3.1.) Regional context

3.1.1) Geographic and geological context.

The city of Santarém is located in the southwest of the Iberian Peninsula, in the Tagus River Valley (Fig 1A). Its proximity to the river was strategically advantageous for trade in the past while the fertile alluvial plain would have been good for food production. The hilltop position of the settlement, elevated approximately 90m above the river, made it easier to defend during its turbulent history [66]. Santarém is situated on the Middle sector of the Lower Tagus Cenozoic Basin and is covered by Miocene and Pleiocene sedimentary deposits followed by Pleistocene fluvial terraces (Fig 2). The Pleiocene plateau on which the city rests contains units of limestone, clay and marl (P2) and layered sandstone and claystone (P1) [67]. Within a few kilometres, there are Miocene outcrops of limestones, sandstones and clays (MP, M4) and modern deposits (a) on the alluvial plain west of the river and Pleistocene alluvial terraces (Q3) [67, 68]. Plant samples (collected and measured by [61]) from within 15Km of Santarém yielded ⁸⁷Sr/⁸⁶Sr ratios of 0.712 and 0.710 where the geological substrate is Miocene calcareous material and sandstone, and a ⁸⁷Sr/⁸⁶Sr ratio of 0.715 where the substrate is Pleistocene gravels, sands and limestone [61].

Download:

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

Fig 1.

Map of a.) Location of Santarém and b.) Excavated area of Avenida 5 de Outubro. Map in Fig 1A). drawn by MacRoberts RA and adapted from Wikimedia Commons (https://commons.wikimedia.org/wiki/File:Portugal-CIA_WFB_Map.png), CC public domain. Map in Fig 1B.) taken from openstreetmap.org licensed under the Open Data Commons Open Database Licence (ODbL).

https://doi.org/10.1371/journal.pone.0299958.g001

Download:

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

Fig 2. Geological map of Santarém and sampled sites for bioavailable ⁸⁷Sr/⁸⁶Sr baseline.

Map drawn by MacRoberts RA and adapted from the LNEG (Laboratório Nacional de Energia e Geologia) 1: 50 000 geological raster maps “31-A” and “31-C”. (https://geoportal.lneg.pt/pt/dados_abertos/cartografia_geologica/cgp50k/) [67]. Both geological raster maps 31-A and 31-C are published under CC By 4.0. (31-A: https://sig.lneg.pt/metadados/catalog/search/resource/details.page?uuid=831fb1df-e743-4e23-97d0-91af58d46f02; 31-C: https://sig.lneg.pt/metadados/catalog/search/resource/details.page?uuid=20486b7f-20be-4506-aa88-5caacc5d6946).

https://doi.org/10.1371/journal.pone.0299958.g002

3.2.) Historical context

In 711 AD, the Islamic conquest of Iberia began, led by the General Tariq Ibn Ziyad. The Berber armies, under the rule of the Damascus-based Umayyad dynasty and united by their Islamic faith, crossed from Tangiers into Gibraltar. What may have begun as raids rapidly escalated into a full-on conquest of the peninsula [76, 77]. After defeating the Visigoth king, Roderic, the Muslim armies, reinforced by Arabs under the command of Musa ibn Nusayr, met little resistance and swiftly conquered Córdoba, Toledo (the Visigothic capital) and Mérida in 713 AD. By 716 AD, the cities of Évora, Santarém and Coimbra had fallen. Within just five years, Muslim forces had gained control of the entire Iberian Peninsula, with the exception of Asturias, a mountainous region in the northwest that remained under Christian control [78]. Favoring the drier climate south of the Tagus river, the Islamic dominion over the region would continue until the 12^th century in most of Iberia and as late as 1492 AD in Granada.

The newly settled population consisted of mainly North African Berbers with Arabs making up the higher social status [76]. The Umayyad dynasty was primarily concerned with military expansion and the conquering of territories rather than proselytisation. It did not favor the conversion of non-Muslims, and Christians and Jews living in “Shantarîn” and other cities would likely have been allowed to keep their religion and property with the protected legal status of “dhimmi”, on the condition that they paid a yearly tax or “jizya” [79]. In fact, conversion was at times actively discouraged by the Umayyads as it would impact state revenue, since Muslim landowners paid a lesser tax or “ushr” [78, 80]. In the mid 8^th century, following Berber Revolts and civil unrest, the Umayyad Caliphate was brought down by the Abbasids who shifted the main seat of control from Damascus to Baghdad [81]. Most of the Umayyad’s were slaughtered in the coup, with the exception of Abd al-Rahman, who fled to Spain and founded a new branch of the Umayyad Dynasty, the “Emirate of Córdoba” in 756 AD which continued to rule Iberia independently for a further three centuries [76]. In the 10^th century, control of al-Andalus was divided amongst the taifa kingdoms (the most important being Badajoz, Toledo, Zaragoza and Seville), followed by the respective rules of the Berber Almoravids and the Almohads [78, 80]. This fragmentation of Muslim control in the 11^th and 12^th centuries paved the way for the Christian conquest and eventual establishment of the Portuguese Kingdom. Santarém, an important cultural centre under Islamic rule, was captured by King Afonso Henriques and the Christian forces on 15 March 1147 [82].

During the period of Islamic rule, al-Andalus likely enjoyed trade and commercial relationships with other Islamic regions [80], facilitating a cultural expansion as much as a military one. Technological innovations brought by the Arabs to the Iberian Peninsula included new irrigation methods, experienced as they were with conserving and distributing water in the Middle East. With new irrigation techniques came the potential for growing crops that required an adequate water supply, hence the Muslims introduced cotton, rice, brinjal, artichoke, apricot, banana, date palm, spinach, coconut, orange and lemon, and this legacy is evident in some of the names of these crops in Portuguese and Spanish that derive from Arabic [80, 83]. Also, among these introduced crops, are two very important C₄ plants: sorghum and sugarcane. Sorghum was cultivated in the Niger Delta and Nile Valley, spread to the Arabian Peninsula and Asia [84], and was later brought to Europe with the Islamic conquest. It is highly drought-tolerant and requires at least 35 cm of rain per year, slightly more than millet. It is still commonly eaten in Africa in the form of bread, boiled like rice or as a porridge. Sugarcane is a tropical grass that was cultivated in southeast Asia around 9000 BC, spread to India and then Persia along the Silk Road around 500 AD [85]. In the 7^th and 8^th centuries, Muslim traders brought sugar to North Africa and southern Iberia, with the most refined quality being produced in high quantities in Egypt. Its poor nutritional value, however, made it a delicacy but not a staple crop.

Millet, a hardy C₄ plant originally cultivated in Africa (pearl millet and finger millet) and Asia (proso millet and foxtail millet), grows well in infertile, sandy soil and requires as little as 25 cm of rain per year. It was grown by Europeans by the third millennium BC but became widespread around 1000 BC [85] and thus is was the only prevalent C₄ plant to be consumed in Europe before the Muslim conquest. In the Middle Ages, it was commonly consumed in porridge and flatbread, and is still a major food source in modern African and Asian populations.

Concerning medieval Islamic burials, tradition dictated that the head of the deceased was turned to the right and pointed towards the Kaaba in Mecca, as instructed by the hadiths, a direction known as the qibla [86]. However, it has been noted that the orientation of the qibla varied across regions and the way it was sought differed greatly [87]. No mathematical solution existed for identifying this direction prior to the 8^th century, and in some ancient Muslim traditions Jerusalem or Petra may have been chosen for the direction of prayer, rather than Mecca [87–89]. Although this may mean that many ancient Muslim burials may not have been in the ‘correct’ canonical position, the most important thing for the believer would be the intention of facing the qibla, even if it is not geographically correct in the end.

At a shorter timescale, it has been observed that variations in the orientation of graves, by as much as 35°, could have occurred as a result of seasonal differences in the angle of the sun because, in the absence of a permanent indicator, grave diggers could orient graves using the rising or setting sun [90].

3.3.) Archaeological context

3.3.1.) Avenida 5 de outubro.

During the 8^th & 9^th centuries, the inhabited area of Santarém would most likely have been concentrated in the alcáçova (castle), located on the highest and most central part of the hilltop settlement. A small neighborhood was located to the northeast and outside of the walled area (Fig 1b), but was later incorporated into the city as the population grew and the defensive line was extended North into the Marvila Hill by the 11^th/12^th century [91]. The site of Avenida 5 de Outubro #2–8 (WGS84: 39.235627; -8.679350), in Alporão, Santarém, was excavated during 2007 and 2008, uncovering a necropolis that contained evidence of use during the Roman, Visigothic and Islamic Periods. 58 adult and 32 non-adult skeletons were recovered from the Islamic necropolis. The graves of interest to this study consisted of simple burial pits, with an absence of grave goods and the individuals placed on their right side, and thus appeared to be consistent with medieval Islamic burial rites whereby the dead are placed in a shroud, unaccompanied by ornamentation and laid on their sides in a simple grave [86]. Some of the graves in the northern part of the excavated area contained skeletons that were oriented S/N, a slight deviation from the orthodox position whereby they should be aligned to the qibla, while skeletons in graves in the southern and south-eastern area are all oriented SW/NE and canonically aligned to the qibla (head to the SW, body on the right side and face turned towards Mecca i.e. facing a south-eastern direction) [91] (Fig 3).

Download:

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

Fig 3. Image of intersecting graves 955 and 956 with differing orientations.

Republished from [91] under a CC BY license, with permission from Medievalista, original copyright 2012.

https://doi.org/10.1371/journal.pone.0299958.g003

Originally interpreted as related to chronology, the misaligned (S/N-oriented) burials were thought to be a consequence of early conquest settlers (themselves probably newly-Islamised Berbers) converting existing Christian temples into mosques and incorrectly estimating the direction of the qibla [91]. In the late 10^th century, under the initiative of Caliph Hisham II, the aljama mosque was constructed in Santarém [92], at which point the mihrāb would have been corrected (a niche in the mosque wall that indicates the direction to pray) and it was initially supposed that burial orientations would have been standardised accordingly i.e. SW/NE-oriented graves in this necropolis [93]. A recent reappraisal of the stratigraphy data (exceeding 2500 units) has led to reconsideration of the chronological relationships and revealed that some of the misaligned burials might in fact be more recent than the canonical ones, such as the S/N-oriented grave of skeleton 955 which overlaps the canonical grave of skeleton 956 (Fig 3). An ongoing investigation by a separate research team at the University of Coimbra will provide radiocarbon dates adding further information on the chronology of the site.

Taking into account the medieval Islamic preference for burying the dead outside the cities of the living [86, 93], and archaeological evidence of the expansion of the city over time, Liberato and Santos (2017) [93] have hypothesised that the necropolis of Avenida 5 de Outubro was used between the 8^th and 10^th century by some of the earliest Muslim inhabitants of Santarém following the conquest (and the earliest to bury their dead in the necropolis). The presence of large negative structures, probably silos, dating to the first taifas/Almoravid period indicate that by the early 11^th century, this space was no longer primarily used as a necropolis [91]. By this point, another necropolis had been established further west on the plateau, a probable result of the growing population and extended defensive line of Santarém.

4.1.) Human skeletons

Approval for collecting, sampling and destructive analysis, was granted by the Câmara Municipal of Santarém, Portugal, where the skeletal specimens under study are curated. According to Portuguese laws and regulations concerning archaeological human remains, no permits were required for the described study. Prior to selection for chemical analysis, the human skeletons from Avenida 5 de Outubro were subject to anthropological assessment of age at death and sex. Sex estimation was determined by evaluating both pelvic and cranial features, following methods established based on Portuguese reference collections [94] and recommended standards [95]. The age at death of adult skeletal remains examined in this research were established using degeneration of the pubic symphysis [96], degeneration of the auricular surface [97], and recommended standards [95]. The results of age at death estimations were grouped into three age categories: young adults ranging from 18 to 29 years, mature adults between 30 and 50 years, and older adults over 50 years of age.

Following this assessment, bones from 58 adult skeletons (25F, 20M, 13U) were selected for δ¹³C, δ¹⁵N and δ³⁴S analysis. Long bones were preferentially selected because of their long turnover rates, incorporating dietary input over a period of more than a decade [98] and high volume of compact bone which facilitates good collagen preservation [99]. In the case of HS-811, no long bones were available and a rib was sampled instead, which may represent a different period of life prior to the death of the individual [98]. Wherever possible, broken bones were chosen for sampling in order to minimise damage to archaeological material. Of the bones selected for δ¹³C and δ¹⁵N analysis of the bone collagen, 45 were also sampled for δ¹³C analysis in the bone apatite, so that whole diet carbon contributions could be considered alongside the protein dietary inputs. Although 25 skeletons did not have teeth preserved, either due to ante-mortem loss or taphonomic processes, tooth enamel from 22 individuals was sampled in order to measure δ¹⁸O and ⁸⁷Sr/⁸⁶Sr for the reconstruction of mobility patterns, as well as δ¹³C in order to assess whole diet inputs during childhood. Loose molars were preferentially selected to avoid damaging well preserved jaws wherever possible. First molars (M1s) were preferred for their potential to yield birthplace ⁸⁷Sr/⁸⁶Sr signatures [100], but when they were not available, M2s or M3s were sampled. The enamel of tooth crown forms between the ages of birth—~2 y in the M1, ~18 mths—~5 y in the PM1, ~3 - ~6.5y in the PM2, ~3.5 - ~6.5y in the M2 and ~9.5 – 12y in the M3 [101]. In order to account for possible oxygen isotope enrichment (higher δ¹⁸O_DW values) in the first molar (M1) tooth enamel as a result of breastfeeding, which may occur as a result of fractionation in the body water of the mother [41], wherever possible a later formed tooth (M2/M3) was sampled in addition to the M1s. Of these 22, 9 individuals were selected for the analysis of two teeth each for δ¹⁸O and δ¹³C, with the aim of checking for differences in teeth formed earlier vs later (as per availability) that could be introduced by a breastfeeding signal, or by changes in diet or location during childhood. Six individuals were also selected for ancient DNA analysis, three of which had maxillofacial prognathism, broad nasal apertures, smooth inferior nasal margins, retracted chins and rectangular occipital orbits, features that are more prevalent in pre-modern populations with African genetic ancestry than Eurasians [102, 103], and thus their genetic lineage was of particular interest in this context. Ancient DNA molecules were extracted from the petrous bones of these six skeletons. All materials are listed in Table 1.

Download:

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

Table 1. Human skeletal material.

https://doi.org/10.1371/journal.pone.0299958.t001

4.1.1.) Grave orientation.

One of the most intriguing features of the burials from the necropolis of Avenida 5 de Outubro was the differing orientation of their graves and in some cases the direct overlap of graves with varying orientations. Overlapping of burials was a relatively rare occurrence in medieval Islamic burials with the exception of very densely populated cities, such as Cordoba in the 10^th century [104]. This feature raised questions over whether there may be any meaningful difference, culturally or chronologically [91], in the way the graves were oriented. The individuals were thus assigned a classification according to their burial orientation. Three categories were thus defined: 1.) unorthodox graves that were clearly S/N oriented containing skeletons with the head to the S and feet to the N, 2.) graves that are not perfectly S/N oriented but still deviate from the canonically ‘correct’ position and contain skeletons that are oriented closer to S/N oriented than to SW/NE oriented, and 3.) canonical graves that were clearly SW/NE oriented containing skeletons with the head to the SW and feet to the NE. Fig 4 illustrates how these classifications were defined. For the ease of referring to these orientation types, these groupings are referred to in the results and discussion as: ‘Group 1’(S/N-oriented), ‘Group 2’ (Maybe S/N-oriented) and ‘Group 3’ (SW/NE-oriented). Fig 5 shows the location of the graves in the necropolis and the number of individuals in each group.

Download:

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

Fig 4. Grave orientation classification, diagram showing the direction that skeletons would face in right lateral position with faces turned towards the east or southeast, and the direction from Santarém towards Mecca.

Drawn by MacRoberts RA. The map image inset was obtained from Wikimedia Commons (https://commons.wikimedia.org/wiki/File:North_Africa_location_map.svg), CC public domain. The skeletons were adapted from [91] with permission from Medievalista, original copyright 2012, republished under a CC BY license.

https://doi.org/10.1371/journal.pone.0299958.g004

Download:

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

Fig 5. Map of Avenida 5 de Outubro necropolis showing position of graves and Bar graph showing # of individuals per group (inset).

https://doi.org/10.1371/journal.pone.0299958.g005

4.2.) Fauna

12 Faunal bones from Vila Rosa Palace, Santarém were selected for δ¹³C and δ¹⁵N analysis, representing both domestic and wild species (4 cows; 4 sheep/goats; 1 rabbit; 2 chickens; 1 fox). Taxa identification and anatomical diagnosis was conducted, prior to chemical analysis, at the Laboratório de Arqueologia at Universidade do Algarve [105]. Faunal material is listed in Table 2.

Download:

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

Table 2. Faunal skeletal material.

https://doi.org/10.1371/journal.pone.0299958.t002

4.3.) Plants

Eight modern plant samples were collected from the various lithologic outcrops present within a 15 km radius of Santarém, in order to create a local bioavailable ⁸⁷Sr/⁸⁶Sr database against which the archaeological human ⁸⁷Sr/⁸⁶Sr data could be compared. Wild plants were selected from locations where soils were not expected to be affected by agricultural activity and/or fertilizers. Samples were collected from public lands not included within the Natura 2000 protected areas network. The location of each plant sample was recorded by GPS and is indicated in Fig 2. Plant material is listed in Table 3.

Download:

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

Table 3. Plant material.

https://doi.org/10.1371/journal.pone.0299958.t003

5.1.) Collagen extraction

Approximately 500 mg of compact bone was cut with a DREMEL® rotary tool fitted with a diamond-coated disc. Pieces were then cleaned with a diamond-coated burr to remove dirt, discoloration, potential contaminants, and spongy bone. A modified Longin method [106, 107] was used to extract the bone collagen, whereby samples were each placed in 10 ml of 0.5M HCl at room temperature with overnight refrigeration, for around 14 days until demineralised. Acid was replaced after one week and samples were vortexed regularly. Once fully demineralised, acid was removed and samples were rinsed to neutrality with Milli-Q® ultrapure water. Bone samples were then placed in 10 ml of 0.125 M NaOH for 20 h at room temperature to remove organic contaminants such as humic and fulvic acids [108]. Samples rinsed to neutrality again and then placed in 10ml of 0.01 M HCl for 48 h at 70°C to facilitate gelatinisation. The solution was filtered using a 9 ml Ezee-Filter™ (60-90μm) (Elkay Laboratory Products) to remove insoluble particles. Collagen was then freeze-dried and the dry collagen was weighed to calculate collagen yield.

5.2.) Stable carbon and nitrogen isotope analysis

Approximately 0.7 mg of collagen was weighed into tin capsules for δ¹³C and δ¹⁵N analysis, conducted using an Elemental Analyser (EA Flash 2000 HT, Thermo Fisher Scientific®) coupled, via a ConFlo IV interface (Thermo Fisher Scientific®), to an isotope ratio mass spectrometer (Delta V Advantage-IRMS, Thermo Fisher Scientific®). All sample preparation and analysis of δ¹³C and δ¹⁵N was conducted at the HERCULES Laboratory, University of Évora. Isotope results are reported as δ values, given in parts per thousand (‰) relative to the international standards VPDB (Vienna Pee-Dee Belemnite) for carbon and AIR (Ambient Inhalable Reservoir) for nitrogen. A three point calibration [109] was used for δ¹³C composition, IAEA-CH-6 (sucrose, δ¹³C: -10.45‰), EEZ-20, (δ¹³C: -12.3‰) and IAEA-600 (caffeine, δ¹³C: -27.77‰), while for δ¹⁵N composition, IAEA-600 (caffeine, δ¹⁵N: +1‰), IAEA-N-1 (ammonium sulphate, δ¹⁵N: +0.43‰) and IAEA-N-2 (ammonium sulphate, δ¹⁵N: +20.3‰) were used for calibration. L-Alanine (δ¹³C: -18.39‰; δ¹⁵N: +0.91‰) was used as an in-house check standard interspersed during each run. Measurement uncertainty was calculated according to the method presented by Szpak et al. (2017) and using the Microsoft Excel spreadsheet templates provided in that paper [109] (See S1 Table). Standard uncertainty (u_c) is estimated as a pooled standard deviation of precision (u(R_w) ±1σ) while accuracy (u(bias) ±1σ) is calculated using the variability in calibration standards and check standards. Precision (u(R_w)) was determined to be ± 0.12‰ for δ¹³C and ±0.09‰ for δ¹⁵N. Accuracy (u(bias)) was determined to be ± 0.17‰ for δ¹³C and ± 0.23‰ for δ¹⁵N and the total uncertainty over 7 analytical sessions was estimated to be ± 0.21‰ for δ¹³C and ± 0.25‰ for δ¹⁵N.

5.3.) Stable sulphur isotope analysis

Sulphur isotope analysis was conducted at SIIAF, University of Lisbon, using an IsoPrime mass spectrometer. Approximately 6mg of collagen was placed into capsules with additional 1mg V₂O₅ and combusted with a pulse of oxygen. δ³⁴S was calibrated using the inorganic international standards NBS 127 (+20.3‰), IAEA S1 (-0.3‰), and B2155 Protein (+6.32‰) as an internal standard. Precision (u(R_w)) was determined to be ± 0.1‰, while accuracy (u(bias)) was ± 0.96‰, providing a total uncertainty of ± 0.97‰ for δ³⁴S over 2 analytical sessions (see S1 Table). Stable sulphur isotope results are reported as δ values, given in parts per thousand (‰) relative to VCDT (Vienna- Canyon Diablo Troilite).

5.4.) Stable carbon and oxygen isotope analysis

Prior to sampling, the preservation of bone apatite was assessed using ATR-FTIR, to ensure suitability for δ¹³C and δ¹⁸O analysis (See S2 Text for methodology). Subsequently, the outer surface of the tooth enamel and bone samples were cleaned with a DREMEL® rotary tool fitted with a diamond coated burr to remove soil particles and contaminants. Approximately 15 mg of bulk tooth crown enamel/bone powder was collected and ground with an agate mortar and pestle to obtain a homogenous sample material. The powder was bleached in 1 ml NaOCl (2–3 vol%) for 24h at room temperature to remove organic matter and contaminants. Samples were then rinsed to neutrality with Milli-Q® ultrapure water and placed in 1 ml of 1M acetic acid buffered with Li-acetate solution for 12 h (for tooth enamel) or 24 h (for bone powder) to remove any adsorbed or diagenetic carbonates [110–112]. Samples were rinsed again to neutrality and dried overnight at 70°C. Subsequently, the stable oxygen and δ¹³C data were determined for the carbonate fraction of the samples. For that purpose, approx. 800–850 μg was dissolved in He-flushed borosilicate exetainers at 72°C using a water-free phosphoric acid. The released CO₂ gas was then measured in a Continuous Flow- Isotope Ratio Mass Spectrometer (Thermo Fisher Scientific® 253) coupled to a GasBench II at the Institute of Geosciences, University of Mainz. δ¹³C and δ¹⁸O were calibrated with a 1-point calibration using the Carrara Marble (δ¹³C: +2.01‰; δ¹⁸O: -1.91‰) because the isotope data of the samples was very close to that of the Carrara Marble. Results are given in parts per thousand (‰) relative to VPDB. The average precision error (1σ) was found to better than 0.04‰ for δ¹³C and 0.04‰ for δ¹⁸O and the accuracy was within 0.01‰ for both isotope systems.

The δ¹⁸O values (relative to VPDB) were first converted to the VSMOW-scale, using the equation (see Eq 1) established by Coplen et al. [113]: (1)

The conversion of carbonate to phosphate oxygen isotope values (δ¹⁸O_c to δ¹⁸O_p) was done using the equation (see Eq 2) proposed by Chenery et al. [44] while the estimated drinking water oxygen isotope values (δ¹⁸O_DW) were calculated using the equation (see Eq 3) established by Chenery et al. [44] which in turn was based on the drinking water equation of Daux et al. [43].

(2)

(3)

5.5.) Radiogenic strontium isotope analysis

For human tooth samples, the enamel surface was cleaned with a DREMEL® rotary tool fitted with a diamond coated burr and around 20 mg of bulk enamel powder was collected from the tooth crown, with care taken to avoid any sand or dirt particles. Enamel was ground with an agate mortar and pestle, then placed into 1.2 ml acetic acid buffered with Li-acetate solution for 12h. Sampled were then rinsed to a neutral pH with Milli-Q® ultrapure water and dried overnight at 70°C. Plant samples were rinsed with Milli-Q® ultrapure water to remove any exogenous dust or contamination, dried at 50°C for 48h, placed into quartz crucibles and ashed for 12h at 550°C in a muffle furnace. Approximately 20 mg of plant ash was collected and placed in microtubes. Sample digestion and strontium extraction for tooth enamel powder and plant ash samples was conducted at the Isotope Geology Laboratory of the University of Aveiro. All samples were digested using HNO₃ acid, and in plant samples an additional step of digesting with H₂O₂ was used. After evaporation to dryness the samples were dissolved in HCL 6.2M and dried again. All reagents used were sub-boiling distilled, and the water produced by a Milli-Q Element (Millipore) apparatus. Columns filled with Sr-Resin (Sr-B10-S, 50-100mesh, TrisKem International) were used for purified strontium from the remaining elements. Then the extracted Sr was loaded onto Ta filaments with H₃PO₄ and the isotopic composition measured with a VG Sector 54 TIMS in dynamic mode with ⁸⁸Sr peak measurements at 1–2 V. Typical runs consisted of acquisition of 60 isotopic ratios and the analytical data were corrected for mass fractionation using the exponential law relative to ⁸⁸Sr/⁸⁶Sr = 0.1194. The standard reference material NIST SRM 1400 (bone ash) was concomitantly measured and provided a value of ⁸⁷Sr/⁸⁶Sr = 0.713160 ± 0.000021, within the GEOREM [114] database range of 0.7131–0.7134. In addition, the mass spectrometry standard SRM 987 gave, in the two periods of experimental work, mean values of ⁸⁷Sr/⁸⁶Sr = 0.710240 ± 31 and 0.710266 ± 11, (conf. limit 95%, N = 13 and 12, respectively).

5.6.) Ancient DNA analysis

Ancient DNA (aDNA) molecules were extracted from the petrous bone using a standard method suited for degraded fragments [115] and partially UDG-treated double-stranded DNA libraries [116] were generated. DNA extraction and sequencing library preparation steps were performed at the Australian Centre for Ancient DNA (ACAD)’s ultra-clean laboratory facilities, while post-amplification experiments were completed in standard molecular biology laboratories at the University of Adelaide following rigorous laboratorial procedures to minimise contamination and ensure high standards of quality for the genetic data. Shotgun sequencing was performed on an Illumina NovaSeq instrument at the Kinghorn Centre for Clinical Genomics (Sydney, NSW, Australia). A detailed methodology is provided in S1 Text.

Genetic sex determination was assigned using X- and Y-ratios, Rx ratios [117] and X chromosome read dosage (Mx) [118] (see S1 Text). A relatively low number of authentic ancient DNA fragments was initially obtained and only 3 libraries passed quality thresholds to undergo re-amplification and enrichment for DNA sequences overlapping 46k informative SNPs located on the Y chromosome and the whole mitochondrial genome using the myBaits Expert Human Affinities Prime Plus Kit by Daicel Arbor Biosciences (Ann Harbor, MI, USA). The enriched libraries were sent to the Kinghorn Centre for Clinical Genomics (Sydney, NSW, Australia) for sequencing on an Illumination NovaSeq instrument. Mitochondrial haplogroups were called using HAPLOFIND [119]. To obtain Y-chromosome haplotypes for the male individual, a VCF file selecting only the Y chromosome sites was input into yhaplo v1.1.2. [120].

5.7.) Reconstruction of solar position

In order to test some theories regarding the position of the sun in different seasons and how this may affect the ability of the gravediggers to estimate of the direction of Mecca, the online model SunCalc [121] was employed. Various times of the year including 1 January (for midwinter) and 20 June (for midsummer) were entered into the model, and the position of the sun at both sunrise and sunset from the location of Santarém was observed.

5.8.) Statistical analysis and data management

The statistical comparison of isotope results, to assess groups within the population, was conducted using the online statistics calculator “Statistics Kingdom, 2017” [122]. Non-parametric Kruskal-Wallis tests (when there were three groups i.e. Grave Orientation) and Mann-Whitney u tests (when there were two groups i.e. Sex) were used for non-normally distributed data to assess the mean ranks between the groups, whereby the shape of distribution within each group is assumed to be equal. In the cases where a significant difference between groups was found, a Post-Hoc Dunn’s test with a Bonferroni correction was used to identify which pairs were significantly different. A Levene’s test was used to assess if there was a statistically significant difference of variability between the groups. These results are provided in S3 Text. All isotope data from this study has been made available on the IsoArcH database [123, 124], DOI 10.48530/isoarch.2023.007 and is accessible via the URL https://isoarch.eu/datasets/2023-007/. This dataset is available under the Creative Commons BY-NC-SA 4.0 license.

6.1.) Bone collagen and apatite preservation

The bone collagen samples presented here (Table 4) had C:N ratios within the acceptable range [125, 126] for good preservation (humans: 3.1 ± 0.1; fauna: 3.1 ± 0.0) and overall had good collagen yields (humans: 10 ± 3%; fauna: 9 ± 4%), with the exception of HS-1979, HS-918, HS-923, HS-958, HS-1027 and HS-1061 which yielded no collagen. The carbon (humans: 39.2 ± 9.4%; fauna: 39.8 ± 2.7%) and nitrogen (humans: 14.8 ± 3.5%; fauna: 14.9 ± 1.0%) amounts indicated good collagen preservation [127], with the exception of one faunal sample (FVRP6- chicken) and the humans HS-819, 955-H, HS-2079, HS-2215, HS-2275, and HS-2316 who were excluded from further analysis and interpretation. The samples HS-2003 and FVRP12 were excluded from comparison due to instrument error in their analysis. This left a total number of 45 human samples and 10 faunal samples with comparable bone collagen results. Regarding the sulphur, modern bones typically have %S values that range from 0.15 to 0.35% as well as C:S and N:S ratios that range from 300 to 900 and 100 to 300 respectively [128, 129]. With the exception of the previously mentioned human samples and HS-2170, and the faunal samples FVRP6 and FVRP12, which were excluded from further comparison, the bone collagen sample values fell within these ranges and were considered to be reliable. Additionally, the preservation of bone apatite for a selection of individuals was assessed using ATR-FTIR. The methodology and results are provided in S2 Text. These results indicated that the bones were not significantly altered by diagenetic processes and thus any δ¹³C_ap values obtained from them are likely to reflect in vivo carbon in the bone apatite.

Download:

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

Table 4. Results-Bones.

https://doi.org/10.1371/journal.pone.0299958.t004

6.2.) Stable carbon (bone collagen and apatite) and nitrogen isotope results

The analysis of δ¹³C and δ¹⁵N in the human bone collagen (n = 45) resulted in a mean δ¹³C_col value of -19.0 ± 0.5‰ (ranging from -19.9‰ to -17.0‰) and a mean δ¹⁵N value of 10.4 ± 1.0‰ (ranging from 8.4‰ to 12.4‰). Regarding the faunal bone collagen isotope values, δ¹³C_col ranged from -20.9‰ to -16.9‰ and δ¹⁵N values ranged from 4.6‰ to 9.1‰ (Table 4). The animals have δ¹³C_col values that are consistent with a C₃ plant-based diet, with the exception of one chicken (FVRP1) which may have been partially fed on C₄ grain (i.e., millet), and the fox (FVRP5) which may have consumed food from various wild sources as well as human food scraps. As Fig 6 illustrates, the human δ¹³C_col values appear to be quite homogenous, with only one clear outlier (HS-921), and overall this population falls within the expected trophic level increase of 0–2‰ in δ¹³C_col [130] over the domestic terrestrial herbivores (Bos and ovicaprids). Concerning the δ¹⁵N values of the fauna, the highest belong to the fox (FVRP5) and chicken (FVRP1), both omnivorous species that are trophically more similar to humans. The Bos (n = 3) δ¹⁵N values are quite homogenous, while the ovicaprids (n = 4) are more diverse, possibly reflective of the more varied grazing patterns of sheep and goats compared to cows. Applying an estimated trophic offset of 3–5‰ [130] to the mean δ¹⁵N values of the domestic herbivores, it appears that the main protein components in the diet of this population likely came from local, C₃ plant-fed, domestic herbivores.

Download:

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

Fig 6. Human and fauna δ¹⁵N and δ¹³C_col.

Squares represent expected range of +1 trophic level (+0–2‰ in δ¹³C_col; +3–5‰ in δ¹⁵N) increase over mean Bos and Ovicaprid values, respectively.

https://doi.org/10.1371/journal.pone.0299958.g006

The stable carbon isotopes measured in the human bone apatite (δ¹³C_ap) yielded a mean value of -11.9 ± 0.7‰, ranging from -13.4‰ to -9.9‰. Comparing δ¹³C_col and δ¹³C_ap values for the bones (Fig 7), the overall homogeneity of diet for this population is still apparent. When plotted against a model developed by [131] to evaluate both protein and energy (whole diet) inputs, the δ¹³C_col values are consistent with C₃ based protein inputs for the population (except HS-921 who appeared to have a higher C₄ input) while the δ¹³C_ap values indicate that there was a dominance of C₃ with some mixing of C₄ based energy sources in the diets of these individuals.

Download:

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

Fig 7. Human δ¹³C_col and δ¹³C_ap plotted against regression lines [131] indicating main protein and energy sources in diet.

https://doi.org/10.1371/journal.pone.0299958.g007

6.3.) Stable sulphur isotope results

The analysis of δ³⁴S in the bone collagen of the Santarém humans (n = 46) yielded an average δ³⁴S value of 10.6 ± 2.1‰, ranging from 4.9‰ to 16‰. The faunal bone collagen yielded an average δ³⁴S value of 12.7 ± 2.4‰, ranging from 9‰ to 16.3‰. The humans have similar δ³⁴S values to the fauna with the exception of four individuals with lower δ³⁴S values (HS-2423; HS-1103; HS-1491 and HS-2181) (see Fig 8).

Download:

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

Fig 8. Human and fauna δ³⁴S and δ¹³C.

https://doi.org/10.1371/journal.pone.0299958.g008

6.4.) Tooth enamel stable carbon and oxygen isotope results

In addition to bone collagen and apatite, stable carbon isotopes were also analysed in the tooth enamel (δ¹³C_en) of 31 teeth, from 22 individuals (9 individuals were selected for multi-tooth analysis) in order to assess whole diet inputs during childhood. These values are all given in Table 5. These results yielded a mean δ¹³C_en value of -11.9 ± 1.0‰, ranging from -14.0‰ to -9.7‰. Of the individuals for which multiple teeth were analysed, the differences in δ¹³C_en values ranged from 0.2‰ to 2.1‰, but no consistent pattern could be identified of either increasing or decreasing values δ¹³C_en amongst individuals that might indicate changes in diet during childhood. Stable oxygen isotopes were analysed to assess the mobility of the individuals. A mean δ¹⁸O_DW value of -5.7 ± 1.0‰ was calculated from the tooth enamel of these individuals, ranging from -8.6‰ to -4.4‰. Concerning the individuals for which two teeth were sampled, the M1 was more enriched than the M3 in only two cases (2334-F and HS-1584), with a difference of +0.1‰ and 0.3‰, respectively. For the other individual in which both an M1 and M3 were sampled, the M1 had more depleted δ¹⁸O_DW values (HS-2079: M1 = -5.8‰, M3 = -5.0‰). If all δ¹⁸O_c values for M1s were corrected by -1‰, a conservative estimate using the breastfeeding enrichment range of 0.5–1.2‰ given in [132, 133] and the δ¹⁸O_DW values recalculated, they would be as follows: 955-H = -9.0‰; HS-1584 = -7.3; HS-1092 = -6.4‰; HS-639 = -6.1‰; HS-2079 = -7.4‰; HS-559 = -7.5‰; 2334-F = -6.7‰ and HS-1976 = -6.9‰.

Download:

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

Table 5. Results-tooth enamel.

https://doi.org/10.1371/journal.pone.0299958.t005

6.5.) Tooth enamel radiogenic strontium isotope results

The ⁸⁷Sr/⁸⁶Sr ratios of the plants collected from various lithologies around Santarém are provided in Table 6. The estimated local bioavailable ⁸⁷Sr/⁸⁶Sr range, based on these plant samples as well as the plant samples (PT16-104; PT16-109; PT16-110) collected close to Santarém in a recent study [61] is considered as 0.709–0.716 for the purpose of this study. The ⁸⁷Sr/⁸⁶Sr ratios obtained from the human tooth enamel, given in Table 5 gave a mean value of 0.710098 ±0.001259, ranging from 0.708103 to 0.712659. When the plant sample ⁸⁷Sr/⁸⁶Sr ratios are grouped according to the geological substrate from which they were sampled (see Figs 2, 9 and Table 3), the human range shows consistency with food sourced on the alluvial plains, the Pleiocene plateau (P2) upon which Santarém is located and the Miocene limestone outcrops to the west of the Tagus river. The plants sampled from the sandstone, clay and conglomerate outcrops (Q/P3/MP) to the east of the river, however, yielded ⁸⁷Sr/⁸⁶Sr ratios higher than any of the humans, therefore they are unlikely to have consumed high proportions of food grown in this region. A few individuals also yielded ⁸⁷Sr/⁸⁶Sr ratios lower than 0.709 which is below the measured regional range.

Download:

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

Fig 9. Human ⁸⁷Sr^/86Sr results compared to bioavailable ⁸⁷Sr^/86Sr for Santarém (plant ash samples from this study and [61].

https://doi.org/10.1371/journal.pone.0299958.g009

Download:

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

Table 6. Results-Plant ash.

https://doi.org/10.1371/journal.pone.0299958.t006

6.6.) Ancient DNA results

Ancient DNA analysis was performed on a subset of individuals to investigate aspects of their biological ancestry, in particular whether uniparental markers could inform on likely ancestral origins outside Iberia. These results are provided in Table 7.

Download:

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

Table 7. Ancient DNA results.

https://doi.org/10.1371/journal.pone.0299958.t007

The proportion of endogenous DNA estimated for this set of samples varied between 0.09% - 6.41% for the shotgun sequencing data. Only three individuals (1092-F, 1647-F and 955-H) passed quality thresholds to undergo genetic analysis due to poor DNA preservation. Mitochondrial DNA (mtDNA) haplogroups were identified for the individuals determined to be female, HS-1584 (H3ap) and HS-931 (J1b1a1), and the male 2334-F (HV24). In the latter case, we also identified the Y-chromosome haplogroup J1. The mtDNA haplogroups provide information about the maternal lineages while Y-chromosome haplogroups reflect paternal lineage. The mtDNA haplogroup H3 is found throughout Europe and in the Maghreb and has been found in Neolithic sites in Portugal. Its particular prevalence in populations in Andalusia and Morocco is thought to be a result of prehistoric migrations between these regions across the Straits of Gibraltar [134–136]. It has a postglacial or early Holocene origin, ~9500 ya in Iberia and ~8100 ya in North Africa, probably as a consequence of old introgression from Iberia to North Africa coinciding with largescale population resettlement within Europe [137]. MtDNA haplogroup J1b1a1 has an estimated Neolithic age of 9175 ± 3092 years and is widespread in present-day Near East, North Africa, the Caucasus and northern and central Europe, having likely spread during postglacial migrations [138, 139]. Concerning the male individual, 2334-F, the mtDNA haplogroup HV (excluding subclades H and V) is not particularly common among present-day Europeans (0–10%), being extremely rare in Iberia and reaching the highest frequencies in southern Italy. This haplogroup is also found at high frequencies in Iran and the Middle East [140]. The same individual harbours the Y-chromosome haplogroup J1, which has distinctly Middle Eastern origins, and a high incidence in North Africa. Specifically, this haplogroup reaches ~40–80% frequency across the Arabian Peninsula and ≥ 30% in Tunisia, Algeria and Morocco. In contrast, it has an average frequency of ~5% in Europe and <10% in Iberia.

6.7.) Reconstruction of solar position results

Using the online model SunCalc [121], it could be observed that in January, or midwinter, the sun rises to the east south-east of Santarém, and could be used to estimate the direction of Mecca quite accurately. In June, or midsummer, the sun rises to the north-east of Santarém but sets to the north-west, so if 180° was calculated from the setting sun in summer, a relatively accurate estimation for the direction of Mecca could also be achieved. In Spring and Autumn, however, the sun rises approximately in the east and sets approximately in the west of Santarém which would reduce the accuracy of predicting the direction of Mecca, regardless of whether the sunrise or sunset was used.

7.1) Grave orientation

7.1.1.) Grave orientation: Patterns.

Medieval Islamic burials are typically sparse by nature. While it is generally accepted that the Muslim conquest of Iberia brought with it a largely Berber population along with an Arab ruling elite class [76, 141], the paucity of grave goods leaves no indication of wealth or social status for those buried at the time. Indeed, it is only in exceptional cases that status can be inferred from medieval Islamic burials at all, for example in São Jorge Castle, Lisbon, where the location of burial was strong evidence of importance [20]. Although it is not possible to distinguish the status of the individuals from Avenida 5 de Outubro, the varying orientations of the graves themselves provides a different parameter within which to consider the above data. As explained in section 4.1.1., the individuals were grouped according to grave/skeleton orientation for the purpose of discussing the data. It should be noted, however, that although three distinct groups could be observed, it is not clear if the unorthodox burials (i.e. Group 1 and Group 2) are meaningfully separate from each other, or just differ as a whole from the canonical burials (Group 3). To this end, results are discussed with consideration to each of the groups and are also represented accordingly in the Figs 10 and 11, but in the latter figure, Group 1 and Group 2 are also represented as a potential combined group of unorthodox burials (Group 1 + 2).

Download:

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

Fig 10.

Human (a.) bone collagen (δ¹³C_col) vs bone apatite (δ¹³C_ap) and (b.) bone collagen (δ¹³C_col) vs tooth enamel (δ¹³C_en), by grave orientation (plotted against C₃ protein regression line [131] indicating main protein and energy sources in diet).

https://doi.org/10.1371/journal.pone.0299958.g010

Download:

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

Fig 11. Isotope results according to grave orientation.

a.)δ¹³C_col, b.) δ¹⁵N, c.) δ³⁴S, d.) δ¹³C_en, e.) δ¹⁸O_DW and f.) ⁸⁷Sr/⁸⁶Sr. M1 teeth are indicated by triangles in 9e.

https://doi.org/10.1371/journal.pone.0299958.g011

Considering the carbon isotope data, at first look (see Fig 7) it appears that the whole diet inputs for this population in both childhood and adulthood are similar mixes of mainly C₃ with some C₄ inputs, but when the results are framed in terms of the grave orientation of the individuals, a clearer picture emerges. When the δ¹³C_col and δ¹³C_ap is compared, there are no clear differences between the individuals with Group 1, 2 and 3 graves, as seen in Fig 10A. However, when the δ¹³C_en is compared to the δ¹³C_col in Fig 10B, the range of δ¹³C_en values is wider for the ‘Group 1’ individuals, over 4‰, than for the ‘Group 3’ individuals at 0.8‰. A Kruskall-Wallis H test identified a significant difference between the three groups in the mean ranks of the δ¹³C_en values (H = 10.795; p = 0.0045) and a Post-hoc Dunn’s test with a Bonferroni correction identified a significant difference between the following pairs: Group1 + Group3 and Group2 + Group3. A significant difference was also found in the mean ranks of the δ¹³C_col values (H = 6.365; p = 0.0415; Post-hoc Dunn’s test with a Bonferroni correction: Group 1 + Group 2) but considering that all the δ¹³C_col values except for HS-921 reflect typical C₃ plant consumption, this latter difference is not thought to be meaningful in a dietary sense (see S3 Text) [122]. With the exception of one individual from Group 1 (2334-F M1: -14‰), all Group 1 and Group 2 individuals had δ¹³C_en values that were more ¹³C-enriched than Group 3, indicating a higher proportion of C₄ plant input in childhood diet, or at least had more varied diets than the individuals who had Group 3 graves, a pattern that seems to disappear in adulthood based on the bone δ¹³C_ap values for all groups. In fact, 2334-F also had a later formed tooth sampled (M3) that was also more ¹³C-enriched (-11.9‰). Due to its high nutritional value, in many African cultures millet porridge is eaten by pregnant and lactating women and children [85]. While millet was consumed in Europe prior to the Muslim conquest, it is possible that the higher δ¹³C_en values of Groups 1 and 2 reflect either a food source preference or cultural practice, of consuming millet or sorghum porridge perhaps, that could be an indicator of individuals spending their childhoods in regions where this practice was more common or that the tradition persisted among some of the Muslims living in Santarém. In contrast, in Group 3, individuals consumed less to no C₄ plants in early childhood.

The δ¹⁸O values in the tooth enamel also display a distinct pattern when considered in terms of grave orientation of the individuals. While the δ¹⁸O_DW values of the Group 3 individuals all fall between -4.4‰ and -5.8‰, which is consistent with modern precipitation in the region around Santarém, paleoclimatic reconstructions have indicated that during the climatic periods of the Dark Ages (DA: 500–900 AD) and the Medieval Climate Anomaly (MCA: 900–1300 AD), conditions were warmer and drier over the southern Iberian Peninsula, resulting from a positive North Atlantic Oscillation (NAO) phase [74, 75, 142, 143]. Given the regional aridity during this time, one could expect less negative δ¹⁸O_DW values in individuals who spent their childhood in Santarém or nearby. Some of the teeth (1/9) sampled from Group 2 and 5/14 from Group 1 reflect δ¹⁸O_DW values that are more negative than -6‰, illustrated in Fig 11E and one female individual (HS 1442) had values as low as -8.4‰ in her M2 and -8.6‰ in her M3. These values indicate possible origins in a region with more humid conditions than Santarém at this time but could be consistent with precipitation δ¹⁸O values in the wetter mountainous regions of the Maghreb [53, 54, 69, 70]. Origins in the Arabian Peninsula seem less likely, as it is very arid and less negative δ¹⁸O_DW values may be expected. It should also be considered that precipitation δ¹⁸O values in the less humid areas of the Maghreb also overlap with expected δ¹⁸O values around Santarém [69], so it is not only the individuals with the most negative δ¹⁸O_DW values that may have migrated from North Africa, but it is not possible to distinguish the others using only oxygen isotopes as a proxy.

Overall, the relative homogeneity of Group 3 in both the δ¹⁸O_DW values and the δ¹³C_en values in their teeth seems to indicate similarity in their drinking water sources and in whole diet plant inputs, which would be consistent with a local population in Santarém who sourced water locally and ate a predominantly C₃-based diet in childhood. In contrast, the Group 1 and Group 2 individuals display more heterogeneity overall (see Fig 11D). While this pattern did not prove to be significant statistically in the δ¹⁸O_DW values (see S3 Text) [122], it is still present practically, with a min-max difference of 4.2‰ in Group 1+2 compared to 1.4‰ in Group 3. The large geographic variability for precipitation δ¹⁸O_DW values, along with the considerable overlap across the regions of interest, undoubtedly makes it very difficult to distinguish which individuals may have been non-local. These isotope values can be interpreted as a general pattern of mobility and behavioural differences amongst the individuals buried with different orientations i.e., some of the Group 1 and Group 2 individuals moved from a more humid region, possibly the wetter parts of the Maghreb, to Santarém, and both of these groups consumed more C₄ plants in childhood, perhaps sorghum or millet porridge, as is still a popular staple food throughout North Africa. If these individuals were indeed Berber settlers, they may have migrated to Iberia from different parts of North Africa, which would explain the more heterogenous isotope values of these two groups compared to the relatively homogenous Group 3.

Regarding the aDNA, it is important to consider that the North African gene pool has been formed by Palaeolithic and Neolithic back-migration of several European lineages, with Berber populations in particular being very genetically heterogenous [135, 144, 145]. The mtDNA haplogroups identified in three individuals from this population are found in Europe but also relatively frequent in the Maghreb. In fact, the H3 and HV haplogroups, which are found in mean frequencies of around 18% and 5%, respectively, in Mediterranean Europe, are within the same range in the Maghreb (13% and 7%, respectively). These three individuals belonged to Group 1 (2334-F and HS-931) and Group 2 (HS-1584) and their maternal lineages do not exclude Berber origins, in fact when combined with the overall previously mentioned patterns of data for these groups, it seems even more likely. Taking into account the clear difference in the M1 δ¹³C_en value for 2334-F, compared to the other Group 1 and Group 2 individuals, as well as his apparent Middle Eastern genetic lineage (see section 5.6), it is plausible that his tooth enamel reflects early childhood dietary practices that were more influenced by Arab culture, in which many of the most popular dishes at the time made use of wheat and/or barley e.g. tharīd, a bread stew [146], rather than North African Berber influences which may have included higher consumption of C₄ plants. It is clear though, that C₄ plant consumption by this individual increased during the time frame of M3 enamel formation.

The ⁸⁷Sr/⁸⁶Sr ratios of this population showed an overall consistency with the bioavailable ⁸⁷Sr/⁸⁶Sr range, as previously mentioned, with the exception of some individuals who had ⁸⁷Sr/⁸⁶Sr ratios below 0.709. When considered by grave orientation, there are no clear differences between the groups (see Fig 11F). It should be noted, however, that the consistency of human ⁸⁷Sr/⁸⁶Sr ratios with measured bioavailable ranges does not serve as absolute confirmation of a local origin, but only as evidence of provenance in a geologically similar region. In fact, ⁸⁷Sr/⁸⁶Sr ratios measured in studies across the Maghreb have also yielded similar ranges, for example 0.710–0.712 in human tooth enamel from southwestern Libya [147], 0.708–0.710 in plant ash from northern Tunisia [70] 0.7078–0.7096 in waters from the Continental Intercalaire (CI) aquifer that flows from the Atlas Mountains in Algeria to the Tunisian Chotts [148] and 0.7093–0.7105 in groundwater from the Lakhssas Plateau in Morocco [149]. This means that Berber populations who migrated from the Maghreb and settled in Santarém would not necessarily have very distinct ⁸⁷Sr/⁸⁶Sr ratios from individuals who had spent their entire lives in the vicinity of Santarém.

Another potential indicator of mobility is sulphur. Fig 12 compares the Santarém human and fauna δ³⁴S results to the data from the sites of Tomar (11^th-17^th C) [19] and Évora (12^th-13^th C) [22] along with the predicted range [47] and demonstrates that the Santarém population is highly variable in comparison to both Tomar and Évora. With the exception of two ovicaprids (FVRP3; FVRP9) and the fox (FVRP5), the fauna fell within the predicted range while many of the humans had δ³⁴S values below 12‰. It is plausible that the human δ³⁴S values are highly variable because the time during which the individuals lived (and when the δ³⁴S was incorporated into their bones) was during the period in which high Saharan aeolian input deposited sulphur from northern Africa in Iberia [143]. This potentially would have caused fluctuating atmospheric and soil δ³⁴S but the extent of this influence is unclear

Download:

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

Fig 12. Human and fauna δ³⁴S from Santarém (this study), Évora [22] and Tomar[19] along with expected regional range [47].

https://doi.org/10.1371/journal.pone.0299958.g012

The δ³⁴S values, were the most heterogenous in the Group 3 individuals (ranging from 4.9‰ to 16‰), while the Group 1 (ranging from 8.8‰ to 13.9‰) and Group 2 individuals had more homogenous δ³⁴S values (ranging from 7.4‰ to 11.2‰) (see Fig 11C). This pattern is in contrast to the δ¹⁸O_DW and δ¹³C_en data, but when Group 1 and Group 2 are considered together, this pattern mostly disappears, and the range is similar to that of Group 3. It is important to reiterate that these δ³⁴S values were measured in the bone collagen and reflect sulphur incorporated into the bones over a period of years before the death of the individual while the tooth enamel is formed over a much shorter space of time and a much earlier period of the individual’s lives. It is possible that diverse dietary input in the years prior to death might also explain such high variability and the fact that some human δ³⁴S values are lower than the modern expected range for Santarém. If this is indeed a dietary pattern, it is not apparent in the δ¹³C_col and δ¹⁵N values, adding an additional layer of complexity to the interpretation of this data.

While δ³⁴S may not have provided clarification regarding the mobility of the Santarém population, the δ³⁴S values obtained are still an important contribution in a broader sense because there is so little published data for medieval Portugal, and no studies (to date) have provided δ³⁴S values for a human population living during the Late DA (650–900 AD) and during the Early MCA (900–1100 AD) in Portugal. This was a notable climatic event, so making δ³⁴S data available for this region during this time period may prove useful to future studies.

7.2.) Sex

Excluding the results of the skeletons classified as ‘Maybe F’, ‘Maybe M’ and ‘U’, the data was assessed for patterns in terms of sex. There were no clear differences in the δ¹³C_col and δ¹⁵N according to sex, with males (n = 14) having a mean value of -18.8 ± 0.6‰ in δ¹³C_col and 10.3 ± 0.8‰ in δ¹⁵N, while females (n = 20) had values of -19.0 ± 0.4‰ and 10.3 ± 1.0‰ for δ¹³C_col and δ¹⁵N, respectively (Fig 13). The δ³⁴S values for males (n = 10) were slightly higher (δ³⁴S = 11.9 ± 2.1‰) than females (n = 18; δ³⁴S = 10.7 ± 1.7‰) but were not significantly different. This indicates that adulthood diet was relatively similar amongst males and females in the Early Islamic Period in Santarém.

Download:

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

Fig 13. Human and fauna δ¹⁵N and δ¹³C_col by sex.

Squares represent expected range of +1 trophic level increase (+0–2‰ in δ¹³C_col; +3–5‰ in δ¹⁵N) over mean Bos and Ovicaprid values, respectively.

https://doi.org/10.1371/journal.pone.0299958.g013

Considering the tooth enamel results for δ¹³C_en and δ¹⁸O_DW, the males (n of teeth = 10, from 7 individuals) have slightly more variable carbon (δ¹³C_en = -12.0 ± 1.3‰) than the females (n of teeth = 21, from 15 individuals; δ¹³C_en = -11.9 ± 0.8‰) while the females show some more variability in their oxygen (δ¹⁸O_DW = -5.9 ± 1.2‰) than the males (δ¹⁸O_DW = -5.1 ± 0.4‰). While it is true that the sample size of teeth was larger for females than for males, the number of individuals with teeth available for sampling was proportionally low and more females had teeth available for sampling than males. Although we cannot assume that this small number of samples is fully representative of the whole Muslim population under study, it is nevertheless interesting that all of the δ¹⁸O_DW values below -6‰, i.e., possibly reflecting origins in a more humid region than Santarém, belong to five female individuals (1442-F, 955-H, HS-2242, HS-1027, HS-2128). The last belongs to Group 2 while the other four were in Group 1 graves. The M1s, for which δ¹⁸O_c values were corrected by -1‰ and the δ¹⁸O_DW values recalculated, are represented by the hollow triangles in Fig 14, which demonstrates the distribution of δ¹⁸O_DW values in terms of both sex and grave orientation. Although it is just a conservative estimation [132, 133], if this correction of -1‰ were a true reflection of breastfeeding enrichment in the M1s, it would mean that most of the individuals have δ¹⁸O_DW values out of the expected precipitation range for Santarém, with the most negative values belonging to the females.

Download:

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

Fig 14. Human δ¹⁸O_DW with multi-tooth comparison.

Individuals with values outside of local range and/or of whom multiple teeth were sampled are in colour and labeled. Individuals within local range and with only one tooth sampled are in grey. Solid triangles = M1s, circles = M2s, squares = M3s, cross = PM1 and diamond = PM2. Hollow triangles represent hypothetical M1 values corrected (-1‰ from δ¹⁸O_c VPDB) for breastfeeding.

https://doi.org/10.1371/journal.pone.0299958.g014

Taking the ⁸⁷Sr/⁸⁶Sr ratios into account too, the females were also slightly more variable than the males. Bioavailable ⁸⁷Sr/⁸⁶Sr ratios below 0.709 were not obtained for any of the geological substrates surrounding Santarém (see Fig 15) but the three individuals who had tooth enamel ⁸⁷Sr/⁸⁶Sr ratios below 0.709 were all female. Although this difference may appear small, it would also be consistent with examples of the ranges of ⁸⁷Sr/⁸⁶Sr in North Africa that were previously discussed. While tentative, the consideration of δ¹⁸O_DW and ⁸⁷Sr/⁸⁶Sr together, as demonstrated in Fig 16, would indicate a pattern of more sedentary males and more widespread origins in the females. Although it is necessary to interpret this data with caution, due to the small number of tooth enamel samples, particularly for the males, it could support the argument for a patrilocal social system whereby the males were more sedentary, and females relocated later in life, possibly upon marriage.

Download:

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

Fig 15. Human ⁸⁷Sr^/86Sr results by sex compared to bioavailable ⁸⁷Sr^/86Sr for Santarém (plant ash samples from this study and James et al.

2022 [61]).

https://doi.org/10.1371/journal.pone.0299958.g015

Download:

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

Fig 16. Human ⁸⁷Sr^/86Sr and δ¹⁸O_DW results by sex.

Grey lines indicate the limit of the expected ‘local’ range for the respective isotopes in Santarém. For ⁸⁷Sr^/86Sr, bioavailable range is determined from plant ash samples from this study and James et al. 2022 [61].

https://doi.org/10.1371/journal.pone.0299958.g016

Furthermore, three of the female individuals (955-H, HS-931 and HS-1584) had cranial morphometric features, such as maxillofacial prognathism, broad nasal apertures, smooth inferior nasal margins, retracted chins and rectangular occipital orbits, that are more prevalent in pre-modern populations with African genetic ancestry than Eurasians [102, 103]. While it is true that these characteristics could be retained in the Muslim population for many generations after migrating to Iberia, the δ¹⁸O_DW value in the M1 of 955-H suggests that at least one of these individuals may have spent their childhood in a different region, hence could have been a first generation migrant to Santarém. The mtDNA results for HS-931 and HS-1584 also support maternal lineages that are frequent in the Maghreb and Arabian Peninsula (haplogroups H3 and J1b respectively). Genetic diversity is higher amongst females than males in Berber populations, due to patrilocality [144, 153], so if this practice continued in the Santarém population it could explain the previously mentioned isotope data patterns for the females as a whole.

The male individual, 2334-F, appears to have Arabian genetic origins, certainly in paternal lineage and probably maternal lineage too. Genetic studies on modern populations in the Maghreb found that J1 (Arab) lineages comprise 17% of the modern male Tunisian gene pool and the Arab male genetic input from 7^th century expansions is thought to be as high as 38% in Tunisia, and also high in Algeria [144, 153]. Considering that the Arabisation of the Maghreb, and the Islamic conquest of Iberia were military ventures, most likely driven by Arab men displacing native Berber men but marrying Berber women, in the case of the former, it makes sense to have females with apparent European/Berber genetic lineages and a male with Arab genetic lineage. As discussed in section 7.1.1, the early childhood diet of this male, 2334-F, is also distinct from the other Group 1 and Group 2 individuals, which could be a consequence of different food practices related to cultural influences (Arab/Berber) if his genetic lineage is a clue to his lived experience. It is impossible to speculate broadly based on three individuals, but the isotope results of the rest of the Santarém population has also been informative. There are records of Muslim men marrying local non-Muslim women who subsequently converted in the years following the conquest of Iberia, for example the widow of King Roderic, Egilona, who married the Berber Abd al-Aziz, son of the conquering general Musa bin Nusayr [154]. This was likely to have been a relatively common practice as it consolidated Muslim authority, ensured legitimacy over inherited lands, and was a vital element for driving the process of social and cultural change in Iberia following the conquest [150, 154]. While this may indeed have been the case for some of the females from Santarém, whose δ¹⁸O_DW values and ⁸⁷Sr/⁸⁶Sr ratios are consistent with expected local ranges, there are at least some (previously mentioned) females who appear to have migrated from a region with higher precipitation and/or possibly lower ⁸⁷Sr/⁸⁶Sr ranges, in the cases of HS-1584, HS-559 and HS-956, for example the Atlas Mountains region of Morocco and Algeria.