evidence. In addition to fieldwalking, we analysed the local micro-relief and its potential impact on the site’s preservation and visibility by performing automatised drone (UAV) flights with Mavic 2 Pro and AUTEL EVO II models. High-resolution RGB orthomaps and Digital Surface Models (DSMs) were then generated using Agisoft Metashape Professional 1.8.5^® photogrammetric software.

A key aspect of our fieldwork was the application of geoarchaeological principles to complement the archaeological surface-driven observations. This began with the geomorphological characterisation of the site environs (e.g., ‘exposed bedrock’, ‘colluvium-driven deposits’, or ‘alluvial surface’). To better understand these processes, geoarchaeological test pits were distributed along the transects for two selected sites (Arsalia and Ayia Varvara, both in the vicinity of Ayios Georgios village). These transects were designed to characterise the local soil catena or toposequence by examining in detail the exposed stratigraphic profiles down to the bedrock [64]. Each profile was described in pedo-sedimentary terms, while charcoal samples were collected in situ and sent for AMS ¹⁴C dating. The results were calibrated and plotted using Oxcal 4.4.2 software [65] with the INTCAL20 curve considering two sigma (2σ) ranges [66]. Finally, the test pits were photographed and their coordinates were recorded using GPS.

4. Results

4.1. Fieldwalking Results

In total, the pedestrian survey covered an area of 11.78 km² and identified 64 sites, of which 39 were previously known while 25 were newly discovered. Consequently, the site index of the area increased by 64%. In four cases (9%), the reported locations could not be identified as an archaeological site. Unless stated otherwise, the following analysis concerns the sum of the sites. More than half of the fieldwalking took place in terrain with optimal visibility (75–100%) (Figure 3). The large-scale classification of land-use practices between 1990 (i.e., during the period of the Grevena Project) and the present (2018), based on the Copernicus Land Cover data, showed that 92% vs. 89% of the sites, respectively, were in agricultural zones, with the rest in forested (8% vs. 9%) or urban (0% vs. 2%) areas. Only 6 out of the 64 sites’ land-use was different at present compared to 1990, amounting to a 9.4% change (Table 2).

On only a few occasions, for example at Ayios Petros (Myrsina), was there a high, continuous concentration of surface material. Most commonly, the site locations exhibited a low or partly discontinuous spatial density of material culture. Furthermore, the evidence from Syndendro and probably that from Ayios Georgios point to the existence of contemporaneous settlements located no more than 300 m apart during certain periods, such as the EIA and Hellenistic (Figure 4 and Figure 5). Beyond the site limits, the survey of the environs indicated a landscape very poor in artefacts. Isolated finds, such as stone tools, lithics, ceramic tiles, and pottery sherds, were present, but their characteristics do not suggest a direct connection to the use of the nearby sites. These finds were subsequently recorded as ‘off-site scatters’ but, due to their occasional resolution, they were excluded from the present analysis (Figure 3).

More than half of the surveyed sites were multi-period (58%), containing materials from more than one cultural period (see also Section 4.2, below). Most of them revealed discontinuous phases of activity, for example, with artefacts belonging to the Bronze Age and then to the LR, without diagnostic samples from the in-between periods. Three exceptions must be noted as cases involving a continuous datable sequence across periods: Ayios Nikolaos (or Kastro) in Ayios Georgios (EBA and LBA to modern), Ayios Petros (or Ampelia) in Myrsina (LBA to modern), and Ayia Triada in Megalo Seirini (EBA and LBA to LR) [16,20]. On the other hand, single-period sites (42%) were mostly dated to the EIA and LR periods (Figure 6). The revisited sites alone provided a very heterogeneous view in terms of the artefact visibility and dating. Drawing from the new finds collections, in 42% of the cases we found fewer cultural periods than the legacy data suggested, in 37% the chronologies were the same and, finally, new phases of use were identified for 21% of the sites.

4.2. Site Dating, Categorisation, and Spatial Distribution

Overall, it was possible to categorise the sites into settlements (meaning broad habitation), cemeteries, a potential ritual context (temple), and unknown activity (when the material culture was too fragmented to be assigned a specific use). The data indicated three peaks of human activity, during the EIA, the Hellenistic, and the MR-LR periods (Figure 6). The Neolithic and most of the Bronze Age (excluding the LBA) were represented by very few sites compared to the later periods. The LBA witnessed a notable increase, accompanied by a high degree of continuity of the same sites (75%) into the following EIA (Figure 4). The number fluctuated between the Archaic and LR periods (Figure 5). The normalised values of the settlements (based on the duration of each period) provide a more robust view of the population dynamics, indicating expansion during the EIA and Hellenistic periods, whereas the MR-LR peak appears less significant than the overall site numbers initially suggested.

Diachronically, the majority of the habitation sites (55%) occupied the tops of ridges (Figure 4, Figure 5, and Figure 7). This was a common feature of all the recorded Neolithic settlements. In general, ridges offered advantages, such as high visibility of the surroundings while maintaining easy access to water sources, and were found not only in the extensive valley bottoms but also at swallow depths underground, due to the hydrological properties of the marls and sandstones. In some cases (16%), naturally protected or fortified locations on hill plateaus and hilltops were selected. Notable examples include Paleoglas in Dasaki, Paleokklisi in Lochmi, Ayios Petros in Myrsina (Figure 7(3)), and Ayios Nikolaos in Ayios Georgios (Figure 7(1)). The hill of Ayios Nikolaos is an exceptional case of a multi-period site, where the in situ preservation of successive anthropogenic strata gives the impression of a tell, though this is also due to the steep local topography. No other examples of this phenomenon are known in the area. Two of these protected locations, Paleoglas and Ayios Nikolaos, are the only sites that featured fortification walls during their use, but their dating remains elusive in the absence of modern excavation. At Paleoglas, the wall was visible a century ago [67], but today it only appears as a topographic irregularity along the hilly plateau.

The Archaic and Classical periods are represented by few sites, none of which was from a single period. Instead, the sites from these periods were typically recovered in areas with pre-existing activity. The Hellenistic period witnessed a notable increase in site numbers. By far the largest settlements during this period were the hill sites of Ayios Petros and Ayios Nikolaos, extending over an area of c. 9.1 and 7.1 ha, respectively. The common recovery of rooftiles, table wares, storage vessels, and loomweights point primarily to the domestic use of these spaces [68,69]. On the other hand, the discovery of stone inscriptions in the same area of Ayios Nikolaos, the oldest of which is dated to the Hellenistic period, implies the existence of a public or religious context [36].

A separate category includes rural sites dated to between the 2nd and 4th centuries CE, during the MR-LR period, and are mostly located near Ayios Georgios (10 of 12 such cases) (Figure 5 and Figure 8). These are distinguished not only by their topographic characteristics, but also by their similar size (0.5 to 1 ha) and common material records. They were situated very close to the headwaters of small ephemeral streams, where water springs or local phreatic levels surface. The artefact collections show the dominance of utilitarian pottery with high percentages of ribbed amphora sherds, followed by plain (undecorated) tablewares and large storage vessels (pithoi). There is a strong possibility that these locations were nearby craft or production facilities, based on the recovery of kiln remains (80% of all the discovered kiln parts were found at such sites) and scatters of metal slags. At this stage, it is not possible to differentiate between pottery and metal kilns, but it is likely, at least, that pottery/tile production and some form of metallurgical craftsmanship occurred locally, as it happened in other regions of Macedonia [70]. Their debris also includes building elements (e.g., rooftiles, bricks, and plasters), and one-third of them are monumental architectural remains (e.g., columns, fringe, and thresholds) that can be attributed to large, complex structures. For example, at Pervanas (Klimataki) we identified a stone base of a column, characteristic of a porticus, as well as fragments of terracotta hypocausts that may point to the presence of baths [70]. Most of these sites (70%) were single period and found in locations that had not been occupied before. Parallel examples from Macedonia and southern Greece support their identification as farmsteads and rural villas (‘villae rusticae’), referring to small nuclei of rural settlements related to the agricultural exploitation of the landscape [2,70,71,72,73].

Finally, in Ayia Marina (Amydgalies) the bibliographic testimony of a monumental Roman structure was positively validated by the presence of two worked marble blocks and building materials (rooftiles and plasters) in the area of an excavation conducted in the early 20th century CE [25]. However, no additional materials were recovered to confirm its function as a temple, as had been suggested [25].

The spatial extent of 37 of the 48 visited settlement sites (77%) was mapped based on the artefact distributions. The sample was divided into single-period/phase sites and multi-period/diachronic use, since the recording resolution could not offer period-per-period estimates for the latter category (Figure 9). The sites excluded from these estimates did not have a sufficient material dispersion or any other features with which to define their approximate size. The results illustrate that, diachronically, habitation extended in an area of less than 1 ha (70% of the cases), while larger sites (30%) generally ranged between 1 and 4 ha, with only two settlements exceeding 7 ha. In many periods (Neolithic broad, EBA, Archaic, Classical, Hellenistic, and ER) the scarcity of data prevented generalisations. However, a notable clustering by site size was observed for the MR-LR periods.

4.3. Geoarchaeological Test Pits in Arsalia

The ridge site of Arsalia is located at the southern fringes of Ayios Georgios (Figure 10A). During the pedestrian survey we mapped a wide range of building materials, artefacts, and traces of production facilities. Earlier testimonies described the existence of columns, decorated capitols, and reliefs [26], none of which were found. On the basis of the surface concentrations, Arsalia was categorised as an MR-LR farmstead, the largest of its kind, covering an area of 2.6 hectares. The dated pottery scatters also indicated human activity at the site during the Hellenistic, ER, Byzantine/Medieval, and Ottoman periods.

By analysing the regional cadastral map of 1934 and the aerial photographs predating the 1970s, we noted that agriculture in the area was practiced parallel to the hillside’s contours, with hedges of trees marking agrarian boundaries [74] (Figure 10C). This traditional, pre-mechanised approach must have served as a lynchet-like field terracing based on the local gradient, which would not require wall construction, especially considering the soft lithologies found in the area [32,75,76]. Land reclamation in the 1970s greatly altered the field arrangements in Arsalia, shifting the plots’ orientation to an upslope–downslope axis. As oral testimonies have supported, this led to the partial levelling of the natural relief and the substantial destruction of the archaeological site.

Five test pits (T1–T5), approximately 2 × 1 m each, were distributed along a NE-SW transect, extending from the highest part of the ridge (T4 and T5), to the footslopes (T3 and T2), and finally to the valley bottom (T1) (Figure 11). In T5 and T4, archaeological levels were found directly above the bedrock and included an accumulation of large tile fragments, mostly horizontally deposited, with building materials (plasters and small blocks of stones), pottery, charcoals, and ash. This unit was interpreted as a ‘destruction layer’, although in situ constructions (e.g., walls) were not found. In T5, a collected sample of pottery and glass indicated a Late Hellenistic–ER dating of the deposit (c. 2nd century BCE–1st century CE). In T3, on the upper slopes, a wall belonging to a linear structure, likely having a domestic context, was excavated in contact with the bedrock, and had a few materials dated to the Roman period (c. 2nd–4th centuries CE).

In the lowest parts of the topography, T2 and T1 revealed colluvial accumulations containing scattered artefacts (tile and pottery fragments) throughout the profile. The base unit of T1 showed evidence of widespread burning, with abundant large pieces of charcoal near the contact area with the bedrock. Macroscopically, this assemblage seemed to correspond to a palaeosol with anthropogenic influence, preserved in situ and, thus, connected to its use in Arsalia. A charcoal sample extracted from the bottom of the stratigraphy of T1 had a calibrated date of 308 ± 100 BCE. The topsoil in all the test pits, corresponding to the tillage horizon, was hard to differentiate in terms of the matrix composition, texture, and inclusions, and gave a homogeneous view of colluvial bodies mixed with disperse tiles, pottery, and rock fragments. The pottery from these units was very fragmented and eroded, with the few datable examples belonging to the Roman and Ottoman periods.

The combined study of archival sources, archaeological remains, and sedimentary stratigraphies suggests that during its earliest period of use in Hellenistic–ER times, Arsalia was established in a setting with little or no available soils. This conclusion is supported by the discovery of architectural remains placed directly on the bedrock (T3), as well as artefacts from the lowest sedimentary facies of the test pits (T4 and T5), none of which predated the Hellenistic period. Additionally, radiocarbon dating from the bottom of the ridge (T1), although slightly older than the relative chronologies of the material culture, indicated that the first sediment accumulation visible today commenced around 308 ± 100 cal BCE. The geomorphological analysis of alluvial landforms in Xerolakkos placed a major episode of extended soil erosion as occurring from the Middle Bronze Age (c. 1800 BCE) onwards, which only ceased during the ER-MR period (c. 2nd–4th centuries CE) [33]. Thus, Arsalia was subject to strong soil loss before the foundation of the site, whereas, with the beginning of human activity, presumably during the Hellenistic period, sediment also started to accumulate in situ. This process continued into the Roman period and was likely associated with the introduction of a new land-use strategy against erosion, like the field terraces documented in the middle 20^th-century CE landscape. Such terraces would have facilitated the progressive accumulation of colluvium in the upslope areas of the hedges [74]. Although the absolute dating of the lifespan of this system is not yet available, it is reasonable to argue that the same or a very similar technique took place in antiquity.

5. Discussion

5.1. General Remarks: Methodological Feedback, Site Visibility, and Models of Settlement

The Grevena Project’s database provided a solid foundation for mapping the archaeological landscape of the study area. In addition to the legacy sites, the new integrated fieldwork revealed a much denser record than previously considered, adding 64% more sites in the survey area. This result reaffirms that the Grevena Project had relied primarily on a technique involving visits to locations directed by local informants and oral testimonies [10]. This methodology, combined with the large-scale scope of the project aiming to cover the entire Grevena region, explains why they missed sites that were identified thanks to our methodology.

New material collections from the same sites contributed to refining the relative chronological distributions. In 25% of the revisited locations, new cultural periods of use were traced for the first time, filling a gap in the knowledge of the site’s history. The other side of the coin, however, is the high percentage (almost in 50% of the old sites) of locations where less diagnostic material was identified, compared to the findings of the Grevena Project some 35 years ago. Similarly, resurveying in Boeotia (central Greece) indicated that approximately half of the sites mapped during the 1970s and 1980s are today either no longer visible as artefact concentrations or have changed in terms of their spatial characteristics due to evolving land-use practices [3,47]. Unlike the Boeotian case, this part of Grevena has seen very little change in land use between the original project and this new survey (Table 2) and, given that most (75%) of our fieldwork was conducted under excellent visibility conditions, the explanation for Boeotia cannot account for the disappearance of surface finds here [10]. Over time, erosional episodes—particularly along the elevated areas—should have dislocated site debris and dispersed materials towards the valley bottoms, where colluvial sediments (e.g., as observed in the Arsalia test pits) would have buried them [33]. However, this is not likely to have occurred within a temporal scale of 35 years. In the absence of other explanations, it seems that the material debris of the studied landscape is generally diminishing, and thus artefact collection during the initial Grevena Project survey may partly explain why fewer artefacts are exposed today (cf., [50]).

The outcomes of this new survey reveal aspects of past habitation models in northern Grevena. The long-term analysis shows that unwalled or naturally exposed settlements were generally preferred, although certain sites were located in more protected areas and, in two cases, fortifications were constructed (e.g., Ayios Nikolaos and Paleoglas, see Section 5.2). It is also noteworthy that no human-made tells (or toumbes) were found. This form of settlement was widespread in the eastern, central, and parts of western Macedonia (e.g., Kozani) during the Neolithic period, but mostly during the LBA [77,78,79,80]. A closer look at the prehistoric settlement traditions of central and eastern Epirus, on the other hand, a landscape with physical characteristics very similar to Grevena, has stronger connections with our findings, especially from the LBA onwards [81,82,83].

Our size estimates are a breakthrough for understanding the surface record of western Macedonia, where comparative large-scale studies are, to our knowledge, lacking. Overall, the data suggest that diachronic habitation was expressed through very small sites, typically under 1 ha. The period-per-period resolution was insufficient to extract the temporal fluctuations in site extent due to the small sample numbers. Nevertheless, some general observations can be made as follows: No settlement exceeded 1 ha before the LBA. The phases with increased single-period sites, like the LBA-EIA and the MR-LR, provided a more detailed view. The LBA-EIA sites displayed some variation, with the bulk of the settlement beings less than 1 ha, yet with two cases occupying twice as much space. On the other hand, the MR-LR sites displayed a high degree of homogeneity in space use, clustering around 0.5 ha. Finally, a few examples, all of which were multi-period in use, appeared significantly larger than the rest (Figure 9). Notably, the lifespan of these outliers (except Upper Dexameni, where Neolithic activity has also been documented, and Ayios Nikolaos, with evidence of use in the EBA) extended from the LBA onward to the Roman period.

Off-site activity must have existed, but the fragmentation of the evidence we have recorded at this stage does not allow for further interpretations comparable, for example, to those for southern Greece, where dense material concentrations reported outside site limits have been interpreted as smaller habitation foci or remnants of land-use activities (e.g., manuring) [45,50,51,52,54,55,59,84]. Our data, if anything, resemble more the landscape of Langadas in central Macedonia, where large parts of the surveyed area were almost or completely devoid of surface finds [85]. In northern Grevena, even within site limits, the surface record in the study area is characterised by low-density, discontinuous scatters. This, combined with the small extent of material dispersions, suggests a form of sparsely nucleated occupation. Furthermore, domestic and productive activities were likely intertwined within the settlement area, as evidenced by the traces of craftmanship (e.g., kiln remains) found at the MR-LR rural sites. In contrast, it was not possible to differentiate between domestic and other contexts at the prehistoric sites.

5.2. Setting a Long-Term Synthesis of the Surface Archaeological Record in Northern Grevena

The combined survey results show that the prehistory, excluding the LBA, are significantly underrepresented in the surface record of the study area, with less than five sites dated per major period. The Early Neolithic (EN) included three cases, two of which continued to be occupied during the following Middle Neolithic (MN) period. This very small number discourages general assumptions, yet all of them were located on prominent topographic spots. Thus, on a micro-regional scale, our data enrich Wilkie’s hypothesis that EN habitation in Grevena primarily developed along the riverways, by the occupying of the upper terraces close to Aliakmon and its major tributaries [16,18,20]. However, our findings indicate that more diverse parts of the landscape might have been settled. The Grevena Project posited that these sites were smaller than 1 ha [16,20], which agrees not only with our own surface estimates from Kivotos-Gazep Rahi (c. 0.6 ha), but also with the excavation reports from other parts of Grevena, Knidi-Kremastos (c. 0.25) [86] and Knidi-Matsouka Rahi (c. 0.15–0.2 ha) [87], as well as, in some cases, from Kozani, like Mavropigi-Filotsairi (c. 0.5 ha) [88].

No Late (LN) or Final Neolithic (FN) surface materials were identified, suggesting at first glance a discontinuity of habitation in the area. According to the Grevena Project, most of the EN sites ceased to be inhabited by the end of the MN, while the evidence for occupation until the end of the Neolithic appeared very restricted in general [16,18,20]. This lacuna has been explained as a failure of the first agropastoral communities to adapt to the physiography of the regional environment, which was consequently abandoned [86]. However, a very different picture emerges from the bordering regions of western Macedonia, where survey and excavation projects over the last three decades have recovered a large number of LN-FN sites, indicating dense population during this period [80,89,90,91].

Besides socioeconomic explanations, other factors might have influenced our current knowledge of prehistoric human activity in Grevena. For example, the pedo-sedimentological archives of Xerolakkos have highlighted the potentially serious impact of taphonomic processes on the visibility of Neolithic and EBA sites [33]. Two major episodes of widespread soil erosion and wash-off were attested to during the EN (c. 6300–6000 BCE) and from the Middle Bronze Age (MBA) to the Roman period (c. 1800 BCE–200 CE). These events must have removed, discarded, and covered archaeological strata of existing or previously occupied settlements, especially in elevated areas (tops of ridges and hills) [92]. Yet, the survival of some locations today with exposed prehistoric materials demonstrates that the spatial extent of such phenomena was localised and patchy, even within the studied catchments. For example, the partly excavated EN site of Knidi-Kremastos, located on the footslopes of a low ridge some 7 km to the east of this case study, includes in situ remains [86]. More notably, in the flat plateaus of Kitrini Limni (Sarigiol), again in Kozani, the Neolithic sequences are well preserved due to the accumulative rather than erosive nature of the local sediments [93]. Thus, post-depositional dynamics must be carefully examined when interpreting the prehistoric surface record in Grevena.

A growing trend in settlement density was observed for the LBA, which was further consolidated during the following EIA, when most (75%) of the LBA sites continued to be inhabited. This is also supported by Wilkie [16], and finds parallels in the wider region of northern Greece, highlighting the absence of significant breaks in habitation and material culture [94]. However, exceptions have been identified in western Macedonia, such as in the Middle Aliakmon Valley, where only 4% of the LBA sites remained in use at the beginning of the EIA [91]. In addition to the high degree of continuity in the study area, the appearance of new settlements during the EIA provides evidence of population expansion across the landscape. Likewise, another important aspect of the end of prehistory and the EIA in this part of Grevena is the emergence of sites that would develop into the largest, multi-period habitation centres during the subsequent historical periods. This is the case for Ayios Petros (Myrsina), Ayios Nikolaos (Ayios Georgios), and Paleokastro (Syndendro). While the spatial character of their initial establishments cannot be determined at present, the period-by-period site distributions, as well as the existing evidence for a kind of locally based political formation during the Hellenistic period (e.g., Ayios Georgios, see below), imply that the established hierarchical relationships of the historical times in the region could have originated in the beginning of the 1st millennium BCE, earlier than the reach of historical phenomena, such as the expansion of the Macedonian Kingdom, have suggested (cf., [36,37]).

Archaic and Classical sites are represented in small numbers in the survey results, implying some kind of settlement decrease. In fact, the original Grevena Project survey argued that no Archaic sherds had been found at all in the wider region [16]. This picture probably stemmed from a combination of dating issues. The first involved a strong bias toward ‘searching’ exclusively for these periods, relying on the well-dated pottery imports from southern Greece. However, these types represent a small minority of the surface artefacts, since in Macedonia they were primarily used in funerary contexts that are harder to trace through surveying [95]. The second barrier concerned the lack of published excavation stratigraphies corresponding to the Archaic–Classical periods. Furthermore, parallels from neighbouring regions reveal the tendency of the local material culture to preserve ‘older’ elements, which was consolidated during the LBA-EIA, and with little typological or technological change over time. This was the case for handmade pottery shapes and decoration patterns that often persisted until the Classical period, making it difficult to discern, for instance, Archaic potsherds from earlier examples—especially in the fragmentary state of surface ceramics [81,96]. This has created the common misconception that ‘it is only in the 4th century [BCE] that that archaeological record seems to pick up again’ [16] (p. 1752). Thus, the most likely explanation for the limited site information we see for the Archaic and Classical periods in Grevena has to do with a weakness to identify more artefacts dated to these periods.

By the mid-4th century BCE, towards the end of the Classical period, the expanding Macedonian Kingdom under Philip II had apparently assimilated the region of Grevena, where a form of political power organised around civic institutions was already being exercised locally. This can at least be inferred from the study of a Hellenistic inscription found at Ayios Nikolaos [36,40,41]. Along the eastern side of the Pindus, a series of new fortified sites have been claimed as marking the territorial boundaries between Macedon and the Kingdom of Epirus to the west, introducing a timeline of antagonism and military hostilities between the two entities as during the early Hellenistic period [16,39,97,98]. However, the distribution of these structures seems to be space specific and concentrated in the Pindus mountainous zones. In the study area, for example, life during the Hellenistic period continued in pre-existing communities, whereas, in rare cases, there is evident use of new locations (e.g., Plastires in Kokkinia). In principle, settlements remained unwalled. At the same time, Ayios Nikolaos and Ayios Petros must have reached their maximum size, to judge from the spatial dispersions of the Hellenistic materials. Therefore, the impact of such historical events, regardless of their importance for the political geography and economy of the wider region [99], did not alter the local settlement dynamics.

During the same period, we also begin to see the origins of a new practice related to the economic exploitation of the landscape. This is advocated by the chronostratigraphic reconstruction of Arsalia, where both the relative and absolute dating indicate that a land management technique was introduced to deal with an ongoing erosion episode, which had left parts of the local environment completely deprived of soils [33]. A form of field terracing, observed in the traditional agriculture of the region, could lie behind such a transformation. Terraces would have been efficient at conserving soils in situ against erosional processes, and would have increased soil weathering through tillage, leading to better crop productivity [75,100,101]. Thus, an intention towards improving the agricultural yields but also investing in the long-term sustainability of a previously degraded environmental setting can be argued for the study area.

The site distribution during the ER period is limited, with an apparent abandonment of locations besides the major, pre-existing settlements. This phenomenon could reflect the aftermath of the Macedonian Wars, a long chain of military conflicts between Macedon and Rome from the end of the 2nd to the middle of the 1st century BCE, which essentially led to the indisputable dominance of Rome in Greece [99]. Consequently, Macedon was divided into four self-governing ‘republics’, of which quartae Pelagoniam incorporated Grevena. Economic and political liberties were severely curtailed and commercial activities and intermarriage between citizens of the different quartae were prohibited [40,41]. These developments would have created a context of economic isolation for the existing communities that, conversely, could explain the limited activity in the study area, if not a decrease in population. Reviews of the intensive urban surveys of Boeotia also indicate that, despite the presence of élite families in the Greek countryside, very little was invested in the rural economy during the early phase of Roman rule. This, along with the significant reorganisation of inner urban structures, also give the impression of a largely deserted landscape [73].

The real change in site distribution occurred during the Middle and the start of the Late Roman period. This phenomenon was connected with the establishment of several single-period rural sites that have been interpreted as villas and farmsteads (c. 2nd–4th centuries CE). The survey data illustrate that the material culture of these sites shared specific characteristics (e.g., architecture, ceramic assemblages, and production facilities), and describe a pattern of structures with very similar spatial configurations. These sites appear to have been clustered outside the core of the main settlements but, at the same time, occupying new parts of the landscape.

The density, along with the lifespan of these sites, indicate one of the busiest periods of human activity in northern Grevena. A similar observation was made in a previous assessment of the regional Roman record [19], though without the site categorisation proposed here. Ayios Georgios was the focal point of this new pattern, with six sites located within a radius of less than 2 km around the village. Considering the wider zone and the Xerolakkos catchment, four additional locations may have been part of the same network, as they were not associated with any other larger settlement nearby (Figure 7 and Figure 8). Among these, Arsalia appears to have been the largest (c. 2.5 ha), covering more than double the area of all the other recorded examples of this type.

All this prompts us to consider the kind of socioeconomic system that could have supported this new regional pattern. Paleoenvironmental proxies from Xerolakkos, for example, have pointed to the extensive use of the surrounding landscape at that time, whereas the pre-existing erosion, suggested in Arsalia, appears to have been minimised during the 2nd–3rd century CE [33]. This reinforces the chronostratigraphic importance of Arsalia, and implies that the introduction of a terracing system was indeed successful not only at preserving but also nourishing the soils, while decreasing the total sediment reaching the valley bottoms. Moreover, a location preference for the several headwaters in the area, where the water supply would have been easy and continuous, could have also supported the expansion of agrarian activity. Following these lines of evidence, we believe that the farmstead phenomenon was primarily driven by an attempt to intensify agriculture in the area, perhaps for achieving a greater surplus destined for extra-regional markets [33,71,72].

The spatial meaning of the new site distribution requires more analysis. Looking at other regions of Macedonia, especially in the central and eastern parts, we see a similar clustering pattern emerging from the end of the 1st century CE onward. This development was favoured by the Pax Romana and its emphasis on the role of big estates in transforming agricultural production and regional economic networks [70]. Arsalia, for instance, could fit the profile of a wealthy villa estate. Yet, the degree of standardisation between various aspects of Arsalia and those of several other smaller villas/farmsteads better describes a broader network of independent farms in Grevena. It would, thus, be reasonable to consider a regional change in land-use distribution or tenure as aimed partly at reinforcing the soil potential, rather than as simply an agricultural expansion by the opening up of the landscape. At this stage, we cannot determine whether these farms were monopolised by a single, élite family leasing to dependent workers, or if they were owned by different middle- and lower-class farmers [2,70,71,73]. Nevertheless, their architectural sophistication could be an indication that at least some farmsteads became the epicentre of wealth investment in the local landscape, and it would not be surprising if parts of or whole families who could afford this living resided in them on a permanent or semi-permanent basis.

This regime must have ended by the 5th century CE, when the activity of most rural sites ceased. Dated materials belonging to this part of the LR period are indeed very scarce for the study area. Arsalia, Ayios Nikolaos, and Paleoglas represent the only examples where some presence was maintained during the Byzantine/Medieval and Ottoman periods. In connection with this change, Rosser [19] argued for the abandonment of unprotected sites and the construction of fortifications as a response to the external raids that started from the late 4th century CE onwards. The evidence from neighbouring regions, such as Kastoria, draws a similar picture of concentration on a few walled towns to cope with the rising danger of the Visigoths [102,103]. The wall structure in Paleoglas, and perhaps in Ayios Nikolaos, could have been dictated by these documented events, yet, in the absence of dated stratigraphy this remains conjectural. Independently of the following historical trajectories, it is worth noting, as a closing remark, that certain aspects of ancient land-use patterns, such as the terracing of Arsalia, may have persisted until recent times according to our remote-sensing analysis.

6. Conclusions

We conducted a multi-scale and multi-disciplinary archaeological survey of northern Grevena based on site-revisiting strategies for the interpretation of the surface record under a diachronic perspective. This work stands out for its integration of geospatial technologies (GIS and remote sensing) with targeted archaeological and geomorphological fieldwork. We discussed our results from both methodological and interpretative angles.

The systematic fieldwalking yielded a considerable new number of sites, shedding light upon an archaeological landscape much denser than the previous regional surveys had suggested. This confirmed, for example, the location-based strategy of the Grevena Project, which resulted in missing multiple sites in the study area. Furthermore, a major contribution of our research is the generation of a wide range of quantitative (e.g., site size) and qualitative characteristics (e.g., site classification), enabling both a period-per-period and long-term analysis of human activity and settlement histories. This is the first such work conducted in the wider area of western Macedonia. Finally, the targeted geoarchaeological approaches highlight the importance of studying local pedo-sedimentary records for weaving a meaningful narrative of the geomorphological, taphonomic, and land-use dynamics involved in site history.

More crucially, the successful implementation of these approaches calls for reconsidering the role of extensive surveys in the study of cultural landscapes. Our workflow showcases how integrated resurveying techniques can not only validate and reassess legacy information but are also capable of producing new datasets on their own grounds. This demonstrates that revisits to old sites must be a necessary rather than secondary strategy in a surveyor’s toolkit.

On the archaeological side, our results show that the prehistoric periods before the Late Bronze Age are less represented than historical times in the study area. However, this fact should not assume long-term discontinuities in settlement. Instead, taphonomic phenomena, such as pronounced soil erosion, must be included, at least in this case, to explain the lack of surface material culture for certain periods. A model of long-term occupation was often noted between the Late Bronze Age and Late Roman periods, coupled with a more distinct division between types of activities (e.g., cemeteries, farmsteads, and small settlements). In detail, the Late Bronze Age saw increased site numbers, with a high degree of continuity into the Early Iron Age, suggesting an overall population expansion. During this period, we begin to see the spatial diversification of a few settlements that would evolve into the man foci of habitation. This phenomenon implies that the hierarchical structures of the historical periods may have already been shaped among the communities of northern Grevena before any influence of the supraregional political entities (Macedonian Kingdom and Roman Empire).

Particularly during the Hellenistic period, habitation was dense and focused on the same spatial pattern as during the previous periods. This leads to questions about the supposed impact of the war episodes between Macedon and Epirus. A rather different view emerges from the Early Roman period, when the socioeconomic aftermath of long-term military events (e.g., Macedonian Wars) could have led to a seemingly restricted use of the landscape, although the previously established settlement order persisted. It was during the Middle and the start of the Late Roman period that habitation shifted, with single-period farmsteads found in new locations, indicating agricultural intensification. This was possibly achieved with the implementation of terracing systems (e.g., in Arsalia) that limited soil erosion and increased field productivity. However, this new landscape configuration was short-lived and must have ended by the 5th century CE, which is marked by site abandonment and the decline of rural activity. At that point, the alleged impact of foreign raiding on local communities becomes elusive due to the scarcity of dated evidence.

Concerning the settlement dynamics of these historical periods, the most important contribution of our research is an understanding of the spatial and temporal effects of historical events across the landscape. Our work in Ayios Georgios, for example, introduces the potential of using multi-proxy survey data to discuss how political and economic power was ultimately expressed not only in the ‘central’ places (e.g., Ayios Nikolaos), but also in relation to the rest of the local population. Future systematic excavations, coupled with interdisciplinary strategies at both site- and landscape-based scales, will enhance the dating resolution of site distributions and provide more clues about land-use practices and site taphonomy. This will improve our knowledge of the long-term landscape evolution and the nature of human habitation in ancient Macedonia.