Archaeometric Investigations on the Significance of Limestones in the Roman Provinces of the Southern Lower Danube

Abstract

In contrast to several types of marbles used in antiquity, where different data collections do exist, similar approaches for provenance analysis are not available for limestones. In general, limestones are characterised visually and by the examination of sedimentological parameters. The usual microscopic investigation of limestones and the characterisation of limestone artefacts are often impossible because of the large samples necessary for these investigations. It is demonstrated in this paper that for the provenance analysis of limestone, the same established methods used for assigning marbles to their origin can also be successfully used in the case of limestone. We investigated a series of Roman quarries (three different quarry regions) in the north of the Haemus Mountain Range in the region of Marcianopolis and Nicopolis ad Istrum with the aim of establishing a first database (isotope analyses and trace element analyses) to correlate limestone artefacts. All in all, 42 quarry samples from the locations Marciana, Samovodene and Koevtsi were investigated. Fifteen artefacts made of limestone were analysed. Two samples from Marcianopolis were assigned to the Marciana quarry, and seven artefacts from Nicopolis ad Istrum originate from the quarries of Samovodene. For six more samples from Nicopolis ad Istrum, no match to one of the sampled and investigated quarries was found; thus, the existence of one or more quarries in the area is presumed. An extension of the database and the investigation of further limestone quarries in the area certainly will compensate for this deficiency.

1. Introduction

In general, a stonemason does not specifically differentiate in terminology between the terms “limestone” and “marble”. The stone industry is much more interested in the properties of the specific rock, particularly with regard to the capability of its surface to take a polish. This is in contrast to the scientific, petrographic nomenclature. Here, the term “marble” is clearly assigned to metamorphic rocks, while “limestone” is a not metamorphosed sedimentary rock and, in the case of metamorphic overprint, is transformed to marble. The geomaterial dealt with in this paper is exclusively limestone in the strict petrographic sense. In general, limestone is often used much more locally compared to pure white marble; however, there are many “coloured marbles” (many of them are petrographically limestones), among them the most valued stones in antiquity, which were traded all over the Empire. Scientific characterisation of limestone is based essentially on sedimentological investigations aiming at identifying the microfauna of the carbonatic sedimentary rock. The main task of these investigations is the visual examination of the macrofauna and the study of microscopic-thin sections of appreciable size (e.g., 4 × 4 cm). Considering the problems when taking samples of much smaller sizes for these investigations from marble artefacts, the sedimentological investigations and sampling of high-quality limestone artefacts may meet considerable difficulties. Thus, in this paper, we do not focus on the macro- or micro-petrographic features of limestone, but we apply chemical and isotopic methods, which are well proven in the investigations of marbles.

For the provenance analysis of white marbles, chemical analyses, isotope analyses, EPR (Electron Paramagnetic Resonance), cathodoluminescence, the analyses of fluid inclusions in the marbles and petrographic methods as a proxy technique were used in the past. In the case of white marbles, the provenance analysis based on the above-mentioned methods is well established. Specifically for the stable isotope composition of the different marbles, an enormous amount of data was published in the past. Recently, data on the chemical composition of the international marble quarries of antiquity were made available. However, despite figures and diagrams that are widely available, there does not exist a comprehensive digital database of marble quarries (e.g., [1,2,3,4]).

For technical reasons, EPR and the analysis of fluid inclusions in marbles (e.g., [5]) are rarely used today. Cathodoluminescence has been used for many years mainly in a qualitative way (e.g., [6]). For a multivariate evaluation, it has to be considered that the resulting descriptive outcome is not suitable for a statistic evaluation simultaneously together with other numerical data. However, recently, a quantitative approach using cathodoluminescence was published and successfully used ([7]). However, establishing the correlation between cathodoluminescence analyses obtained by different labs must be prioritised for a feasible comparison of samples (e.g., internal standards, instrument setup, etc.).

The intention of this work is, on the one hand, to investigate whether the established methods for provenance analysis for white marbles can also be applied to assign limestone to the corresponding source, and on the other hand, to examine the importance of the region’s limestone quarries for the production of local artefacts and architecture.

This work is the result of a close collaboration between the Austrian and the Bulgarian Academies of Science to unveil the production and use of the marbles in Roman Thrace (FWF project P 33042 “Fingerprinting White Marbles. Quarries and Cities of Roman Thrace, 1st–3rd century AD”). Due to the outstanding importance of different limestone quarrying sites within the reach of the Lower Danube north of the Haemus Mountain Range, these limestone quarries were also sampled and separately investigated. Furthermore, exemplary artefacts made of limestone in that region were analysed.

The two case studies presented here are related to the Roman cities of Nicopolis ad Istrum and Marcianopolis. Nicopolis ad Istrum was founded around 110 AD by Emperor Trajan to celebrate his victories over the Sarmatians and Dacians. Located near modern-day Veliko Tarnovo, Bulgaria, the city was strategically positioned on key trade and military routes that connected the Danube with major cities in the Roman Empire, facilitating trade with Pannonia and the Black Sea. Nicopolis ad Istrum became a significant regional hub, benefiting from navigable rivers like the Yantra and Rositsa and in proximity to vital mountain passes. The city’s territory extended northward to Sexaginta Pristina, south to Augusta Traiana, and west toward Melta, eventually becoming part of Moesia Secunda under Diocletian’s reforms. Settled by immigrants from Asia Minor, Nicopolis ad Istrum adopted a Greek-inspired urban layout [8,9,10].

Marcianopolis, one of the major Thracian cities founded by Emperor Trajan after his victory in the Second Dacian War, was first documented in a mid-second century inscription from Lambaesis. Strategically located 27 kilometres west of Odessus (modern Varna), Marcianopolis sat at a crucial crossroads linking southern routes to Scythia Minor and the Danube with the road from Odessus to Moesia Inferior. This advantageous position, coupled with an abundant supply of fresh springs from the karst terrain, supported the city’s growth as the largest urban centre north of the Haemus Mountains. The city’s gradual monumentalization and economic development, especially in the second century under the Antonines, suggest its significance within the Roman urbanisation programme. Nearby boundary stones mark the limits between Odessos and Thrace, further indicating Marcianopolis’ expanding influence within the region, although its territorial reach was likely modest, bordered by the territories of Odessus and Nicopolis ad Istrum [11,12,13].

2. Limestone Quarries of the Roman, Late Roman and Early Byzantine Cities of Nicopolis Ad Istrum and Marcianopolis

Three different types of rocks were the main materials used in Nicopolis ad Istrum and rural sites in its territory in the period from the first quarter of 2nd c. AD until the beginning of the 7th c. AD. These are limestone, sandstone and basalt ([14] (p. 268), [15] (p. 16)). Usually, scholars have not paid attention to their usage, and it has generally been accepted that Nicopolis ad Istrum had the limestone quarries near the village of Hotnitsa, situated 9.5 km southwest of the city, as the stone supply site ([16] (pp. 37–38), [17] (pp. 109–121), [18] (pp. 304–306), [19] (p. 22), [20] (pp. 179–181), [21] (p. 126, p. 129)) (Figure 1). However, a detailed view shows that sandstone and basalt were used exclusively for masonry ([14] (p. 268), [15] (p. 16–19)). Sandstone was used in the city centre and in almost all other sites of the rural territory ([15] (pp. 16–19), [22]). Basalt is the predominant local stone in the western part of the territory of Nicopolis ad Istrum, mainly in the area of the modern villages of Butovo, Varbovka, Nedan, Slomer and some others in the vicinity ([15] (pp. 16–19), [22]).

However, the material for the production of monolithic elements in Nicopolis ad Istrum and sites in its administrative territory was the organogenous detrital limestone, quarried near the villages of Samovodene, Hotnitsa and Koevtsi ([23] (pp. 75–88), [24] (pp. 203–210), [25] (pp. 197–203), [26] (pp. 79–106), [14] (p. 268), [15] (pp. 16–19), [22], [27] (pp. 127–130), [28] (pp. 121–138), [29] (pp. 139–141), [30] (pp. 476–477), [31] (p. 33, No 4) (Figure 1).

It is anticipated that these regional sources were utilised in the city and can be identified macroscopically. This study aims to establish a correlation between artefacts and their specific material sources. The diversity of the produced wares is as follows: all elements of the architectural order (bases, columns, capitals, cornices, friezes, architraves, pediments, ceilings, etc.), street pavements, water conduit elements, inscribed and uninscribed burial, votive and monuments from everyday life (stelae, sarcophagi, funerary and cult altars and monuments, measuring tables, mortaria) ([14] (p. 268), [15] (p. 18) (Figure 2 and Figure 3)). In single cases, this limestone was used for the production of votive plaques and 3D sculpture—a votive plaque from a settlement site near the quarries between Samovodene and Hotnitsa (Figure 3 (1)) and a piece of a sphinx relief from the military base of Dimum (Figure 3 (2)) ([32] (pp. 259–265), [14] (p. 276), [33] (p. 135, Nos 4–5, Figure 4)). Monumental statues, made of limestone, have not been found so far. In Late Antiquity, the organogenous detrital limestone was also used for masonry at sites situated close to the natural resources, sometimes as primary extraction, but almost in all other cases as spolia (Figure 3 (4,5)) ([14] (p. 268, pp. 277–278), [15] (p. 18), [34] (p. 38)).

The organogenous detrital limestone was used mainly in Nicopolis ad Istrum and its rural territory. The city centre was the main recipient. A collegium of stone cutters is known from there ([35] (No 674, No 690), [36] (220), [34] (p. 40), [31] (pp. 33–34)). They were settled during the foundation of Nicopolis ad Istrum from Asia Minor, and they brought their skills to Thrace. The second big site with supply from the stone quarries for organogenous detrital limestone was the military camp of Novae on the Danube River, where this is the second most used stone material for monolithic elements ([24] (p. 204), [25] (p. 199), [28] (pp. 121–126), [29] (pp. 139–142)). The territorial distribution of the limestone shows that it has been found at sites near the town of Levski to the west, the military base of Dimum to the northwest, Novae to the north, the village of Nova Varbovka to the east and the area of the village of Slaveykovo and the towns of Sevlievo and Lyaskovets to the south (Figure 4) ([14] (pp. 276–277), [15] (pp. 18–19), [27] (pp. 128–129), [28] (pp. 124–125), [29] (pp. 139–142)). It was used most probably in the military camp of Sostra to the southwest, but further analyses are needed in that direction (Figure 4).

The chronological distribution shows that the earliest inscriptions on this limestone are from AD 135/136 ([24] (p. 204), [25] (p. 199), [14] (p. 277), [15] (p. 18), [35] (No 601), [37] (No 72), [38] (Nos 357, 358, 386, 429), [39] (p. 32), [28] (pp. 121–126), [29] (pp. 139–141), [40]). It infers that its quarries began to operate as a result of its foundation ([14] (p. 277), [15] (p. 18)). The latest examples are from AD 270/271 ([15] (p. 18), [35] (No 734), [41] (No 5216)). AD 135/136–270/271 was the active period of stone extraction from the quarries ([15] (p. 18)). Most probably, they were exploited in the fourth and the first half of the fifth c. as well because Nicopolis ad Istrum had a massive rebuild [15] (8). Limestone use is noticed in some late antique fortresses, and the latest known examples so far of elaborated wares and spolia are known from Nicopolis ad Istrum (Figure 3 (3)) and the Early Byzantine city on the Tsarevets Hill in Veliko Tarnovo (Figure 3 (4,5)) ([14] (pp. 275–278)). The primary wares are church architectural elements from the sixth c. (Figure 3 (3)).

Located near Dewnya, the Golyamo Kesme quarry (or Marciana quarry)—the largest in the region—was a crucial source of high-quality limestone for monuments and architectural elements across Moesia, including cities on the Black Sea coast during the second and third centuries [42]. Situated about 1.5 kilometres northwest of Marcianopolis, Golyamo Kesme employed both vertical and horizontal stone extraction techniques, making it a site of particular interest. A 2013 project, funded by the Ministry of Education and Science and led by Zdravko Dimitrov from the National Archaeological Institute with Museum, Bulgarian Academy of Sciences, conducted extensive surveys of the quarry.

The quarry features a prominent vertical rock face, over 100 metres wide, from which stone was extracted in segmented sections. The vertical cutting process mirrored techniques used in Philippopolis, where wooden wedges were inserted high into the rock face. When wetted, the wedges expanded, cracking the rock from the top edge and enabling further extraction. This wedge-splitting method at Golyamo Kesme serves as a leading example of this technique in Bulgaria. Distinct saw marks from stonemasons are also visible, showing where the rock was further refined after splitting. In addition to vertical extraction, large blocks were removed horizontally from separate pits spread across the terrain, a method also seen at Nicopolis ad Istrum [42].

3. Locations and Geologic Positions of the Investigated Limestone Quarries

The investigated quarries are located in the northern foreland between the Haemus Mountain Range and the River Danube. In

Table 1. The coordinates of the locations of the investigated quarries in the Lower Danube area.

Marciana quarry	N43.25324-E27.57508
Koevtsi quarry	N43.15289-E25.11560
Samovodene I	N43.13617-E25.56536
Samovodene II	N43.13471-E25.57715
Samovodene III	N43.13424-E25.57837

, the coordinates of the visited locations are given; nevertheless, some other ancient sites where limestone was mined in antiquity may have occurred. Actually, the analytical data of some artefact samples indicate the existence of one or more of these so far unknown sites (see below). In all three sampled areas, traces of Roman quarrying activities are clearly visible and not hampered by modern mining activities.

Geologically, these quarries occur in the Fore-Balkan units, the external fold–thrust belt of the Alpine Orogen south of the Moesian Authochton, a microcontinent and promontory of the European platform (Figure 5). The sedimentation of these formations took place in Mesozoic and Cenozoic times, and the original deposition took place along the southern margin of the southernmost European continental block [2].

Preparing thin sections is particularly useful for limestones, as it allows for detailed petrographic analysis of their mineral composition, texture, and diagenetic features. Thin sections were prepared from both geological (FWM1124 and FWM1129) and archaeological samples (FWM1007 and FWM0916) to highlight their similarities. Due to the limited sample availability, only two archaeological samples were made into thin sections. Considering the macroscopic and microscopic similarities between these categories enabled a more comprehensive evaluation of the geochemical, geological, and archaeological data.

The big Marciana quarry north of Devnya most probably is the biggest Roman quarry in Thrace that was not disturbed by modern activities (Figure 6). Here, very well-preserved high quarry faces with Roman pick traces and big heaps of debris in front of the quarry can be studied. The limestones of the big Marciana quarry represent intraclastic organogenic and aphanitic limestones from Lower Cretacious (Hauterivium) strata. The fossil inventory consists largely of foraminifera and shell fragments indicating a shallow water environment (Figure 7). By visual examination, this limestone is of brownish/greyish colours. Occasionally, small vugs or cavities can be seen.

The Koevitsi and Samovodene limestones (Figure 8) also belong to the Lower Cretacious formation; however, in this case, their position is in Barremian–Aptian formations according to the geological map. These series are composed of limestones, sandstones, marls and sandy limestones.

The limestone exhibits a very fine micritic groundmass, dense and without vugs, with bioclasts of appreciable size. Under the microscope, the limestone is composed mainly of foraminifera and large to very large shell fragments. The organic assemblage testifies a shallow water environment. In Figure 9, a large, conical foraminifera can be seen.

4. The Applied Methods

For a detailed description of the analytical methods, applied references are made to Prochaska and Attanasio (2021) [3]. An abbreviated summary of the methods used is given below. When the data of the database published here are used for comparison, it is of pivotal importance to strictly follow the preparatory procedures and the analytical designs used for the analyses of the database samples. Furthermore, it has to be considered that, with the small sample size when sampling artefacts, problems with surface contamination and with the inhomogeneity may occur and utmost care has to be applied.

Stable isotope analysis: The analysis of the stable isotopes of O and C is a standard method in marble provenance analysis. For this investigation, the samples were analysed at the Bayerische Staatssammlung für Paläontologie, Munich, Germany. Carbon dioxide was extracted from carbonates by reaction with phosphoric acid at 72 °C using an automated online device (Finnigan Gasbench II). Oxygen and carbon isotope ratios were measured with a Finnigan DeltaplusXP mass spectrometer (Thermo Fisher Scientific, Waltham, USA) operated in a continuous He flow mode. A house standard was used as the running standard.

Trace element analysis: The chemical analyses of all samples taken during this investigation were performed by ICP-MS analysis after dissolving the carbonate phase using hot HNO₃. As this is a “near total analysis”, the trace element data obtained by this method must not be compared with data obtained by bulk analytical methods (e.g., HF dissolution, XRF analysis, etc.). Utmost care has to be exercised in sample preparation and handling in general, as many significant trace elements occur only in the sub-ppm range.

Quantitative analyses of trace elements were conducted using an iCAP Qc ICP-MS system (Thermo Fisher Scientific, Bremen, Germany) equipped with a concentric nebulizer and a quartz cyclonic spray chamber. Sample uptake was accomplished via an ESI SC-2-DX autosampler (ESI, Omaha, NE, USA) and an ESI FAST sample introduction system (1 mL sample loop) (ESI, Omaha, NE, USA). Signal quantification was based on external calibration with aqueous standard solutions in a general concentration range of 0.002 to 0.5 mg/L. Data acquisition was performed using Qtegra software provided by the instrument manufacturer.

For the ICP-MS analysis, a plasma power of 1550 W, a nebulizer gas flow of 1.0 L/min, a cool gas flow of 14 L/min argon and an auxiliary gas glow of 0.8 L/min argon were applied, and the dwell time was set to 10 ms. ⁷²Ge, ¹¹⁵In and ¹⁸⁵Re were used as internal standards.

Data evaluation: The large number of acquired data invariably required for the use of a statistical evaluation of the data. Therefore, a simultaneous evaluation of a large number of variables was performed through a multivariate discriminant analysis. In this work, the programme packages STATISTICA (version 7) and SPSS (version 11.0.) were used.

5. The Analytical Results and the Geochemical Characteristics of the Investigated Samples

The analytical results for the three different groups of limestones analysed for this work are discussed below. The analytical data of the analysed artefacts are given in the Supplementary Material in Table S1. Numerical results of the corresponding quarry samples for comparison are also presented here in the Supplementary Material in Table S2.

The results of the stable isotope investigations are graphically displayed in Figure 10. The corresponding data fields are shown as statistical 90% ellipses. Two samples from Marcianopolis (nos. 1 and 2) are displayed in blue, and the sample set from Nicopolis ad Istrum is separated into two groups representing the samples’ origination from the Samovodene quarries in red and those of unknown origin in grey. This separation is, in fact, an anticipation of the results of the discrimination analysis shown below. For the sake of clarity, this assignment is already shown in the isotope diagram, although this separation is only achieved by the multivariate analysis and the numerical calculation of the provenance as shown below (Figure 11,

Table 2. The numerical results of the multivariate discrimination analysis (for explanation, see text below). As for Marciana and Samovodene samples the relative probability is 100% there is no second choice (marked with --).

	Distance	Abs. Probab.	Rel. Probab.	Provenance	Rel. Probab.	Provenance
	1. choice				2. choice
1	0.98	61.2	100.0	Marciana	--	--
2	1.25	53.4	100.0	Marciana	--	--
3	3.88	14.3	100.0	Samovodene	--	--
4	3.09	21.4	100.0	Samovodene	--	--
5	2.91	23.3	100	Samovodene	--	--
6	1.93	38.2	100.0	Samovodene	--	--
7	2.88	23.7	100.0	Samovodene	--	--
8	0.44	80.1	100.0	Samovodene	--	--
9	3.70	15.8	100.0	Samovodene	--	--
10	10.51	0.5	99.9	unknown	--	--
11	43.0	0.0	100.0	unknown	--	--
12	53.1	0.0	99.9	unknown	--	--
13	8.73	1.3	100	unknown	--	--
14	8.14	1.7	100	unknown	--	--
15	13.74	0.1	55.2	unknown	--	--

).

In the isotope diagram, (Figure 10), it is clearly demonstrated that the three quarry areas under consideration can be clearly separated on account of the stable isotope numbers. The 90% ellipses are clearly separated, and the discrimination is 100%. The samples from the big Marciana quarry tend to have a lighter isotopic composition compared to the Koevtsi and Samovodene samples. It is remarkable that the latter two locations clearly can be separated on account of their isotopic composition, although the limestones of both locations are hosted in Barremian–Aptian formations.

To enhance the discrimination of the different groups of marble sources and to improve the separation of the data fields, further variables were included in the calculations. Seventeen main and trace elements were analysed and are given in the Supplementary Material (Table S1 for the artefacts and Table S2 for the quarry samples). The multivariate discrimination analysis computes a large number of analysed variables simultaneously and reduces them to a smaller number of artificial variables, so called factors. For this evaluation procedure, the variables δ¹⁸O ‰, δ¹³C ‰, Mg, Mn, Fe, Sr, Cr, Y, Ba, La, Ce, Pr and Yb were used.

As demonstrated in Figure 11, the data fields of the different types of quarry samples can be completely separated. As only two variables can be shown in a bivariate graph, this graphical display is a two-dimensional simplification of a multidimensional system, and it is only an approximation of a multidimensional reality.

The precise (multidimensional) conditions of the state of assignment of a sample to a quarry (or quarry sample collective) can only be expressed numerically after a multivariate statistical procedure. In this work, we used the programme package SPSS. The numerical results of the probabilities of the provenance calculations of each sample are presented in Table 2. The results of the probability calculations presented in Table 2 show that the relative probability of the corresponding samples is 100% already in the first choice, thus the second choice is nil for the artefacts from Marcianopolis and from a group of samples from Nikopolis ad Istrum. Another group of samples from Nikopolis ad Istrum (sample projections in grey in Figure 10 and Figure 11) also reveals Samovodene provenance although plotting outside the 90% Samovodene ellipse. This second group most probably is coming from a so far not-sampled location in the Smovodene area.

Distance: Distance of the sample under consideration from the centre of the ellipse that represents the quarry probability field. The central point of the ellipse expresses the average and, hence, most characteristic values of a quarry. The closer a point is to the centre of an ellipse, the more likely the provenance from that marble site.

Relative (posterior) probability: Probability of the sample belonging to some group within the assumption that it originates in any case from one of the groups in the selection. The threshold is 60%. Lower values indicate that the sample’s assignment is in doubt between two or more groups or does not belong to any of the chosen groups.

Absolute (typical) probability: This parameter indicates if the sample belongs to the chosen group or, in other words, is a typical representative of the group properties. The threshold is 10%, corresponding to samples on the edge of the 90% probability ellipse. Lower values indicate anomalous samples (outliers) or samples possibly not belonging to any group under consideration.

6. Discussion and Conclusions

As already mentioned initially, the different aims of this paper are as follows: one issue is to investigate if the standard methods for the provenance analysis of marbles can also be applied for pinpointing the origin of limestone used in antiquity. Another important aim is to start a database for these investigations for the area north of the Haemus Mountains in the Roman province of Thrace. When using the analytical data presented here for further extension of the database or comparing analyses from artefacts, it is essential for the sake of comparability to follow exactly the same analytical procedures applied here. This essential circumstance is discussed in more detail in Prochaska and Attanasio (2021) [3]. Data from sources using different analytical methods, especially bulk methods like total digestion decomposition or XRF analyses, will yield different results and cannot be computed with the data given here.

The obtained analytical results clearly show that the different sampled quarry sites clearly can be separated using the combination of the analytical procedures explained above and a subsequent statistical evaluation. As this investigation is only a very first and preliminary inventory of the limestone quarries in that region, it can be expected that more regional limestone sources will be made publicly known in the future.

Another aim of the present investigation was to pinpoint the source of the artefacts made of limestone we sampled during the FWF project P 33042. The vast majority of approx. The 180 artefacts sampled in Marcianopolis and Nicopolis ad Istrum within the course of the FWF project P 33042 were made of marble. Among these, two samples made from limestone from Marcianopolis were investigated and were attributed to the Marciana quarry. The 13 limestone artefacts from Nicopolis ad Istrum 7 can be ascribed to the Samovodene quarries with very good probability results. Six limestone samples from Nicopolis ad Istrum, however, are clearly outside of the considered quarry data fields. Nevertheless, as the best fit is the Samovodene data field (although with very low probability), it can be assumed that other, so far not investigated sources occur in close vicinity.

Investigations of numerous artefacts from Stara Zagora, located on the southern flanks of the Haemus Mountains, have revealed a significant number of objects made from limestone. Notably, the material does not align with previously established limestone sources as presented in this paper. Although relevant quarries exist in the area, there are currently no available data from these locations to confirm their connection to the artefacts. As such, further investigations into these quarries will be necessary to establish any provenance links in future research.