Introduction

1The Thessaloniki Plain (also called Plain of Central Macedonia), located Northern Greece (Fig. 1), has been an important area for human occupation since its earliest occupation during the Neolithic, around the middle of the 7^th millennium BC (Demoule and Perlès, 1993; Grammenos, 1997; Perlès, 2001). From the early Neolithic, settlements were established in the western part of the plain (Rodden and Wardle, 1996). Tribes of farmers were the first inhabitants of the area; subsequently, States and Empires succeeded and dominated the area from Macedonian times until the Ottoman period. The Macedonian State, with kings Philip the 2^nd and Alexander the Great (Alexander the 3^rd), was the most famous culture in the area, leaving behind many archaeological remains such as cities and roads. The discovery of ancient Pella’s ruins by Photios Petsas in the 1960s (Petsas, 1978; Akamitis et al., 2004) and Vergina’s tombs by Andronikos in the late 1970s (Andronikos, 1984, 1993) revealed the historical importance of the changing morphology of the area. Contemporaneous authors (Strück, 1908; Hammond, 1972; Bottema, 1974; Bintliff, 1976) based their different scenarios of the progradation of Thessaloniki’s plain over the last millennia only on the interpretation of literary sources from Thucydides, Herodotus, Strabo, and Livy (Fig. 1). The interpretation of ancient sources is a subjective method and frequently leads to various contradictions of many environmental aspects. Recently, chronostratigraphic evidence (Ghilardi, 2007; Ghilardi et al., 2008 a and b; Ghilardi, 2010; Ghilardi et al., 2010, 2012) has helped to reconstruct the historical developments in the region vs. the changing landscape and the present article provides and suggests new perspective on archaeological research in the area.

Geological and geomorphological settings

2The Thessaloniki plain is the largest (~2200 km²) deltaic area of Greece. It is situated in the northern part of Thermaikos Gulf (Fig. 1) and limited by high mountain ranges (Figs. 1 and 2). To the west is the Vermion (2050 m) Mountain (Mesozoic limestone); the Paikon (1650 m) Mountain (Mesozoic limestone) lies in the north; and the Neogene hilly ranges of Pieria (1750 m) and Chortiatis (1050 m) surround the south and east, respectively. During the early Neogene, the broader area of the Thessaloniki plain and Thermaikos Gulf formed an elongated NNW-SSE trending tectonic depression (graben), called the Axios-Thermaikos depression (Syrides, 1990; Dinter and Royden, 1993). The depression was gradually filled up with mainly clastic deposits (conglomerates, sands, clays) and locally derived calcareous sediments (limestones, marls), which were deposited in successive complex depositional palaeoenvironments during the Miocene (fluvial, continental red beds, brackish clays, sands, limestones), Pliocene (fluvio-lacustrine sands, silts, lacustrine marly-limestones), and Pleistocene (fluviatile) periods. In the southeast part, the Thermaikos Gulf connects the two main deltas of the plain (Fig. 1) created by the Aliakmon and Axios. The Thermaikos Gulf, which is part of the north Aegean Sea, receives freshwater from four major rivers: the Aliakmon, Axios, Gallikos, and Loudias (Poulos et al., 2000; Lykousis et al., 2005). The total drainage surface area of these rivers is approximately 35000 km² (Ghilardi, 2010). To avoid flooding during the heavy rains, various public hydraulic works were installed by the New York Foundation Company in the 1920s, as peripheral channels and artificial river banks (Ancel, 1930). Therefore, the topography and morphology of the plain were totally modified. Today we just have some remnant oxbow lakes resulting from from Little Ice Age hydrodynamics. The Giannitsa Lake (Borboros Limen and Loudias Lake in Antic times) was totally drained for agricultural purposes in the 1920s. The plain provides the largest agricultural area of Greece based on irrigation for intensive cultivation (Füldner, 1967), which also modifies the flow of the rivers Aliakmon and Axios on which several dams were built to control seasonal water discharges during. NEDECO (1970) studied 200 drill holes (drilled to a maximum depth of 12 m) and found that the sediments were mainly deposited under marine and lagoonal conditions during the Holocene (Bottema, 1974). These researchers found marine sediments at 10 m depths near Pella, where the area was covered with lagoonal, limnic, and fluviatile deposits. The sediments carried by the rivers Aliakmon and Axios spread over the marine deposits during the lagoonal stages over all the plain. These gradually filled the former bay, giving the plain its present shape. In the NEDECO (1970) study, the phases of this construction were not known and had not yet been studied. Also, no dating of the material was available at this time. Since the end of the maximum of the last postglacial transgression, studies have been undertaken to evaluate the rise of the sea level in the northern part of the Aegean Sea (Lykousis et al., 2005; Vouvalidis et al., 2005). Using malacological analyses and ¹⁴C dating, the results showed a classical sea-level rise in this area, comparable with findings from other studies around the world (Pirazzoli, 1996; Bruse et al., 2001).

Human occupation since the Neolithic in Central Macedonia: The presence of numerous important archaeological sites and related geoarchaeological questions

3The oldest archaeological remains in the Macedonian plain are dated from the Neolithic period. The excavation of the Nea Nikomedia settlement, by Robert Rodden and the joint Cambridge/Harvard team in the 1960s, led to the assumption of proximity of the site to the sea or to a lake. The earliest occupation of the site began probably during the early Neolithic ca. 6500-6400 BC and the upper phase of reoccupation date from late Neolithic period, ca. 5500-5300 BC (Bintliff, 1976; Rodden and Wardle, 1996). The typical houses of the mound revealed that the sediments used by the inhabitants were extracted very close to mud and reeds (Rodden and Wardle, 1996). Many shellfish remains were identified and Robbert Rodden with his team concluded that the Nea Nikomedia settlement was probably close to a marshy lake or inlet during its earliest occupation (Rodden and Wardle, 1996). Shackleton (1970) closely examined the fauna that was identified and dated by Rodden and he confirmed the theory of a position close to the sea, with brackish environmental conditions. Hammond (1991), a well-known specialist in the archaeology of Macedonia, assumed that the alluvial plain created in central Macedonia was occupied first by settlers during the late Bronze Age, but neither literary sources nor bibliographic references supported this theory. During Roman times, two roads were constructed through the Thessaloniki plain. The first one, called the Early Roman Road, was built at the foothill close to Edessa, Pella, and Thessaloniki. This road was built around the 1^st c. BC. The second road built by the Roman Empire around the 3^rd c. AD (Bintliff, 1976) was shorter and was called the Late Roman Road. For the first time, we have a road passing through the central part of the plain, direct from Pydna to Thessaloniki. In the middle of the plain, we have the Klidi Arch (N40 34’434/E22 37’352) dated from the 3^rd c. AD. The arch was the only remnant from a Roman bridge built with eight or nine arches. Its direction was WSW-ENE, with an initial length of 187-190 m, and the remains are probably the 3^rd or the 4^th arch of the edifice (Delacoulonche, 1859). However, no river has been found under this arch. The hypothesis advanced by Strück (1908) and Bintliff (1976) refers to a bridge spanning the Aliakmon River or a junction between the rivers Aliakmon and Loudias at the arch. However, no manuscripts or any description of this part of the plain can provide evidence of this. The Klidi Bridge was already known during the Medieval times when Emperor Alexis and his army traversed the road in AD 1078 (Delacoulonche, 1859).

Palaeoenvironmental methods in deltaic context

4Twenty boreholes (Fig. 2) were drilled from September 2004 to April 2008 in the western part of the Thessaloniki Plain (west of the modern Axios river course) and revealed the stratigraphy to a maximum depth of 11 m (Ghilardi, 2007; Ghilardi et al., 2008 a and b, 2010, 2012). The locations were decided on the basis of the NEDECO drill hole data (NEDECO, 1970) and where contact between marine and terrestrial facies was likely to be recorded. To find out more about lateral and vertical facies correlations in the western part of the hessaloniki plain, palaeoenvironmental reconstruction was done based on lithoand biostratigraphical studies. All the samples were wet-sieved through a wire screen (0.4 mm mesh). The residue was examined under a binocular microscope and all identifiable shells and characteristic fragments were picked. Palaeontological determinations were made to a generic and specific level in order to identify the different facies. The chronostratigraphy of the cores was determined using a series of 27 AMS and conventional radiocarbon determinations derived from in situ shells and peat samples (Ghilardi, 2007; Ghilardi et al., 2010, 2012). Granulometry involved analysing the grain-size distribution using laser diffraction, thus allowing a characterisation of the energy of deposition. Magnetic susceptibility measurements were performed at CEREGE (Aix-en-Provence, France) and at the Laboratory of Physical Geography in Meudon in order to identify different sediment sources. The sediment cores were sampled (at ~5 cm resolution, except in levels, which included reworked material) providing 500 samples in total. In addition to the low field magnetic susceptibility, normally measured at the 976 Hz frequency, measurements were also taken at the 15616 Hz frequency. The sensitivity of the MFK1 susceptibility meter is of ~3x10^-8 SI at 976 Hz. Finally, stable isotopes analyses (¹³C and ¹⁸O) were performed on carbonate together with X-Ray Diffraction measurements (XRD) in order to identify dry/wet climate phases and to characterise the mineral composition of the carbonate-rich lacustrine sediments. The aim of these analyses was to reconstruct the palaeoclimatic evolution of the area and to distinguish wet and/or dry periods during Holocene that could also be recorded on a regional scale (Psomiadis et al., 2009).

Holocene palaeogeographic reconstruction of the Thessaloniki plain

5The borehole evidence, especially the stable isotope data indicate a high influx rate and overflowing water levels at the base of core NN2 (Ghilardi et al., 2012), suggesting that a shallow fresh water lake occupied the western part of the modern Thessaloniki Plain close to the site of Nea Nikomedeia from the Younger Dryas deglaciation until the early Holocene (8000-7500 BC; Fig. 3). The fluvial sediments rich in carbonates observed in core NN1 (Fig. 2) also suggest that small streams flowing from the Vermion Mountains were very active and flooded into this lake. This interpretation agrees well with regional climatic reconstructions; the European lake level database (Harrison et al., 1991) indicates wetter conditions with higher lake levels at ca. 10000 cal. BP (8000 BC) than at present in southern Europe. Similarly, it seems that associated vegetation in the region changed from a glacial landscape to interglacial forests (Lawson et al., 2004). Although wetter lacustrine conditions at that time is widely acknowledged, the extent of this carbonate-rich lake cannot be precisely defined, since only two cores indicate lacustrine environments. The pretransgressive surface was not reached in the different cores (Fig. 2), preventing the recognition of potential lacustrine sediments beneath the lagoon/marine deposits. The existence of the lake can be attributed to several factors, such as the presence of springs located on large travertine deposits in the vicinity of the cities of Veria, Edessa and Naoussa (Faugères, 1978). These carbonate deposits were formed during the Pleistocene and were probably created by local streams flowing over calcareous rocks (Faugères, 1978). During the subsequent Mediterranean marine transgression, the sea gradually inundated the whole area from the SE to the NW (Lykousis et al., 2005; Ghilardi, 2007; Ghilardi et al., 2008 a and b, 2012), depositing the shallow marine and lagoonal/brackish sediments dating from ca. 6000-5800 BC 4 km east of the Nea Nikomedeia settlement (Fig. 3). The maximum extent of the Holocene marine transgression is dated to ca. 6000-5800 BC. Frequent intrusions of salty/brackish water into the fresh water lake strongly modified the quality and the salinity of the water at that time. Core NN4 (Fig. 2) indicates the presence of brackish environments, probably dating from the beginning of the 6^th millennium BC. It is likely that the desertion of Nea Nikomedeia could be explained by these environmental changes (i.e., salt intrusion into the lake) due to rapid landscape changes from a freshwater lake to a brackish lake. The interpretation of cores NN5 and NN6 suggests that such a shallow marine/lagoonal stage occupied the western/central parts of the plain from ~5800 BC until ~4100 BC (Fig. 3). Subsequently, a second freshwater lake stage started at around 4100 BC, before a main phase of peat accumulation developed from ca. 2000 BC to 1500 BC (Ghilardi et al., 2008b, 2012). These peat deposits might be linked to a gradual return to mild climatic conditions more similar to the modern environment, with less precipitation. Studies of the Mediterranean region indicate a pronounced arid phase occurred at ca. 1600-1400 BC, coinciding with low water-levels in African lakes and also seen in speleothem records (Gasse, 2000; Drysdale et al., 2006; Zhornyak et al., 2008). This drier period has been suggested as a key factor in the collapse of Old World civilisations in the Mediterranean region (Drysdale et al., 2006). According to Ghilardi et al. (2008 a and b, 2010), the area of the Roman Bridge of Klidhi (situated in the lower part of Thessaloniki Plain) was under marine conditions during the Archaic and classical periods (from the 7^th to the 4^th c. BC; Herodotus, Histories, book VII. Fragments 123, 124 and 127). During the early Roman Times (from 168 BC to the 1^st c. AD), there is a general deltaic progradation, which affects the whole area and Strabo considers that in those times Pella was connected to the sea and the ships used to reach the ancient Capital after sailing a narrow channel 22.2 km long (Strabo, Geography, VII, frag. 20 and 22). Ghilardi et al. (2008a, 2010) revealed the shape of the coastline during that times and it is evident that we have marine to shallow marine environments in the area of the Klidhi Bridge. Several lagoons were being formed and coastal barriers created (Fig. 3). But the edifice did not exist at this time and neither Livy (History of Roma, books 44 and 46), nor Strabo (Geography, VII, frag. 20 and 22) mention the existence of a shorter road, along the shoreline allowing us to infer that the bridge was not constructed during the 1^st c. AD. Gradually, due to alluviation of the Aliakmon and Axios Rivers and the constant deltaic progradation, transition to lagoonal conditions occurred during the Hellenistic period, and was well attested during Roman Times (Fig. 3). Such palaeoenvironmental conditions were probably caused by the creation of a series of coastal barrier, as the NEDECO report (1970) suggested, one of which was the so-called barrier of Klidhi-Kimina-Chalastra-Sindos (Ghilardi et al., 2010). Several isolated lagoons developed, which were largely filled by sediments, ranging from silts to medium and coarse sands, and had connections to the sea probably created by frequent coastal barrier erosion. It is this environment this could justify the construction of the Klidhi Bridge. Based on the microfaunal identification and sediment analyses from the different cores, we surmise that two lagoons were separated by an ancient coastal barrier. A shorter road (Bintliff, 1976) was then constructed and branched from Pydna to Thessaloniki, and then to the Egnatia Road. Our work shows that the Roman road ran between two lagoons and was built over a coastal barrier ca. AD 300 (Fig. 3). The Roman engineers decided to establish a road over the newly created lower part of the Thessaloniki Plain. In order to avoid swampy/lagoonal areas, close the shoreline, they decided to develop the road some kilometres inland. The change from lagoonal/brackish environments to alluvial sedimentation is shown by the passage of a river depositing coarse sand after the 15^th c. AD (during the Byzantine and Ottoman periods; Ghilardi et al., 2010) under the central part of the bridge. The presence of a large wood fragment of Quercus sp. found below the ancient bridge, and which was probably a part of the bridge deck (Bintliff, 1976), at the transition between coastal barrier deposits and fluvial sediments, suggests that a violent flood probably damaged part of the bridge.

Conclusion

6Since the height of the last post-glacial transgression, the palaeogeography of the alluvial plain of Thessaloniki has been fashioned over the last 7500 years by the rapid growth of deltaic areas on a former extensive embayment of the sea. Our research offers a new perspective and understanding of the shifts of the shoreline since the Pre-Neolithic times based on a chronostratigraphic study. Twenty boreholes concentrated on the western and central parts of the plain around the ancient sites were analysed to reconstruct shoreline displacements and landscape evolution during the last 10 millennia. Herodotus, Thucydides, Aeschines and Aristotle considered that the sea was still close to Pella during King Philip II’s time. However, the chronostratigraphy revealed a lacustrine occupation rather than a marine presence at this time. However, we do not know the exact definition of of ‘sea’ that these ancient writers were using. It is probable that a lake occupying a large area, from Pella to the Neolithic site of Nea Nikomedeia, could have been considered a sea by these ancient authors. The fact that we cannot define the limits of the lake and the contact zone with the sea has prompted us to think that the “sea” of the ancient authors may have corresponded to a freshwater lake or a lagoon. We are sure that Pella was totally disconnected from the sea and the potential harbour was inland and totally silted-up. The bridge of Klidi constructed by the Roman Empire demonstrated that communications were more directed eastward than during the Macedonian period. The evident change in environmental conditions was confirmed by the position of Pella ca. 28 km away from the sea at the end the Roman period.